The Peter Attia Drive - Cancer screening with full-body MRI scans and a seminar on the field of radiology ｜ Rajpaul Attariwala, M.D., Ph.D. (#61 rebroadcast)

00:00:00.000 Hey, everyone. Welcome to the drive podcast. I'm your host, Peter Atiyah. This podcast,

00:00:15.500 my website, and my weekly newsletter all focus on the goal of translating the science of longevity

00:00:19.820 into something accessible for everyone. Our goal is to provide the best content in health and

00:00:24.780 wellness, full stop. And we've assembled a great team of analysts to make this happen.

00:00:28.900 If you enjoy this podcast, we've created a membership program that brings you far more

00:00:33.320 in-depth content. If you want to take your knowledge of the space to the next level at

00:00:37.340 the end of this episode, I'll explain what those benefits are. Or if you want to learn more now,

00:00:41.740 head over to peteratiyahmd.com forward slash subscribe. Now, without further delay,

00:00:47.780 here's today's episode. Welcome to another special episode of the drive. For this week's episode,

00:00:54.280 we're going to rebroadcast my conversation with Raj Atariwala, which was released back in July 2019.

00:01:02.160 Raj is a dual board certified radiologist and nuclear medicine physician boarded in both Canada

00:01:07.720 and the United States. He's the co-founder of Prenuvo and the medical director at AIM Medical

00:01:13.320 Imaging. Over the past decade, he has been effectively creating a new way of doing MRI by fine tuning the

00:01:20.380 hardware and building unique software to create a completely revolutionary product and process by

00:01:25.480 which to look at the body using the technology of magnetic resonance. While I think this is a bit

00:01:30.760 of a technical episode, I also think it's one that anybody who's ever had an x-ray, a CT scan,

00:01:36.640 an ultrasound, or an MRI in their life needs to listen to. You can divide this episode sort of

00:01:41.220 into two halves as follows. The first half is the history of radiology. So we start with talking

00:01:45.260 about what an x-ray is, how it works, what the radiation does and doesn't mean, how a CT scan

00:01:51.880 became an evolution of that, ultrasounds, PET scanners, nuclear medicine scans, and all these

00:01:55.740 things. We've done this in the way that is really geared towards you, the patient. I think we do a

00:02:01.300 pretty good job of always bringing it back to language that makes sense, and we don't get terribly

00:02:05.720 lost in the physics of these intricate machines. The second half of the episode is a real deep dive

00:02:11.500 around cancer screening and the use of a particular type of MRI technology that Raj has played an

00:02:18.060 enormous role in developing. What I enjoy about this episode and why I think it's an important

00:02:22.740 rebroadcast is my attention and interest in cancer screening has only grown deeper in the last four

00:02:30.360 years since this episode was originally released, and I still find myself having discussions with patients

00:02:35.880 on a nearly weekly basis about the importance of MRI for screening and its limitations. So without

00:02:42.660 further delay, please enjoy or re-enjoy my conversation with Raj Atariwala.

00:02:53.060 Hey Raj, thanks for carving some time out today in the middle of your busy day. I've probably done what

00:02:58.180 no one else has done before, which is shut down this clinic, huh? No, it's fantastic to see you again and

00:03:03.340 to be here. It's a pleasure as always. What's the deal with Vancouver and Uber? I got off the plane

00:03:08.680 this morning and tried to get an Uber to come here, and apparently there's no Uber in Vancouver.

00:03:15.100 It's stunning. Everybody complains. We all complain, but I don't know, somebody somewhere is

00:03:20.180 not allowing it. It was amazing. I was in India, and you can get Uber in India.

00:03:25.860 Yeah, and it's not a Canadian thing because I Ubered in Toronto.

00:03:29.220 Toronto, Calgary, almost all the country, just Vancouver.

00:03:31.900 Yeah, very well. I'll reserve my editorial comments for myself. Well, I've been looking

00:03:36.520 forward to this for a long time, Raj. We've known each other for about four years. I think we were

00:03:39.860 introduced through a mutual acquaintance who's a good friend of mine and has become very interested

00:03:45.180 in your technology. And I'm not even sure if you remember the context, but the context was

00:03:50.440 basically this person reaching out to me to say, hey, there's this really fancy MRI scanner up in

00:03:58.140 Vancouver. Can you go check it out for me? And at the time, I was focusing a lot on different

00:04:07.900 technologies that might be able to aid in the detection of atherosclerosis. And he may have

00:04:13.820 misunderstood what I was interested in, but it was sort of pitched to me through that lens.

00:04:18.420 You and I hopped on a call. I still remember where I was sitting actually in my office at the time.

00:04:23.240 It was probably, yeah, it was in January. By the end of that call, I was really interested. And you

00:04:28.600 said, look, I mean, I think you should just come up to Vancouver, get scanned and let's spend a day

00:04:32.420 discussing it. And the rest is history. So, you know, what I wanted to do today was obviously talk

00:04:37.240 a lot about AIM, which is the, well, I guess it's the name, is that the name of the company or is that?

00:04:43.320 Yeah. So actually what I did is I actually set up AIM as a private MRI company. And we basically put

00:04:48.460 an MRI machine so I could play with it. That's one of the problems of being an engineer. And so

00:04:53.820 AIM is the name of the scanning company, the clinic, but we've actually now kind of moved it

00:04:57.900 into PreNuvo. And what PreNuvo is about is actually to basically sort of put the power of preventative

00:05:03.000 medicine into patients' hands. Yeah. Well, we'll talk a heck of a lot about this, but let's talk a

00:05:08.140 little bit about your background. Cause that was one of the things that right off the bat made me

00:05:12.120 realize that this was going to be an interesting discussion. I, even in medical school, just took a huge

00:05:17.100 interest in radiology. I never thought for a moment I would become a radiologist, but I knew that

00:05:22.000 whatever type of medicine you practice, you have a choice, which is basically, you can just be

00:05:26.860 confused and intimidated by all of these scans that your patients get, or you can at least try to

00:05:32.360 understand them and have some hope of appreciating the risks of them, the benefits of them, the

00:05:38.140 subtleties, et cetera. So when I did my radiology rotation, I was probably the most eager student who

00:05:44.840 wasn't going into radiology. And I remember, and I still have my notes, but I took, I mean,

00:05:50.760 maybe 50 pages of notes, drawing coils and all sorts of things. So let's start at the beginning.

00:05:57.900 You have a background in engineering, which a lot of radiologists seem to have, right?

00:06:01.980 It's actually common in radiology. It's like, basically there's a lot of technology in radiology.

00:06:06.620 And so as a result, it attracts those of us who actually are technophiles. And so my background,

00:06:12.040 and actually I started out in chemical engineering, and then during that period of time, kind of

00:06:16.960 realized that I actually kind of liked the body and physiology and how that works. And as a result,

00:06:21.360 I then went into biomedical engineering, where I did my master's and PhD in biomedical engineering in

00:06:26.240 Northwestern. During that period of time, I was actually working on fluid mechanics. We're actually

00:06:31.320 looking at blood flow and hemodynamics and all sorts of complicated things. And we also actually did what

00:06:36.460 engineers do. We actually had all these manual systems that we're using to measure blood flow

00:06:41.500 and blood pressure. And we decided, okay, let's build a robot to do it instead. So we did.

00:06:46.900 And this robot that we actually built allowed us to do keyhole surgery in the eye that wound up

00:06:52.120 attracting a lot of attention from physicians, from top tier universities all over the United States,

00:06:57.700 from mass general everywhere. And as engineers, the attitude was, we can build anything you want.

00:07:03.300 What do you want? A lot of times physicians can never answer that. I was working with a lot of

00:07:08.160 these top tier guys, like editor in chief of the different medical journals and ophthalmology.

00:07:13.640 And it was like, they were speaking a different language and I just didn't understand what they

00:07:17.600 were talking about. And so I actually kind of decided, okay, well, let me apply to medical school

00:07:23.500 and see what happens. So I can actually kind of learn this language and kind of learn more about

00:07:27.840 medicine. And so I applied and then I actually kind of got accepted. And I was like,

00:07:33.140 do I want to do this? It's like, I'm really an engineer and was born an engineer. So my PhD

00:07:39.120 advisor's famous last words were that the engineering world will always take you back.

00:07:44.040 So I went off to medical school and actually hated every minute of it. It was kind of like,

00:07:48.440 I need to know more. I'm one of those kids who's kind of like, why, why, why, why, why?

00:07:53.060 How does this work? Explain this to me. I don't get it. And realize that there's a lot of things in

00:07:57.720 in the body and physiology and pathophysiology that we just don't understand. Despite the massive

00:08:04.280 amount of literature that's out there in the medical world, you really couldn't find a lot

00:08:07.860 of the answers of why things go wrong and how they go wrong. So then I'd wind up sort of exploring,

00:08:13.080 okay, how do we advance understanding of what's going on in the body? Like the simple aging process,

00:08:18.240 what happens? Why does everything change? Why? People can answer it. Instead, it was like,

00:08:23.120 okay, memorize this list of changes. I'm like, okay, great. Let me try and remember all that.

00:08:29.880 And what I would actually start to boil down to is that I would actually go back to my engineering

00:08:34.340 pathophysiology texts and I actually read them and talk to the PhD guys. And they would actually

00:08:39.960 sort of give me the theories on what they thought was happening. And when you actually got that theory,

00:08:44.820 it was almost like planting a seed. Then you actually kind of understood how the entire tree would

00:08:48.560 look. And that's when I said, okay, maybe this is good, but I still need my technology.

00:08:53.120 Where is technology going? We worked on some of the very first surgical robotic machines ever

00:08:58.120 built. My colleagues presented the first telerobotics telepresence conference ever held

00:09:03.760 in the United States. At the same time, the group that made the Da Vinci was there. I kind of looked

00:09:08.420 and said, there's not a lot of space for robotics because people don't want machines operating on

00:09:13.900 them. As a former surgeon, I'm sure you probably feel the same way. You don't want a machine operating

00:09:17.480 on you unless it's going to be better. So that's when I decided, okay, the technology that people

00:09:22.280 understand, doctors understand is a picture. And that picture is radiology. And that's really

00:09:26.880 where all the technology is. And so I actually started in an area called nuclear medicine,

00:09:31.120 which is a sort of a small specialty within radiology, which is where you're actually looking

00:09:35.780 at functional imaging, how things work, how do things change, what happens over time,

00:09:41.180 and really enjoy that area because it was sort of showing you what's happening when things are normal

00:09:46.640 and then when things become abnormal. And it was actually one of these other amazing fields that in

00:09:51.420 medicine, there's a lot of shades of gray, whereas in nuclear medicine, it's almost black and white.

00:09:55.480 It's there or it's not. It's one of these very few areas where you actually get a binary choice of

00:10:00.120 what's happening. Is there a problem? Yes or no. As opposed to, well, there might be. So that's actually

00:10:06.620 what I liked about it. But then I also kind of realized that radiology is basically very much like

00:10:10.800 anatomy. You actually see what's going on. You actually see the changes and you see the shapes of

00:10:14.780 things. And you use that very much as a blueprint for a building to actually kind of see what it

00:10:18.660 does. Whereas nuclear medicine is kind of, instead of the blueprint, you actually kind of know there's

00:10:23.580 all these people carrying letters moving in and out of this building. We don't really know the

00:10:26.920 detail of where the building walls are, but we know that there must be something happening there.

00:10:31.620 And hey, when you put the radiology with the architecture together of the blueprint of the

00:10:36.140 building, then you combine that with the people going in and out of it, you realize it's the post office

00:10:40.700 and you realize that these are postal workers carrying things. And here's the geometry of the

00:10:45.440 building. So that power really is actually quite useful in the fact that it's really that there's

00:10:50.680 this famous equation when people actually realized to put functional imaging or nuclear medicine

00:10:55.300 imaging together. The first device that did that was positron emission tomography and CT or PET-CT.

00:11:01.820 And the famous equation is that one plus one equals three. These two separate modalities of

00:11:06.220 functional imaging and anatomic imaging come together to actually make something better than

00:11:10.600 each part individually. And so that's what really attracted me to, to the whole field,

00:11:14.700 just because you could see what's going on and you can actually see the power of what's happening.

00:11:19.020 That was certainly one of the more powerful lessons I remember as a medical student when I

00:11:24.380 went sort of headlong into radiology, which was what I really liked about this rotation. I remember,

00:11:29.020 I wish I could remember the names of the residents, but they took me under their wing,

00:11:31.980 even though they knew I wasn't going to be a radiologist. And as you know, from being in medical

00:11:35.180 school, that's a little unusual. Typically the residents tend to gravitate to the students that

00:11:39.320 are going to follow in their footsteps. But I think they saw in me genuine curiosity and they thought,

00:11:44.300 well, look, the smarter we can make this surgeon, the better down the line for us radiologists.

00:11:49.820 And so I remember them taking me aside and saying, look, Peter, anytime you order a test in the back of

00:11:54.780 your mind, you have to be asking yourself, do you want anatomical information or functional information

00:11:58.940 or both? And the example you gave is a great one. And in the show notes for this podcast,

00:12:04.600 we'll link to tons of pictures so that people understand what is meant by anatomic imaging.

00:12:11.080 And the way I would explain this to a person is an anatomic image has nice sharp edges. It looks like

00:12:17.320 what it is you're trying to take a picture of. So the anatomic image of the brain shows all of the

00:12:23.440 substructures. And when the radiologist looks at it, he or she can make out every little blip and bend

00:12:30.520 and crevice inside of the brain. And they can comment on different structures. Well,

00:12:34.440 there's a tumor here, or there's a blood vessel that's slightly dilated there. If you contrast

00:12:39.160 that with the PET scan, as you pointed out, that's looking at a function of the brain, which is how

00:12:43.560 much glucose does it uptake? And instead it lights up darker, the more glucose that's being taken there.

00:12:49.720 So in your analogy, the CT scan is showing the architecture of the post office, but the PET scan

00:12:57.560 is showing you the distribution of people moving into different areas of it. And by putting it

00:13:02.360 together, that gives you a really powerful picture. Exactly. And that's actually exactly how it works.

00:13:07.000 And one of the things that might be useful as well, is if we kind of look at the different sort of

00:13:10.520 techniques that are used between nuclear medicine and radiology, we all know that imaging started with

00:13:15.560 the simple X-ray, but that was actually groundbreaking for the field of medicine.

00:13:19.880 Well, let's start with that. So everybody has seen an X-ray and I think most people,

00:13:25.480 if they can close their eyes and picture it now, or look at an image kind of know directionally that

00:13:31.880 an X-ray is nothing more than a series of contrasts going from black to white and everything in between.

00:13:39.880 So at the highest level, how do we take an X-ray? What are we doing with those little electrons

00:13:45.560 going through someone? How does it produce that image?

00:13:47.640 For sure. And it's quite amazing, actually, how it was discovered by Runt. And effectively,

00:13:51.240 what it is, is we're taking these high energy wavelengths and it actually penetrates right

00:13:55.960 through the body. And anywhere where there's something dense, very hard, like bone, the X-ray

00:14:01.400 beam can't make it through. We've all sort of taken like a flashlight and sort of shown it through

00:14:05.880 our finger and we can actually see the red light coming through. Well, effectively, that's what an X-ray is.

00:14:10.360 We're just taking higher energy wavelength that we can't see with our eyes. And it's actually allowing

00:14:15.080 the areas where there's things like air or soft tissues, the X-ray penetrates right through it and

00:14:19.720 goes right through, shines through. Whereas in places where there's bones, the X-rays can't make

00:14:24.040 it through. So they actually get left behind. And that's what gives us like the white pictures of

00:14:28.040 the bone because the film that's exposed on the other side, soft tissues, the X-rays have gone through,

00:14:34.120 they turn it from white to black. Whereas in bones, the X-ray doesn't make it through because it

00:14:38.600 gets left behind in the person. And therefore on the film on the other side, it stays white.

00:14:42.520 And back in the days of Rankin, did they appreciate the damage of ionizing radiation or were there

00:14:48.760 a number of casualties along the way from people being far too exposed to this type of energy?

00:14:53.800 Yeah. Unfortunately, there were lots of casualties and actually our understanding of radiation

00:14:58.120 has actually been moved by really traumatic events such as Hiroshima and Nagasaki and things

00:15:03.800 like that where there's been real radiation damage and we've actually been able to watch people over time.

00:15:08.040 And recently in the past 10 to 15 years in imaging, we've actually realized the danger of

00:15:14.600 this high powered ionizing radiation and how it can damage cells. And when the DNA of the cells get

00:15:19.880 damaged, there's a risk of inducing cancers with them. And so we're actually starting to understand

00:15:24.600 that more and more. And we're actually starting to see that a lot of patients, individuals are like,

00:15:28.600 look, I know about the potential damaging effects of X-ray radiation. Therefore, I don't want a lot of

00:15:33.800 X-rays, CT type scans. Now, the way I generally talk about this with my patients, I have a graph

00:15:39.400 that I show them that on the X-axis lists a number of different technologies. So a chest X-ray,

00:15:46.040 an abdominal X-ray, a mammogram, and they're generally in increasing amounts of radiation.

00:15:51.000 So at the top end of that spectrum, you'd have a whole body PET-CT just for, if you want it to go

00:15:55.640 as ionizing as possible. And then the Y-axis, I use these units called millisieverts. Can you explain

00:16:01.560 what a millisievert is? Exactly. So a millisievert is actually the unit of measurement of radiation.

00:16:06.600 And it's actually set by the standards group, System Internationale in France,

00:16:11.080 that actually sets what a standard dose of radiation is. And now it's actually important

00:16:15.800 that a lot of people really don't understand radiation. There's good radiation, there's bad

00:16:19.960 radiation, but radiation is anytime you actually have any ion that's actually releasing a component

00:16:25.560 of its energy, that energy has to get deposited somewhere. It actually comes off as a photon,

00:16:31.320 of energy that has to by, I guess, energy effects go from, it's never created, it's never destroyed,

00:16:39.320 but it's actually transferred. And so that energy, if it doesn't go through you, it's actually going

00:16:43.560 to deposit in you. And if it actually deposits in you, that's where it can actually cause the damage.

00:16:48.200 And now when we look at all the different types of imaging, as you talked about, on one end,

00:16:52.520 you have the mammogram, which actually has a very low amount of radiation in millisieverts.

00:16:57.160 It's usually about 0.05, which is quite negligible. Whereas on the other end of the scale,

00:17:02.920 you'll have the PET-CT. And now the PET-CT basically couples the radiation from the CT.

00:17:08.680 And the CT scanner is basically a powerful x-ray that's spinning around the body and creating

00:17:12.840 a three-dimensional view of you. Combined with the PET, which is the positron emission

00:17:17.400 tomography, where you're taking radioactive glucose and you're actually labeling with fluorine.

00:17:23.000 And that fluorine is a radioactive fluorine 18, which actually now gives off a positron.

00:17:27.880 And that positron has tremendous amount of energy. It's the highest energy that you can actually have

00:17:32.120 in imaging for radiation. And that's 511 kilo electron volts. So very, very high.

00:17:37.640 And so when you combine those two together, you wind up getting, typically, when we actually give

00:17:42.520 somebody radioactive glucose for a whole body scan for, let's say they have cancer, it's roughly one

00:17:48.440 millisievert per mega becquerel. But I guess I'm trying to convert the Canadian.

00:17:53.560 Oh, that's okay. Yeah. We've got an international audience ship here. You don't have to make it

00:17:57.160 Americanized. Right. So we do it in mega becquerels up here in Canada. And so it's

00:18:03.080 typically about 35 mega becquerels per millisievert. So typically somebody will actually get the US

00:18:08.840 dose would be about 12 millimoles. And so that's about the 12 millisieverts of radiation

00:18:13.400 in addition to what they get on the CT scan. So they could easily get 30, 40 millisieverts in total

00:18:20.040 in that scan if they were doing chest, abdomen, pelvis, for example. It's possible. Yeah. And a lot of

00:18:24.440 that radiation would come from the CT. Whereas with nuclear medicine, what we typically do,

00:18:28.280 since we're actually going to inject somebody with the radioactive material,

00:18:31.240 as a result, we've exposed them to that radiation. So all parts of the body. So as a result, as it

00:18:35.880 circulates through their entire body, we want to take pictures of everything from head to foot.

00:18:40.120 Because if we're going to give somebody radiation, we want to maximize the amount of

00:18:44.120 data we're going to extract from that exposure. Now, I don't know in Canada what the number is,

00:18:49.080 but in the United States, the NRC recommends that no one receive more than 50 millisieverts in a year.

00:18:55.880 But of course, not all of that can be assigned to radiography because living at sea level exposes

00:19:02.360 you to what, two millisieverts a year, maybe three. I mean, I think even if you live at elevation,

00:19:07.480 people in Denver are getting probably six or seven millisieverts a year at background. I'm

00:19:11.720 probably a bit off on that. It might be a bit less. That's right. Yep. So the higher you go,

00:19:15.080 actually get more cosmic radiation. And then as well, certain geographies will actually have radon,

00:19:20.680 which is another exposure to the millisieverts. And as well, when you actually travel, when you

00:19:25.880 actually go up in altitude in planes, we actually get a lot more radiation exposure. So that's why

00:19:31.800 pilots, for example, they actually, when they wear glasses, they actually have to block the UV

00:19:36.920 radiation, the UVA, UVB, and also the x-ray radiation, which is much more prevalent at higher

00:19:42.280 altitude. The route matters. I know I've calculated this for myself doing a lot of East West Coast

00:19:48.760 travel. Fortunately, not very much exposure. I believe it's less than 0.1 millisievert per round

00:19:54.920 trip. But if you do LA over the pole, right, all of a sudden it goes up by, again, I don't want to

00:20:01.240 misquote it because we have the data so I could just post it, but it goes up by a non-linear amount.

00:20:06.620 It's much more radiation when you cross the North Pole than just the extra distance you travel.

00:20:13.100 Exactly. And that's because of the ozone. The less ozone you have at the poles, the more

00:20:16.780 exposure you're going to get because the ozone actually absorbs the radiation.

00:20:20.780 So there's actually a calculator available online for people to actually determine how much radiation

00:20:24.900 they're getting from exposure. And it's actually required that pilots and flight attendants

00:20:28.580 calculate their dose.

00:20:29.900 We'll make sure we find that and link to that. I just, because the NRC says 50 is the limit, I've

00:20:35.960 never really thought of that as we should go up to 50. I've thought of that as they probably have

00:20:40.720 some reason to believe that successive years with exposure to 50 millisieverts is not a good idea.

00:20:46.060 Do we have a sense of what the implication of that is in terms of normal physiology?

00:20:50.380 We do and we don't. And actually a lot of the time, our understanding of that really comes from,

00:20:54.500 like I said, the tragedies of Hiroshima and Nagasaki, as well as other people like in

00:20:58.780 a Fuji reactor who actually got exposed. And that's where understanding of the damage comes

00:21:03.500 from. Now there's actually a landmark paper out of Columbia that actually looked at the

00:21:07.280 amount of radiation that people actually receive from CT scanners. And it actually forced radiology

00:21:11.960 as a field to actually look at the potential damaging effect of X-ray radiation. And what

00:21:16.880 they actually kind of found is that the younger you are, the greater the risk of cancer induction

00:21:21.360 from CT scanners, which is why in the pediatric world, we actually try and really minimize the

00:21:26.440 amount of dose that children in particular are getting. And the sex as well matters. So females

00:21:31.120 are actually more sensitive to radiation than men.

00:21:33.320 Meaning if you took a 20-year-old male and a 20-year-old female, so both in their reproductive

00:21:39.300 prime, are you saying that the ovaries of the woman are more sensitive to DNA damage in the egg

00:21:47.260 than the sperm are in the testes of the male?

00:21:50.820 Exactly.

00:21:51.420 I didn't know that. Why is that?

00:21:52.520 And that actually has to do with the fact that the egg was actually produced during embryonic

00:21:57.280 stage. And as a result, that DNA is effectively frozen in time. And so as women are getting

00:22:03.200 older and older, they're releasing these DNA in the eggs. So therefore, the younger you

00:22:07.400 are, the fresher, the DNA, whereas when they're prime sort of around 12, which is when this

00:22:13.200 DNA comes out of the, I guess, frozen state during the beginning of menstruation, that's

00:22:18.700 when these eggs start to be released. So actually 12 is sort of the worst time for females.

00:22:23.200 Which is really sad because of course, anybody who spent time in a hospital knows that there

00:22:26.760 are invariably kids that need to undergo radiographic studies. You only need to spend a few days in a

00:22:32.300 cancer ward to realize all these poor children that are right at that age and they're being

00:22:36.040 exposed to it. And unfortunately, there's not much of a choice. And trauma is another area where

00:22:40.860 the child comes in having sustained a bad injury in a car accident. You go out of your way to

00:22:45.320 use ultrasound whenever possible, but invariably sometimes patients do require x-ray and CT

00:22:50.740 radiographic studies. So let's go back to the x-ray because I like where you started there

00:22:56.960 historically. And then you alluded to the fact that basically if you understand how an x-ray works,

00:23:01.720 if you truly understand what's happening, then you understand what a CT scan is doing because it's

00:23:05.460 just doing it in three dimensions spirally. So the other thing about x-ray that I think is very

00:23:11.420 interesting for anyone who spends time looking at it just to appreciate it, even though it seems so

00:23:15.080 simple, is nothing in the body is two-dimensional. So you talked about how this photon is going

00:23:21.600 through the body and if it hits a rib, well, that's going to show up as white. Whereas if it's passing

00:23:26.320 through the lung between the ribs, it's going to show up as black. But of course, you could hit a rib

00:23:31.280 on the front, but not on the back. You can hit a rib on the back and not on the front. You can hit

00:23:35.300 the sternum. So when you look at an x-ray, even to this day, I'm still constantly amazed

00:23:41.720 at what the collage looks like of overlapping layers of tissue. And I mean, I just don't think

00:23:50.440 I ever got good at reading x-rays. It was actually easier for me to, I think, because we just spent

00:23:54.880 more time reading CT scans and there's so much more anatomic detail. But when I look at these old

00:23:59.360 time radiologists, look at x-rays and the stuff that they could pull out of it, I was blown away.

00:24:04.540 Exactly. It's true. It's an amazing skill to actually be able to pull that 3D information

00:24:08.680 out of a 2D picture. And if we actually kind of think about it, our brain is designed to always

00:24:13.760 imagine in 3D, always think in 3D. Even if you actually, somebody lost an eye, they can still

00:24:19.260 see in 3D. And that actually has to do with the fact that that's how our brain was wired or I was

00:24:23.600 wired. And so the amount of information in x-ray is phenomenal. But the biggest problem is that

00:24:29.520 sometimes you're actually overlapping different things and you just can't see. So quite often,

00:24:33.520 the reason we do two x-rays, when you do a chest x-ray, you do a frontal, a PA, as well as a lateral,

00:24:39.380 is so you can actually try and mimic those two together to become a three-dimensional object.

00:24:44.800 And so when CT came around, that's when it actually really allowed us to look at things in 3D. And

00:24:49.200 that's where surgeons and everybody else who actually operates and deals with people in 3D,

00:24:54.280 they can actually start to look at these and actually start to imagine what they're going to be

00:24:58.160 operating on. They actually kind of get the power of where the 3D image comes in. And so a simple way

00:25:03.300 to think about an x-ray, as I try to tell people, it's almost like an x-ray is basically like a flash.

00:25:08.060 Take a single flash and you get a picture. Whereas a CT scan is really like a searchlight in a boat

00:25:13.100 kind of going around an island. You actually get all the images the whole way around. And then the

00:25:17.620 equipment, what it does, it actually sort of sees how the intensity passes through, let's say,

00:25:21.520 two panes of glass and comes out the other side. And then you can actually start to evaluate what's

00:25:27.420 going on inside that entire building. The police officers use this all the time when they actually

00:25:31.740 need to stake out a building or a site, they actually use it to determine where the occupants

00:25:36.080 are. And that's exactly what an x-ray does. It's triangulation effectively by spinning around in

00:25:41.500 multiple different cycles by going around 360 degrees.

00:25:44.180 So we should clarify our semantics for people. We use the term CT very loosely, but it stands for

00:25:49.340 commuted tomography. And sometimes people back in the old days used to refer to it as CAT scan or CAT.

00:25:54.540 And the A was for axial, of course, because it's going up and down the axial dimension of the

00:25:59.460 person. When was the first CT scan put into clinical practice? I mean, would it have been in

00:26:05.100 the early 80s or something like that or earlier? Earlier than that, it was actually EMI, the

00:26:09.800 phonographic company that actually built the very first CT or CAT scanner. And I believe it was in the

00:26:15.020 70s or even earlier. It's actually been around for quite a while. Realistically, that three-dimensional

00:26:20.140 image really revolutionized medicine, as did all imaging.

00:26:22.940 And the other thing that people will often hear about, and I guess maybe patients don't hear about

00:26:28.080 it as often, but it certainly gets touted to them as a feature, is they talk about the speed of these

00:26:33.440 things. They say this is this many bits or that many bits. And presumably the first one was a four-bit?

00:26:39.360 I think it was actually even less than that. It might have been two-bit. And it actually just took a

00:26:42.480 long time to go around. And it was actually first used in the brain.

00:26:45.960 So let's explain what that means. So two-bit means you really only have two flashlights.

00:26:50.380 Exactly. For one flashlight, one detector.

00:26:52.800 Exactly. Yeah. It'll be easier, I think, when people look at pictures and we'll make this clear.

00:26:56.640 But you have this cylinder that goes around the patient who's laying down. And there's one place

00:27:01.800 where you shine the ionizing radiation. And on the backside of the cylinder is where you read it.

00:27:08.640 And then that thing has to rotate. So it's moving in its rotational plane, but also moving up and down

00:27:15.160 the Z-plane in time, correct? Right.

00:27:17.660 This makes so much more sense when I can use my hands, by the way.

00:27:20.720 And so it's exactly like that. Basically 180 degrees apart, you have the X-ray and the detector.

00:27:25.840 And then it actually starts to spin around in rapid revolutions. So when you actually look at a

00:27:29.820 machine, there's the donut hole in the middle, but around it, there's actually the casing and these

00:27:35.420 two parts, the detector and the machine that are actually spinning around the body very, very quickly.

00:27:39.540 It's actually quite fascinating. Maybe we can find a picture of one of these actually with a cover off.

00:27:44.060 We'll find one. And I feel like even when I got to residency, which is, so let's just say

00:27:50.940 directionally 20 years ago, I mean, I still think people were using 16 and 32-bit scanners, right?

00:27:57.380 Yeah, they were.

00:27:58.120 So 16-bit means you've got eight, you're spaced out equally eight units, and then you've got your

00:28:03.300 eight detecting surfaces.

00:28:04.920 Right. And so what it winds up doing is, so if you start at the, let's say the top of your head and

00:28:09.160 go to the bottom, we'll call that the axial dimension. What the eight bit would mean is that,

00:28:13.780 or eight slice, I guess is probably the more correct term, is that you basically have like

00:28:17.840 one slice and then you're actually measuring immediately below it and immediately below

00:28:21.060 that. And so as a result, as you're circulating around the person or the patient, you're actually

00:28:26.040 doing eight slices at a time, or 16 slices at a time, or 32 slices at a time. So what that means

00:28:32.040 is as you're rotating around, the amount of coverage you get is more, the higher the number

00:28:37.440 of slices, or you can actually also have thinner and thinner slices. And the more thin you get,

00:28:43.300 the more detail you can get from an imaging perspective, but the more radiation you require

00:28:48.300 to overcome the signal and the noise background.

00:28:51.220 Yeah. So what you're trading off is an optimization problem, which is speed, resolution, radiation.

00:28:57.800 Now, where are we today? I assume 256 is pretty common.

00:29:02.360 256 is one of the common ones, and it's actually used for things that are rapidly moving,

00:29:06.060 typically the heart. The heart. Do we have a 512 yet?

00:29:08.780 In research.

00:29:09.440 Really?

00:29:09.940 Yeah. You can actually make them as high as you want. The problem is you eventually sort of get

00:29:13.060 to these diminishing returns of how thin the slices are and how many images you need.

00:29:17.360 So if someone needed to scan a part of the body with a CT scan that wasn't moving functionally,

00:29:23.320 so something that's anatomically complicated, like the pancreas, but it's not moving like the heart,

00:29:29.120 are you good enough at 128 or 64? Like does 256 offer an advantage?

00:29:33.580 It doesn't really offer any advantage, no. You can actually,

00:29:36.600 eight slices work quite well. As long as the person can hold their breath and there's not

00:29:39.800 a lot of movement, eight slices work well.

00:29:41.320 One of the things I remember at Hopkins, there was a radiologist there. I don't know if he's

00:29:44.980 still there. I think his name was Elliot Fishman, and he was sort of the god of pancreatic

00:29:49.400 reconstruction. And of course, this was important at Hopkins because at the time, Hopkins was the

00:29:53.240 epicenter of pancreatic cancer surgery. And as important as it was to have a great surgeon,

00:29:59.600 John Cameron, Charlie Yeo, et cetera, that could do this operation, it was as important to have

00:30:04.800 an exceptional radiologist. Because as you know, and maybe some of the listeners know,

00:30:10.440 many patients with pancreatic cancer technically shouldn't be operated on.

00:30:14.180 And you'd really like to know that before you enter the patient's abdomen, which is not always

00:30:18.560 possible. But I remember that Elliot Fishman's images, he would have these 3D reconstructions of

00:30:24.380 the pancreas at a time. I mean, today that's pretty common, but at the time, like nobody was

00:30:28.340 contemplating this kind of resolution. And we would sit there on rounds and look at these images when

00:30:33.980 they were still printed out on that sort of vellum, whatever the hell that plastic paper is. And you

00:30:38.540 just couldn't believe it. So to think that he was doing that with relatively few slices, right?

00:30:44.200 Exactly. And like the whole power of that is, again, as I mentioned at the very beginning, is that

00:30:47.800 our brain thinks in 3D, right? And so the power of that, as opposed to looking slice by slice at a 2D

00:30:53.120 image, became very useful because now it became real. It became what our eyes could see, what our

00:30:58.160 brains could see. And it actually really helped everybody plan their surgery. And so it really

00:31:02.940 was revolutionary to actually start to look at things in three dimensions.

00:31:05.660 Now there's another element that we're going to introduce to this, which is contrast. So what is

00:31:11.940 contrast? Why do we use it?

00:31:13.560 What contrast is, and for the CT world, it's actually an iodinated material. And so what iodine does,

00:31:19.320 it actually absorbs the photons and so therefore makes things look white on a CT scan.

00:31:24.400 So it's like having liquid bone in your bloodstream.

00:31:27.200 Pretty much. A liquid photon absorber. And so what it does is, so we actually inject it into

00:31:32.220 the vein and then we actually, the heart pumps it around. And so that means we're actually able to

00:31:36.620 time when we actually take the CT image in order to be able to see what type of organ we're looking

00:31:41.420 for. So you can think of it, if you actually get the arterial phase, you'll actually see where the

00:31:46.100 arteries are connecting to an organ or else you can wait for the venous phase or when the veins are

00:31:51.020 returning blood flow from that organ. And you actually see the entire detail of the organ.

00:31:55.580 And what contrast really is, is basically a way to light up the blood vessels and light up the

00:32:00.560 capillary net and everything in between, between the artery and the vein to allow us to see the

00:32:05.640 anatomic detail from the perspective of adding blood to it.

00:32:08.900 And the name of course, explains exactly why it's to create contrast.

00:32:13.520 In the absence of contrast, blood looks functionally like water. I feel like I just

00:32:19.920 want to go into this in so much detail, but I'm also trying to be mindful of not going deeper than

00:32:24.060 we need to. But I guess we can talk about a Hounsfield unit because that will allow us to

00:32:28.040 explain this contrast thing and tissue differences, right?

00:32:31.040 Exactly. So what actually happened in CT is that Hounsfield came along and kind of said,

00:32:35.120 okay, how do we calibrate this? And so there's a range of Hounsfield units from minus 1,000,

00:32:39.880 zero to 2,000. And what that actually has to do with is density. And so zero is defined as water.

00:32:47.320 A thousand is basically air. And so as a result, we can actually see the difference between the

00:32:52.260 density of bone and air with water effectively in the middle.

00:32:56.740 So you have at plus 1,000, it is black.

00:33:00.280 Right.

00:33:00.640 At minus 1,000, it is pure white.

00:33:03.740 Pretty much. Yeah. And the biggest problem is that the eye actually can't see that range.

00:33:07.060 So on our computers, we actually will narrow down and look at, we'll actually look at the

00:33:11.680 higher Hounsfield units and we'll actually see lung. Then we go down and we look at the denser

00:33:15.760 material and that becomes bone, but we can't see the whole thing.

00:33:19.560 Right. So you could technically specify multiple parameters. You could specify the width of your

00:33:25.340 window and where it is centered, for example. So if you wanted to look at lung, you would center

00:33:31.400 it much closer to positive numbers and you don't need a very wide range, do you?

00:33:36.740 No, not at all.

00:33:37.440 So what's a typical lung window? Like 800 plus or minus a hundred or something to that? I mean,

00:33:42.040 I have no idea. I don't even remember anymore, but it's, that's the gist of it, right?

00:33:45.660 Exactly. Yeah. And actually, I don't even remember because you push a preset.

00:33:49.700 Once you set it, you never really change it much.

00:33:52.140 So I had committed all of these to memory in medical school. I was so obsessed with knowing

00:33:55.560 the window for optimizing the viewing of every tissue.

00:33:58.520 As did I. And then when you actually start to practically use it, you kind of, you wing it

00:34:02.060 because what happens is that every person is actually slightly different appreciation for

00:34:06.580 contrast. Yeah.

00:34:07.480 And also it's like the amount of photons lost based on the patient's body size, the amount of

00:34:12.360 absorption changes things. So the numbers actually kind of move around and you wind up building a

00:34:17.140 database in your mind of actually what you're looking at. And so when you first start, you actually

00:34:21.060 push the buttons and it's like 60, 40 for the abdomen. And you actually know all these numbers

00:34:24.780 and then as you kind of go through, you're like, nah, I just need to see what I need to see.

00:34:28.920 What you said a moment ago, it really brought back those memories of how horrible it looked

00:34:34.300 if you tried to set the window to be the entire plus minus 1000, you could appreciate nothing.

00:34:40.280 And that sort of struck me as a metaphor for life at times, which is like at a thousand feet,

00:34:45.100 sometimes you can tell I'm looking at a human, but that was about, that was about the limits of

00:34:49.060 detection, but you could appreciate so many different things by zeroing in on the capillary

00:34:54.640 of the lung, but you had to be in the right resolution versus if you wanted to look and

00:34:58.440 see if they actually had a fracture. People forget CTs are great for bones, right? And

00:35:03.220 we're going to talk about how MRI, for example, is less great for bones. So now we've got the

00:35:08.180 CT thing. So what we've established is that x-ray is a purely anatomic study. There's nothing

00:35:14.420 functional about it. As you go into CT by itself, it's also a very anatomic study. You can add

00:35:19.840 contrast to get even better information about the vasculature. And you now have so much information

00:35:26.320 that you can basically titrate or calibrate the window in which you look into that collection

00:35:34.080 of radiation and specify your tissues. CT scans are generally pretty quick, right? People who are

00:35:39.580 claustrophobic don't tend to struggle that much in a CT scanner, correct?

00:35:43.420 Exactly. And that's actually the real power of CT is the speed. So for example, in trauma settings,

00:35:48.040 that's what you want. Basically, if the patient's not doing well in a trauma, you want to put them

00:35:52.180 through the CT scanner as fast as possible to get the information out as fast as possible. And they're

00:35:56.420 very fast. So something that's even faster than CT and comes without at least one of its most

00:36:02.780 significant drawbacks, which is radiation, is ultrasound. So how does ultrasound work? Where does

00:36:09.720 it fit into this? And what are its limitations? The way ultrasound works is basically it's a high

00:36:15.800 frequency, so higher than what our ears can hear. And effectively, it's penetrating solid tissue.

00:36:21.800 So it's a high frequency sound wave versus an ionizing wave of energy.

00:36:27.180 Exactly. And so then it's actually going and every tissue interface, it actually reflects back.

00:36:32.220 So it's very much like an echo. So if you're standing in a mountain range, and you yell out, you actually

00:36:36.860 hear the echo coming back. And you can actually from that time, you can determine how deep that tissue

00:36:41.160 is. And so with ultrasound, we're just doing that very, very fast. And so at every tissue interface,

00:36:46.500 or every mountain range, if we could, you actually hear that reflection coming back, you actually are

00:36:50.980 able to then composite that as a representative of how deep things are away from you.

00:36:56.460 Now, there are animals that do this, right? Including some of our closest relatives, right?

00:37:02.160 They do. And as bats also do them, that's actually how they see.

00:37:05.420 And the bats' resolution on this is what compared to, say, a dolphin. I've read, and I feel like I read

00:37:11.700 this in the journal Science many years ago, so I'm almost assuredly not remembering this correctly,

00:37:16.080 that the resolution with which a whale or a dolphin could undergo sonography rivaled that of our finest

00:37:24.480 medical equipment. I mean, their ability to discern was remarkable. I found that amazing. And of course,

00:37:30.880 in part, that's because the medium through which they travel is water, as opposed to what the bat

00:37:35.980 has to do, which is go through this poorly, poorly dense air, right?

00:37:40.080 Exactly. So traveling through a different material actually makes a very good way to explain it,

00:37:44.860 because in the air, actually ultrasound doesn't penetrate very far because it's this high frequency

00:37:49.580 and it just, you lose it because there's no reflection coming back. Whereas in more solid material

00:37:54.480 like water or even dense material like organs in the body, it actually reflects back easier.

00:37:59.680 And so it's actually that reflection that actually allows you to discern the different tissue types

00:38:03.760 based on how quickly it reflects back. Now, ultrasound can't harm you, right? You don't have

00:38:08.800 ionized. You could ultrasound yourself all day, every day for the rest of your life, and you're not

00:38:12.420 increasing your risk of cancer. Whereas if you did a CT scan of yourself every month, you're going to

00:38:18.060 be in trouble after several months. Exactly. The drawback of ultrasound is the resolution doesn't

00:38:23.260 seem to be as high. Right. With ultrasound, you're only looking at one slice, right? So you're only looking

00:38:27.900 at one slice in time and you're basically kind of sweeping through an organ, trying to composite

00:38:32.400 those slices together in your brain to try and build that 3D model. Because like I said, our brain

00:38:36.980 always wants to build a 3D model. So with ultrasound, as you sweep through, you get one layer, then you get

00:38:41.560 another layer, then you get another layer, and then eventually that's composited together to see what's

00:38:45.660 going on. An ultrasound also seems to really struggle when it encounters air inside the body. So if you're

00:38:52.820 trying to do an ultrasound of someone's aorta, but their bowel is in front of it, it becomes difficult

00:38:58.220 to see. For the same reason the bat can't really use high-frequency ultrasound to fly. Exactly. And so

00:39:04.020 that's why when, for example, you look at a female pelvis, you want their bladder to be full. Because

00:39:08.560 in their bladder being full of fluid, it actually acts as a nice window to allow the ultrasound beam

00:39:13.080 to pass right through to be able to see the uterus behind it. I think any woman listening to this who's

00:39:18.040 been pregnant, that's got to be one of their most vexing parts of prenatal ultrasound is they always

00:39:23.540 had to sit there in a waiting room with a full bladder waiting to have that ultrasound. And of

00:39:28.760 course, that's why we like to do ultrasound on a fetus, right? Is you're not causing any harm.

00:39:33.620 I mean, certainly one thing I came to appreciate in the hospital was the skill that was required on the

00:39:38.980 part of the person doing it. So if you gave me the best ultrasound device money could buy today,

00:39:45.180 like you literally went, bought whatever ultrasound was at the absolute limit of technology. And then you

00:39:51.500 walked down to the local hospital here, Vancouver General, and you grabbed just a middle-of-the-road

00:39:56.540 radiology ultrasound tech, someone who's maybe been out of school for a year. And you gave him or her the

00:40:03.740 worst ultrasound machine on the market. There's no comparison who would be able to see more.

00:40:09.520 This is where skill and experience is invaluable. Basically also dealing with the difficult patient

00:40:15.560 body type is really critical and you can't replace that experience.

00:40:19.400 Now there's a special subset of ultrasound that we do on the heart. Where did that idea come about?

00:40:25.140 Who figured that idea out?

00:40:26.340 Again, the value was like, once you could actually find like a nice window that would actually allow

00:40:29.820 you to miss the air in the lungs and actually kind of look at the heart, you realize that, boy,

00:40:35.580 you can actually start to see this two-dimensional plane of the heart quite well. And as a result,

00:40:39.640 you can actually see where things were moving, such as the valves in the heart or the walls of the

00:40:43.820 heart. And then as well, you actually add something called Doppler, which is basically the frequency

00:40:48.880 bouncing off of blood vessel. If it's going or coming, the frequency is going to be different.

00:40:54.040 And so as a result, you can actually now start to see how blood is moving. And so that's what

00:40:58.840 ECHO does. And so it actually allows you to look at the heart in detail with a very,

00:41:02.280 very thin window, usually underneath the chest and around the lung.

00:41:06.960 Yeah. So anybody listening to this, who's had an echocardiogram, they know

00:41:10.100 that the person doing it is really pressing quite hard. It's somewhat uncomfortable for you,

00:41:16.020 the person getting the echo done. And the reason is they've got that jelly on you,

00:41:20.600 which again is doing everything to eliminate even a drop of air between the interface. And secondly,

00:41:26.220 they're pushing and they're grinding it in between the ribs and they want to get that view

00:41:29.900 versus, so that's a trans thoracic echo where you're doing it over the chest.

00:41:34.860 In surgery, often, if we needed to look at the heart, you would add the anesthesiologist would

00:41:39.060 actually put the echocardiogram in the esophagus and you get an even better view of the heart.

00:41:44.260 The esophagus sits right underneath it and there's nothing in between. And it's a beautiful view. And

00:41:49.680 the patient, obviously, because they're asleep, they don't have to worry about having this huge

00:41:53.860 probe stuck in their esophagus.

00:41:55.580 Right. And because you're closer to the organ you want to see being the heart,

00:41:58.780 the details can be fantastic. Because one of the other things with ultrasound is that

00:42:02.660 the deeper you go, the beam effectively fans out and gets thinner and thinner. So you actually get

00:42:07.640 less sort of detail on the edges. Whereas right in the center, right underneath the probe is where

00:42:11.840 your maximum amount of detail is going to be. So by having it right in the esophagus, which is right

00:42:16.280 beside the heart, you're going to get fantastic detail.

00:42:18.320 As important as the CT scanner was in trauma, the ultrasound was actually the most important

00:42:24.520 radiographic tool we had in trauma. And that was the one thing that even the surgical residents

00:42:30.900 needed to know how to do. And it was called a fast ultrasound. There was an algorithm for this

00:42:35.980 because in a busy trauma center like Hopkins, you're going to see trauma so often, penetrating trauma

00:42:42.540 in particular, where you have to know, does this person need to go up to surgery? Is there fluid

00:42:48.180 in the abdominal cavity? That's generally one of the things you care about. You certainly also care

00:42:52.460 if there's fluid around the pericardium, this non-stretchy sac that surrounds the heart. These

00:42:58.200 are surgical emergencies, especially fluid around the pericardium. So I think sometime in our second

00:43:03.700 year of residency, we would go off and do this course where on pigs, we would have to practice this

00:43:09.320 over and over again until you learned the four places that you were looking for fluid inside the

00:43:14.300 abdomen. I guess we got pretty good at it. I think I got okay at it, but I still always felt like a

00:43:21.180 little nervous when push came to shove because I always felt like I wish I could go and spend a year

00:43:26.540 just being an ultrasound tech to really, really dial this in because the stakes are so high, especially

00:43:32.900 if you miss the slight amount of fluid in the pericardium, that's a lethal injury.

00:43:36.960 And one of the real tricks as well is that depending on the composition of the fluid,

00:43:41.600 there's like frank blood or coagulating blood. It can be really tricky to actually pick it out.

00:43:45.560 And so a lot of times you'd actually see that fast ultrasound would be done.

00:43:49.060 And if people weren't a hundred percent confident that there's a problem or not a problem,

00:43:53.600 they go straight to CT scan because you just couldn't make that error. And that happens time

00:43:57.480 and time again. Yeah. And it's funny, one of the last traumas I was ever involved in as a resident

00:44:02.940 was just one of those cases where the patient came in and he was responsive. He had the tiniest,

00:44:09.680 tiniest stab wound. So less than a centimeter wide under the xiphoid. That's it. So this is a guy who

00:44:16.460 walks in who has a sub centimeter sub xiphoid incision. I mean, it could have been a shaving cut,

00:44:22.480 but he's been stabbed and he's more or less seems pretty normal. Vital signs are more or less what

00:44:31.040 you would expect. When I lay him down and do this ultrasound of his heart, it really looks like

00:44:36.240 there's something there, but I can't quite figure it out. And now the question is, well, he's obviously

00:44:42.120 too responsive to warrant cutting his chest open, which was what you would do in the emergency

00:44:46.980 situation. So you have to do the CT scan. But of course the risk in the CT scanner is

00:44:52.400 as fast as the CT scanner is, he is still laying down on a scanner, potentially ready to have a

00:44:59.740 cardiac arrest for at least a minute and a half. And usually what would happen is I don't recall in

00:45:05.960 this case, but a lot of times, if you're going to go through the trouble of doing that, you're going

00:45:08.820 to do a contrast CT scan as well. You're not just going to do what we would call a dry scan with no

00:45:13.900 contrast. So now you've got the fumbling around of getting the iodine machine hooked up to him,

00:45:18.920 et cetera, et cetera. And Eddie Cornwell, who was the chairman of surgery at Hopkins,

00:45:23.300 he's now the chairman of surgery at Howard and just an incredible human being. I remember one of the

00:45:28.100 things that he told us when we were junior residents is beware of the patient who gets wildly anxious when

00:45:35.220 you lay them down. And sure enough, when we go and lay this guy down in the scanner, he just starts

00:45:40.740 freaking out. And when you sat him up, he sort of calmed down a little bit. It was do not pass go,

00:45:48.120 do not collect $200 and took him to the OR, opened him up immediately. And sure enough,

00:45:52.700 that knife had actually hit his pulmonary vein. And so that pulmonary vein was bleeding into his

00:45:58.480 pericardium. And so he would have had a cardiac tamponade if we'd left him on that table. And amazingly,

00:46:03.400 that patient went home three days later. Yeah, no, exactly. And you can see like the power of the

00:46:08.280 clinical skill as well as like just the basic imaging, right? The power of imaging

00:46:11.600 plus clinical is pretty much where medicine is right now and how we actually are able to

00:46:17.720 diagnose things quickly and efficiently. So we've talked about two technologies that most

00:46:21.680 women are very familiar with when it comes to breast cancer, which is of the cancers where

00:46:27.780 screening is done vis-a-vis imaging technology, breast cancer is head and shoulders above the others

00:46:33.080 in terms of the frequency and ubiquity of the scan. So let's talk a little bit about,

00:46:37.200 because you've already explained what an ultrasound and an x-ray is. So now explain

00:46:40.640 what mammography is and why we would sometimes say mammography is sufficient versus insufficient.

00:46:46.680 And why do some women get told, well, you also need an ultrasound? Right. So basically mammography

00:46:51.080 is a lower attenuation x-ray. We're actually taking x-ray, but a weaker strength of it because

00:46:57.320 we never have to penetrate bone. And actually now it shines through the breast tissue, which is all

00:47:01.200 soft tissue. And so one of the things that actually is maybe stepping back is to actually kind of

00:47:05.800 look at breast tissue in particular. So breast tissue is composed of normal subcutaneous tissue,

00:47:11.580 which is mainly fat, and as well as glandular tissue. And so when women are in their childbearing

00:47:16.400 age, it's almost all glandular tissue to produce milk for eventually feeding a baby. And as women

00:47:21.680 then go through menopause, that glandular tissue can invariably involute. So it's one of these things,

00:47:27.000 you don't use it, it gets replaced with fat. But in some women, it actually doesn't get replaced

00:47:30.840 with fat. And that is what we call the dense breast tissue. So mammograms are very, very good at

00:47:36.880 shining through fat. And it actually allows you to see very, very simple things like calcification

00:47:41.100 in fat, because they are just so dense and it actually just stands out. Whereas in glandular

00:47:46.400 tissue, sometimes that photon of low energy x-ray doesn't make it through. And as a result,

00:47:52.060 the tissue is very, very hard to see through. And that's what we call the dense breast tissue.

00:47:55.760 And the reason it's hard to see through is because of all that glandular tissue that in some women

00:48:00.280 over menopause or even older, they just retain. Nobody really knows why they retain that extra

00:48:05.460 glandular tissue, why in some women it gets replaced with fat, in other women it doesn't.

00:48:10.060 We don't know why. And so many states, and I think they're actually over 38 and possibly soon to become

00:48:15.220 a federal law in the United States, is going to require that the very first line on a mammogram

00:48:19.660 report is going to be that the women's breast tissue is dense, limiting mammographic sensitivity.

00:48:25.760 Or the breast tissue is almost entirely fat, in which case mammograms are helpful. Because that

00:48:31.280 actually will really allow women to determine, was this test good enough? And so for women who

00:48:36.440 actually have dense breast tissue, so they, for some reason, if they're post-menopausal or if

00:48:40.780 they're pre-menopausal in their childbearing age, they still have a lot of glandular tissue. That means

00:48:45.160 a mammogram might not be enough, and therefore they need another second imaging modality to look

00:48:49.800 through the tissue. And that's where ultrasound comes in. And as well as MRI would come in as well to be

00:48:54.720 able to see through that dense glandular tissue that the mammogram can't see through.

00:48:59.080 Now, the last time I looked, and these data could be, they could just be simply dated,

00:49:02.840 but I think directionally this is right. A mammogram had a sensitivity of about 80,

00:49:10.020 call it 84, 85 percent, and a specificity of about 90, 91 percent. Does that still sound about right to

00:49:17.060 you? It depends, actually.

00:49:18.180 Yeah, that was like all comers, was the point I was going to make, which is it becomes almost

00:49:23.120 impossible to interpret what that means, because what you need to know is, what if I had a thousand

00:49:29.120 women that looked exactly like the women I'm scanning right now?

00:49:33.000 Exactly. And so that's where it actually becomes really critical in the fact that depending on the

00:49:36.820 breast density, and that's why it's important for women to know what that is, you're going to know

00:49:41.300 how helpful the mammogram is or may not be. But one of the other things that becomes actually very,

00:49:46.180 very powerful in the mammogram is to actually use comparison over time. So that's why they

00:49:50.860 recommend screening intervals of either one or two years, and it's a matter of academic debate.

00:49:56.760 Because if you actually have like a mammogram taken, and then let's say two years later,

00:50:01.120 you do another one, it's actually far more sensitive to see that subtle change over time

00:50:05.780 than it is to actually look at an individual mammogram all on its own. So a single mammogram on

00:50:10.260 a dense woman, its sensitivity is about 55 percent. It's actually quite poor, whereas on a woman

00:50:15.300 who actually has fatty tissue, it's very high.

00:50:18.600 How high?

00:50:18.960 It can actually be over 95 percent.

00:50:20.700 So let me explain what sensitivity and specificity means so that a person understands what this is

00:50:25.000 about. Let's just use the numbers 80 and 90, because those are generally accepted as an aggregate. So

00:50:31.360 when we say a mammogram has an 80 percent sensitivity, here's what we mean. If there are a hundred women

00:50:38.080 who have breast cancer, so there's a hundred women and we absolutely know that they have breast cancer,

00:50:42.720 and we subject them to a mammogram, 80 of them will test positive. 80 of them will have a true

00:50:49.240 positive, and 20 of them will test falsely negative. So the sensitivity is the true positive rate

00:50:57.680 over the true positives plus the false negatives, correct?

00:51:02.460 Exactly.

00:51:02.840 So the higher the sensitivity, the less likely you are to take someone who has the cancer and miss

00:51:12.460 them. That's the juice on sensitivity. Let's now talk about specificity. So mammography, a moment

00:51:18.340 ago you gave a staggeringly sad example, so we'll come back to that. But let's use the better one,

00:51:22.780 right? Let's say 90 percent. So now what does it mean to have 90 percent specificity? So that means

00:51:29.040 you take a hundred women who we absolutely know do not have breast cancer and you scan them. 90 of them

00:51:37.620 will correctly identify as not having breast cancer. 10 of them will incorrectly identify as having

00:51:45.360 breast cancer. So 10 will be false positives. 90 will be true negatives. So the sensitivity is the

00:51:53.680 number of true negatives over the true negatives plus the false positives. And so the example you

00:52:00.400 gave a moment ago is if you have a woman whose breasts are very fatty, not glandular, therefore

00:52:06.580 she's the poster child for mammography, you're driving that specificity up, which means you are reducing

00:52:14.260 the number of false positives. But in the example you gave earlier, which is a woman who might have very,

00:52:21.260 very dense breast tissue, imagine what it means to take a hundred women who have very, very dense

00:52:27.480 breast tissue and drive your specificity down to 50 percent. That means on a given day, half the women

00:52:35.540 that walk into your clinic are going to be told they have cancer if they don't. Exactly. So it's like

00:52:41.020 flipping a coin. And by the way, one of the greatest examples of this, and I mean, I attribute it to Bob

00:52:46.060 Kaplan, but maybe he heard it somewhere else, but I love it is you can make a test that is a hundred

00:52:51.620 percent sensitive. If you're willing to have 0% specificity and vice versa. For example, you could

00:52:58.040 send a letter. You could have a little card that says you have cancer and you show it to every single

00:53:03.880 person you meet. You have a hundred percent sensitivity, right? Right. You will never have a

00:53:10.280 false negative. The problem is that's so clinically useless because you have no specificity. And

00:53:16.600 similarly, you could have a little magic card that you show everyone you ever meet for the rest of your

00:53:20.580 life that says you do not have cancer. Guess what? You have a 100% specific test. It just has zero

00:53:26.940 sensitivity. So it's as useful as a warm bucket of hamster vomit. And so it's this trade-off between

00:53:32.260 sensitivity and specificity, which I'm teeing this up because I know we're going to come and talk about

00:53:35.860 this when we get into the more advanced MRI stuff. But the example of mammography is amazing to me

00:53:41.020 because it makes you realize you can't just rely on one test, especially when that test has such low

00:53:47.440 sensitivity and specificity depending on the individual. Exactly. And I think that that's the

00:53:52.200 real important lesson is that it's actually very individually tailored, right? And so if you have one

00:53:56.920 test and you don't know what your fingerprint or what the tissue that your breast is made of,

00:54:02.360 you really have no idea what you're looking for. So you always need the one to kind of find out,

00:54:06.520 is this good enough? People always talk about machine learning and AI and how it invariably it

00:54:11.100 has to infiltrate medicine. And it seems to me that one of the best places for it to do so is in

00:54:16.340 at least comparative radiology. So given the ubiquity of mammograms, hopefully every woman above a

00:54:24.640 certain age in the United States, Canada is getting regular mammography. There's no shortage of data.

00:54:29.820 Are there companies out there that are working on basically doing that once you have a baseline,

00:54:35.740 which would be almost impossible for a machine to read, but once you have that baseline,

00:54:40.620 longitudinally comparing it? There are actually a lot of companies that are actually doing that.

00:54:44.340 And even in some states that are actually using machine learning techniques to actually

00:54:49.640 help the radiologist and they actually can be used as even a second reader. There are a fair

00:54:53.820 number of companies working on that, but it's not perfect. What do you think that could improve

00:54:58.480 the sensitivities and specificities of mammography too? I mean, can we get to the point where,

00:55:03.500 I don't know, I mean, most people would say you've got to be north of 97, 98% on both to really feel

00:55:10.580 confident. I think that's a pretty high target to achieve. And the reason that would be is just

00:55:15.020 because of the way the tests are done and the individual variability of people. It's going to be

00:55:18.600 tough. Can machine learning get that good? It'll take a while. It's going to need volumes and volumes of

00:55:23.220 data that's actually reproduced the exact same way. And I think that's the biggest problem is

00:55:26.840 because we're unique. The way the breast is compressed, the way everything is done when

00:55:31.300 a mammogram is taken is somewhat different each time. So the amount of coverage is a little bit

00:55:35.960 different each time. It's possible. Are we there yet? No, we're nowhere near close, but we're getting

00:55:41.600 better. And that's also one of the challenges that we have when we look at the data on mammography

00:55:47.740 is it's so backwards looking. And so if you want to look at the most comprehensive study of mammography

00:55:55.480 and breast cancer screening, by definition, you are looking at a trial that was enrolling patients

00:56:00.700 15 to 20 years ago. And therefore you have to be able to say, well, how relevant is the technology

00:56:07.560 that was being used then relative to today? And in the case of mammography, it shouldn't be changing

00:56:13.160 that much, but I mean, things do get better. I mean, we're reading pure digital now. We have

00:56:18.620 much better capacity to read even an x-ray than we did 20 years ago, don't we?

00:56:22.540 We do. And basically now what's actually happening is we're actually doing a fluoroscopic version of a

00:56:26.940 mammogram where we're basically sort of trying to slice through this three-dimensional object and

00:56:31.460 actually get the detail of it at a three-dimensional layer. Whereas before the mammogram were typically just

00:56:36.540 like the chest x-ray was done, two different views compositing that three-dimensional

00:56:41.420 picture together. So that three-dimensional view of the mammogram is actually better than it was.

00:56:47.240 Now there's something else that I've never actually seen done clinically, but I've read about it called

00:56:51.800 MBI, molecular breast imaging. Is that used any longer? And what is it? The reason it came across

00:56:58.080 my radar was many years ago when I was just trying to get the landscape on ionizing radiation, this came

00:57:04.620 up as a test that was done as a follow-up to a mammogram. But I thought there must be a typo based on how

00:57:10.820 much, it had like something like 20 millisieverts of radiation. I mean, it was 40% of your annual

00:57:15.680 radiation limit.

00:57:17.140 What that is now, that's a functional test. So in the realm of nuclear medicine.

00:57:21.240 And so what that is doing is we're actually taking radioactive material and then we're actually

00:57:25.880 injecting it into the body. And so tissues that actually have increased mitochondria actually

00:57:31.120 concentrate this radio tracer. And so that typically happens in breast cancer. So it's actually used

00:57:37.280 with a radio tracer called Sestamibi. You've heard of this Mibi scan, which is where we actually inject

00:57:42.300 this into the heart. And we actually look at whether or not the heart is being properly perfused.

00:57:47.140 And so areas that aren't being perfused, so therefore the muscle is not alive, basically don't take up

00:57:53.080 the radio tracer. Whereas muscle that is alive does take up the radio tracer. And that muscle that's

00:57:58.100 alive, that's moving has a very high mitochondrial rate. So therefore it actually concentrates this

00:58:02.860 material. So breast cancer was actually doing a very similar thing. They'd actually have a high

00:58:07.600 metabolism. And so this tissue would actually concentrate, or this radio tracer would concentrate

00:58:12.140 in that tissue. So that was the MPI exam for breast.

00:58:14.920 Is that test still done?

00:58:16.340 Rarely, but it can be done in women who actually have very, very dense breast tissue and you actually

00:58:19.960 need to see what's going on. It can be done, but a lot of it's actually been replaced with

00:58:24.220 positron emission tomography scans.

00:58:26.240 So right now, how many women, young women, if we just say, because the young women are going to be more

00:58:31.320 likely to have dense breast tissue, do we have a sense of what percentage of them really are being

00:58:35.480 uncovered, meaning they're not getting adequate sort of surveillance with just mammography and

00:58:40.360 would require at least ultrasound? Is that a third of women? I mean, do we have a sense of what that

00:58:44.060 number is?

00:58:44.780 Depending on the jurisdiction. So the guideline for when you actually start screening for

00:58:48.320 mammography can either be for the age of 40 and higher or 50 and higher. Each jurisdiction is a

00:58:53.200 little bit different. So what that means is that basically anybody under the age of 40, unless you have a

00:58:57.860 family history of somebody having breast cancer at an early age, they're not getting screened at all.

00:59:02.440 So everything we've talked about from a technology standpoint, in some ways, pales in comparison to

00:59:08.400 what we're about to talk about, which is about as complicated a set of physics as you're going to

00:59:13.460 find within the walls of a hospital, right? I mean, it doesn't get a lot more complicated than an MRI,

00:59:17.880 does it?

00:59:18.540 No, it doesn't. It's really an engineered delight.

00:59:21.040 Yeah. And I certainly, again, thinking back to my brief six weeks of doing radiology,

00:59:25.980 I feel like more of my notes were scribbling down an explanation of how this thing worked

00:59:32.620 than anything else. So let's go back to the beginning. Who the heck thought of this?

00:59:37.620 There were actually three people actually thought about it, but the Mansfield is actually one of the

00:59:41.760 main creators of it in UK. And so the MRI machine really, it's actually quite the amazing tool.

00:59:48.200 And it actually wasn't initially developed for imaging. It was actually just sort of developed

00:59:51.760 on a benchtop where they're actually just kind of looking to see what the effect of electromagnetic

00:59:56.400 waves does to anything. And somebody wind up sticking tissue in it saying, hey, look, it's like

01:00:01.480 we can actually, what goes in one side comes out a little bit differently on the other side.

01:00:05.500 And as a result, we can actually determine what that composition of material was.

01:00:09.820 So does that mean like the NMR that we were looking at when we did organic chemistry was

01:00:14.900 really the precursor to the MRI that we're sitting outside of right now, listening to it hum?

01:00:19.700 It's the exact same device. The NMR basically is just a two-dimensional version of an MRI,

01:00:25.860 which is three dimensions because our brain likes three dimensions.

01:00:28.400 So let's go back to organic chemistry. So again, we'll link to a picture of an NMR spec so that

01:00:35.780 people can see what we're talking about. But I guess there's also no easy way around this, right?

01:00:40.080 I think you have to sort of roll your sleeves up a little bit on physics to understand how an MRI works.

01:00:45.060 There is no, I'm sure there's a kid's book out there waiting to be written on the topic,

01:00:49.620 which would be amazing, but it's pretty tough. So you take a molecule like alcohol. Okay. So it's got

01:00:58.300 these two carbons that are joined. The first carbon has three hydrogens around it. The next carbon has

01:01:04.860 a hydrogen and a hydrogen, but then instead of the third hydrogen, it gets an oxygen, which is bound to

01:01:10.840 a hydrogen. That is the stuff that people drink and get drunk on. Now put that into a nuclear

01:01:17.480 magnetic resonance machine, and you're going to see different peaks, right? It's going to show you

01:01:23.260 that there is a methyl group somewhere. It's going to say, it can't tell you what it sees, but it tells

01:01:28.500 you that there's a carbon bound to three hydrogens, right? Right. How does it do that?

01:01:34.940 Perhaps I might actually take up the challenge of a children's book, but the problem is I dislike

01:01:38.920 writing, but maybe for children, I'll be okay. The way it actually does it is actually quite

01:01:43.100 fascinating and it's actually relatively simple. So what it actually does is the hydrogen in

01:01:50.320 particular. So we're going to sort of fixate on hydrogen because that's the atom that we're

01:01:54.140 really interested in. And I'm just going to say one thing because you're going to do this anyway,

01:01:57.740 and I just want to preface it. You're going to use hydrogen and proton interchangeably, aren't you?

01:02:02.180 I will. Can you just tell someone why you're going to use hydrogen and proton

01:02:05.920 interchangeably? Sure. So the atom basically of the hydrogen is you have one proton and one

01:02:12.060 electron. And in the hydrogen proton, we don't really care about the electron. It just sort of

01:02:16.400 disappears. So the hydrogen, I guess, nucleus is a proton. And it has no neutron. Its mass is one.

01:02:23.660 Its mass is determined by the one and only proton it carries, correct? Exactly. Okay. So now hydrogen and

01:02:29.660 proton, they're the same thing for the purpose of this discussion. Exactly. So when we look at the NMR,

01:02:33.960 we actually have hydrogen bound to an oxygen or hydrogen bound to a carbon. And so the behavior

01:02:39.320 of that nucleus is going to be a little different. So there's basically a magnet that's creating a

01:02:45.660 field. And somehow through that, we can see how the hydrogen is bonded to either the oxygen

01:02:52.300 or the carbon in the ethanol molecule. Right. So we were talking about the NMR. And so what the NMR

01:02:58.180 does is that it really is a hydrogen or proton imager or actually just detector. And so the way

01:03:04.660 the magnetic field of hydrogen behaves, if it's attached to either the oxygen and OH of alcohol

01:03:10.940 or the CH3 of carbon is completely different. It actually gives off a different wavelength.

01:03:17.540 And so as a result, that's how we're actually able to get this, what we call NMR spectra.

01:03:21.140 And so what happened is that from there, there's a person named Damadian, who many consider to be the

01:03:28.360 father of MRI. He actually said, well, you know, look, if we can actually take what Mansfield and

01:03:33.140 Lauterberg did on a benchtop, can we actually put a human in it and actually start to see the soft

01:03:37.380 tissue? Because we know that our body's composed of roughly 70% water. There's a lot of hydrogen on

01:03:42.340 fats. So can we see that frequency difference? We can see it on a benchtop, but what about in people?

01:03:46.860 So we have more hydrogen in us. If we were just going to count up the atoms in us, hydrogen wins

01:03:51.340 all day long. Because as you said, if we're 70% water, that's two to one hydrogen over oxygen there.

01:03:57.260 And then all the fat that's in us is all the hydrogen to the carbon there. And there's basically

01:04:02.800 hydrogen in protein as well. I mean, so if you have a hydrogen detector, that's basically the way

01:04:10.220 you would describe an MRI. An MRI is exactly that. It's a hydrogen imager. So basically we're looking

01:04:14.720 at hydrogen nuclei, which is a proton. And so a lot of times as well, people actually talk about

01:04:20.860 proton spectroscopy, which is NMR. An MRI is just basically a simple hydrogen imager.

01:04:27.120 So I think anyone who's had an MRI knows that there's a magnet involved and it's generally a

01:04:32.920 non-trivial magnet. Some people have probably heard of some of these real horror stories where

01:04:37.020 accidentally in the hospital, a patient's wheeled in and there's a loose oxygen tank under the gurney

01:04:42.460 that's wheeling them in and it goes flying across the room and hits somebody and can kill

01:04:46.260 someone. So how strong is the magnet and why does it need to be so strong?

01:04:50.880 Right. So the magnets, they come in different flavors. So typically it's regarded as a Tesla.

01:04:55.720 And so a Tesla is roughly 10,000 Gauss. And so a Gauss is effectively what the North Pole can produce.

01:05:03.700 And it's actually the typical measurement for most magnets. But when we actually get into the MRI

01:05:08.120 field, it becomes so much stronger. And the reason we actually kind of need that high Tesla field

01:05:13.020 strength is because we're actually taking hydrogen, which is typically not that magnetic compared to

01:05:19.020 like a magnet that we think about like a bar magnet. And what we're trying to do is we're actually trying

01:05:23.440 to orient that little dipole of the water molecule or fat molecule a certain direction. And that's what

01:05:31.060 the static field does. And that's why it has to be so strong. So they come in flavors 1.5 Tesla,

01:05:35.540 3 Tesla, which is double the strength, 7 Tesla. And so the higher the Tesla they go,

01:05:41.960 the more it's actually able to pull all of the hydrogens and orient them in one direction.

01:05:46.540 Because as we're sitting here or anywhere, normally our hydrogen molecules are on water.

01:05:51.420 We're just kind of bouncing around randomly, kind of pointing at any which direction. And then

01:05:55.100 there is no kind of magnetic component to us. The hydrogens are spinning around just based on

01:06:01.160 Brownian motion. And so when you actually go into a magnet, that strong magnetic field,

01:06:05.800 these hydrogens basically kind of turn and orient themselves in that direction. And so that's what

01:06:11.100 actually provides the initial basis for an MRI. And that's why, contrary to what we see on TV,

01:06:16.000 that magnet is always on. You can never turn it off because that magnetic field always has to be

01:06:21.200 there in order to provide that orientation. And as well, the way it works is you actually have

01:06:26.900 a superconducting wire that's actually running just above absolute zero degrees Kelvin.

01:06:31.540 So it's roughly two Kelvin. And so you can't turn that off. This is superconducting wire. It has

01:06:38.320 like pumps that are pumping all the time to keep it that cold so that when you actually put an electric

01:06:42.700 field in like a loop of wire, that electric field is what actually generates the magnetic field in a

01:06:48.280 perpendicular direction.

01:06:49.640 So you have a generator that backs up if you lose power, which invariably you're going to lose at some

01:06:54.380 point.

01:06:54.860 Exactly. You have to have that backup power to always keep this pump moving, this liquid helium

01:06:59.300 that's surrounding the wires, to always keep that liquid helium floating around, circulating around

01:07:03.980 that wire to keep that wire in your absolute zero Kelvin.

01:07:06.640 Now, if we were to walk in the scanner today, and everyone can sort of picture this, there's a bed

01:07:10.720 running through a donut. What is the direction right now that that magnet is being oriented?

01:07:16.840 The donut. So that's where that loop of superconducting wire is sitting in. And so depending on how it's put in,

01:07:22.280 most of the time the north will actually face away from the control center. And that's the direction

01:07:27.120 of north.

01:07:28.180 Got it. And if I recall, there's like a right-hand rule on this, isn't there?

01:07:31.480 There is. Very nice. So it can tell you which way the power in the coil is going. And it's actually

01:07:36.000 going, I guess if you're looking from the foot of the bed, it's actually going in a clockwise circle.

01:07:40.820 And that's the right-hand rule.

01:07:41.960 The right-hand rule. It's pointing in the axis. Okay. Now, aside from the fact, let's pretend we weren't

01:07:47.340 wearing anything metal. If you walked into a room where there was a 10 or a 20 Tesla magnet,

01:07:54.300 would you feel anything? And would it do anything to you that is harmful?

01:07:58.700 The actual magnetic field won't do that much. Now, when you get very, very strong to a moving magnetic

01:08:04.480 field, you can actually start to feel it because it can actually trigger your nerve impulses to start

01:08:09.860 moving. So sometimes people actually, if the magnetic field is too strong, they'll actually get

01:08:14.260 twitching. And sometimes people who are actually around magnets all the time, they actually become

01:08:18.880 more and more attuned to this type of a thing. And so if you have MRI technologists or people who are

01:08:23.740 working with the high fields all the time, they can say, you know, when I go to the head of the

01:08:27.740 magnet or like the North side, I actually kind of feel something pulling. And then sometimes people

01:08:33.460 describe getting temporary headaches. And as soon as they step away from the field, it all goes away.

01:08:37.940 Usually when patients ask me if there are any side effects or harm of an MRI, I mean,

01:08:42.920 our lib answer is to say, no, no, no, no, no. Especially with a non-contrast, like if there's,

01:08:47.260 I mean, not that the risk of gadolinium is high, but if you're just having a dry MRI, you say nothing,

01:08:51.800 nothing, nothing. But I actually had a patient who had very, very, very severe migraine headaches.

01:08:56.920 And she actually had a migraine triggered by an MRI. And I truly believe that wasn't just a

01:09:01.420 coincidence. I mean, I think her headaches were so severe. So, so I now sort of always couch my

01:09:05.700 response as there's virtually no short-term or long-term consequence that can come from an MRI,

01:09:11.000 but at least in the case of that person, you could trigger a headache.

01:09:14.560 You can, because what it does is it can actually stimulate the nervous response.

01:09:18.160 And depending on how strong the field is. So for example, if you were to do a seven Tesla magnet,

01:09:23.080 you're definitely going to notice it. And as a result, I'll tell people, look, you can't be in

01:09:26.700 this too long because of the fact that you're actually going to stimulate.

01:09:29.520 So I remember going back to learning about this. One of the other things that sort of strikes anybody

01:09:34.700 who's had an MRI is they take a long time. So why is it that if you wanted to get an MRI,

01:09:40.320 let's nevermind whole body, which we'll come to, but you just want to get an MRI of the abdomen

01:09:44.760 that could easily take 40 minutes. Whereas a CT scan of the abdomen can take two minutes.

01:09:51.260 Yeah. And it's basically the way the images are acquired to completely different mechanism.

01:09:55.320 So if we go back and talk about the x-ray, it's basically like a single flash. We're actually

01:09:58.600 looking through everything. The CT is basically this x-ray that's constantly on spinning around and

01:10:04.200 basically sort of circulating around you. Whereas the MRI behaves completely differently.

01:10:09.380 And during the period of time of that acquisition, what it's doing is everything that's inside the

01:10:14.300 center of that donut is being pulled in a certain direction, all the hydrogens on your water and fat.

01:10:19.120 And then the loud part of the MRI is actually a temporary magnetic field, which is countering

01:10:24.800 that static field. And so it's actually now pulling all the hydrogens in the opposite direction.

01:10:29.780 And then in that opposite direction, it actually turns off. And then the hydrogens reorient to where

01:10:34.840 they were in that static field. And as they reorient, they actually give off a different

01:10:38.560 frequency. And that different frequency takes a while to gather. And that's what we call the TR

01:10:43.960 or repetition time or the TE of the echo time. And that you can't speed up.

01:10:48.380 So let's talk about TR and TE because it's the TR and the TE that determine what sequence you're

01:10:54.440 looking at. So again, I think the average person probably won't recognize these terms, but certainly

01:10:59.520 anyone in the medical profession will know the difference between a T1 weighted versus a T2

01:11:05.780 weighted image versus a spin versus an echo versus all of these things. So let's just talk about

01:11:11.180 the difference between TR, that repetition pulse time, and then TE. Is TE the time it takes to relax

01:11:18.460 back to its original position? So just discussing TR and TE, how do they differ in acquiring a T1

01:11:26.300 weighted image, which is the one that's really anatomically beautiful? It's the closest thing

01:11:31.560 you see to, wow, I know what I'm looking at and I'm not a radiologist versus the T2 weighted image,

01:11:36.920 which seems to highlight water more. So things that are water look more white, but it doesn't

01:11:43.500 have the anatomic resolution. How would you differentiate those?

01:11:46.720 Right. So the simplest thing to do, and it's actually quite fascinating, is I went through

01:11:49.680 residency. People are always sort of stunned with, is this a T1 image or a T2 image? And went

01:11:55.500 through all that and it was kind of like, sort of as a bit obscure. And then when they started to do

01:12:00.160 MRI much more, it's like it became actually pretty simple. On the T1 image, we actually see nothing but

01:12:06.200 fat. So fat gives a lot of signal, which is what makes it nice and bright. So we see a single

01:12:11.140 element, or I guess a single element that the hydrogen would be bonded to that we're looking

01:12:15.020 at. Whereas a T2, we're actually now seeing two elements. We're seeing fat and water. And so those

01:12:22.060 two elements are actually coming off at different frequencies from the MRI machine. And you have to

01:12:26.900 wait a longer echo time to be able to pick up the water because it returns back to normal much more

01:12:32.680 slowly than the fat does. So that's our TE time.

01:12:35.940 So T2 weighted images take longer to acquire because the TE is long because you have to wait

01:12:42.740 to get both fat and water.

01:12:44.900 What's the difference in the TR between the T1 and T2?

01:12:47.880 It's all going to be entirely dependent on the machine. So you actually have to customize those

01:12:51.380 parameters for every single machine. And so that actually kind of takes a while to actually kind

01:12:55.680 of go and calibrate and kind of get used to what your eye is used to seeing. And it's also going to be

01:13:00.320 dependent on the signal, the overall magnetic field, as well as the overall signal to noise for that

01:13:05.580 coil set that you're using. So each one has to be effectively tuned.

01:13:09.720 And then what does it mean when you have these other things that come out? And God, it's been so

01:13:14.040 long since I've done it that I don't even remember when we would look at spin, spin, echo,

01:13:19.400 flare, all of these. I vaguely remember all of these other sequences we would order. I don't actually

01:13:24.080 recall what they were. Give us a brief rundown of that.

01:13:26.560 One of the things that MRI absolutely loves is just the different acronyms for everything.

01:13:30.420 You got to have it.

01:13:32.220 It's almost like whoever can come up with the coolest acronym wins like LAVA and all these

01:13:36.860 other things. But effectively, there's three main categories of MRI sequencing. One is what we'll

01:13:43.500 call conventional. So conventional spin, echo. And so that's basically just waiting as long as you can

01:13:48.200 for the hydrogen to completely relax and give off both its water and fat signal. And then we have

01:13:54.240 what's called granite imaging. And so that's actually, you're not waiting for it to completely

01:13:58.000 return back to normal, but somewhere in the middle, you're actually kind of repulsing again. So you're

01:14:02.920 basically hearing that noise of the machine turning back on and saying, you know, we're not going to

01:14:07.040 wait to completely relax. We're going to fire up again.

01:14:09.360 And just give us a sense of the actual time. How many milliseconds, if you're trying to get a T2 signal

01:14:15.080 and you're waiting for that full relaxation, how many milliseconds is that directionally?

01:14:19.760 It's actually going to be, it can be up as high as like 60 milliseconds or even longer for some of

01:14:24.780 them. Okay. And when you do these gradient-based tests where you're going to repulse, how quickly

01:14:30.560 are you repulsing? You can actually repulse in like two milliseconds or even faster. And then there's

01:14:36.300 actually the third category, which is actually called EPI or echo planar imaging. And this is

01:14:41.580 actually amazing. So this part actually allows you not just to be looking at a single slice of a

01:14:48.120 person, but you're actually going and you're actually now running multiple slices simultaneously

01:14:52.980 where you're actually putting two different fields on the person at the same time. And so as a result,

01:15:00.060 and it kind of gets complicated because we use the word gradient all the time. And so what a gradient

01:15:04.920 basically means, it's effective like a ramp from a low number to a high number. And so if I was kind

01:15:10.380 of looking at you and I sort of said, we're going to start the image from your top down. First, we're going

01:15:14.500 to put a gradient on from top to bottom. So it's going to be a little bit of a higher frequency at

01:15:18.400 the top, a little bit of a lower frequency, lower down. And then we're actually also going to look

01:15:22.840 at phase from right to left. And depending on how your body's oriented and where the blood flow is

01:15:27.360 going to be, we're going to look at phase and frequency, which now bring us into the realm of

01:15:31.960 a Fourier transform. So these are now with the pulses, we're effectively looking at all these repetitive

01:15:37.340 sine waves. And we're actually plotting that in frequency and phase domain.

01:15:42.320 Right. And for the listener, we always talk that Laplace was only half the man Fourier was.

01:15:48.720 That's like the nerdiest math joke I'm going to tell today. All kidding aside,

01:15:53.100 how does one get into MRI radiology without a background in mathematics and physics? It seems

01:15:58.680 it would be impossible. It's a struggle. I can actually remember with the group of residents that

01:16:03.520 I was with training and great people that we actually had a physicist come in and was talking about

01:16:08.760 MRI physics for a couple of days and trying to teach us all. And I thought, boy, this guy's

01:16:13.060 really watering it down. I can barely hang on to what this guy's saying. Then I kind of looked and

01:16:17.820 talked to all my colleagues and they were just bewildered. They had no idea what he was talking

01:16:21.880 about. Because as far as an engineer is concerned, it's like Fourier domain, it's kind of like,

01:16:26.640 that's sort of like the alphabet.

01:16:28.220 Yeah. That's our bread and butter back in the day.

01:16:29.980 Yeah. The difference with the MRI is that you start in this Fourier domain. And

01:16:33.360 because we're a three-dimensional object, when you're looking at a two-dimensional plane,

01:16:37.300 that's the Fourier transform. When you now add that third dimension, it becomes what we call

01:16:41.140 K-space. So it's effectively a two-dimensional Fourier transform, which is what the MRI world

01:16:45.900 operates on. It's called K-space.

01:16:47.820 So what we're going to do in the show notes here is we're going to link to some very common

01:16:53.140 types of MRIs that people have, right? The most common ones that people have is you've tweaked your

01:16:58.240 knee. You're going to get the knee MRI. You've hurt your back. You're going to get the MRI of your back.

01:17:02.760 You're having headaches. They're going to do the MRI of the head, those sorts of things. So when

01:17:06.660 you think about the bread and butter clinical practice of medicine, what types of MRIs,

01:17:11.440 describe what those three would be. What sequences would be run to it? If you wanted to evaluate

01:17:15.300 someone's ACL, the ligament in the knee, what are you going to look at?

01:17:18.980 The first thing you're going to do, and this sort of relates back to the plane X-ray,

01:17:22.540 you're always going to look at things in two dimensions. So you're going to look at two planes in this

01:17:26.280 case. And so you'll be slicing from right to left, top to bottom, side to side. And the beauty of

01:17:31.880 the MRI is that based on how you orient your gradients, you can easily slice those three

01:17:36.780 directions or actually in any direction you want, which is why we call it multi-planar.

01:17:41.600 And so if we were to look at a simple thing, like a simple image, like an E, so we always like our

01:17:46.700 anatomy image. So that's our plane T1 because fat is beautiful and it actually allows us to see

01:17:51.680 everything really well. And it looks like what we're actually accustomed to seeing. And so we do a

01:17:56.640 a T1 sequence, then we'd also look at a T2 sequence or a T2 fat sat sequence. And what

01:18:02.440 that actually allows us to do is on a T2 fat sat, we actually go...

01:18:05.980 And that's fat saturation.

01:18:07.660 Yeah. So T2 fat saturation, what we're doing is we're taking T2. So now the T2 looks at two

01:18:12.500 things, so fat and water, and then we actually suppress the fat. So really all we're seeing is

01:18:16.820 water. And why that's most interesting is because edema. So when there's something going

01:18:21.640 wrong in the body almost anywhere, edema happens. It's like if you bang your hand and it swells up,

01:18:26.240 that swelling is edema. You injure your knee, it swells up, that's edema. And so if you actually

01:18:31.300 bang the bones on your knee, so you've actually injured the cartilage, you're going to get edema

01:18:35.420 happening in the ends of the bone as well as in the cartilage. So that's why the T2 with fat saturation

01:18:41.160 or removing the fat signal becomes so powerful because it effectively now turns into a edema imager.

01:18:47.080 And when we know when there's edema, there's a problem. And that's kind of a simple concept in

01:18:51.320 MRI that is quite often lost, but it's very, very important.

01:18:55.200 And then when you look at somebody's brain, for example, we just reviewed my MRI recently.

01:19:00.120 So one thing that stands out, I think, is the exquisite anatomic detail you're getting that

01:19:06.000 seems to look far better than it does in CT scan. And secondly, it's the fact that without any contrast,

01:19:14.160 you're able to see as though you did an angiogram on all the vessels in my brain.

01:19:20.960 Now, is that something that any MRI can do? Or is that just something that the MRI here can do?

01:19:26.920 Most MRIs should be able to do that. And so when we can sort of think back to what an MRI is,

01:19:32.240 so again, hydrogen imager, but it's also a big, powerful magnet. And so what makes our blood red

01:19:38.400 is actually the iron that's contained within it. So what you can actually do is you can take all the

01:19:43.060 blood that's, let's say, flowing to a particular organ like the head. So anything that's flowing up

01:19:47.280 and you say, okay, I'm going to actually excite anything going up north to the brain. And so

01:19:52.500 therefore that's arteries. And so then you'll actually get to see all the exquisite arteries

01:19:56.180 in your brain just by exciting that blood.

01:19:58.760 I had never realized something so obvious as you just said it, but that's one part of the body where

01:20:05.920 it's really easy. I mean, the limbs would be the same where directionally it's so clear.

01:20:11.620 You know, you know, which way your magnet is oriented. You know, which way is blood flow away

01:20:16.800 from the heart. And you've got this beautiful iron floating around in water.

01:20:22.920 That's one of the beauties of MRI is that there's all these different things that you can actually

01:20:26.100 add to it. And so not only that, you can actually excite anything flowing in one direction,

01:20:30.340 but you can actually also pick off the frequency that's different between oxygenated arterial blood

01:20:35.620 and deoxygenated venous blood. And so that comes into a different sequence called SWI or susceptibility

01:20:41.660 weighted imaging, where you can actually look at the deoxygenation status of venous blood and you can

01:20:47.180 get spectacular contrast to the small blood vessels in the brain using that sequence.

01:20:52.120 I remember the first time you did an MRI of my head. I don't know why I was more nervous about

01:20:57.500 seeing that than anything else, because we sort of remember the most extreme, horrible stories. But

01:21:03.440 I mean, I know people who have died of aneurysms. You and I were even speaking about a patient a few

01:21:08.620 hours ago about this. I used to ride a bike with a guy who was maybe six years, seven years older

01:21:14.580 than me and was at Disneyland one day with his kids and had a horrible headache and dropped dead.

01:21:18.840 And he had an aneurysm, which is congenital. And so, yeah, I was sort of like really nervous,

01:21:24.520 even though I realized straight probability basis, the odds were quite low. They weren't zero. And I

01:21:31.020 figured, well, it's better to find this now because you can treat these things electively quite easily.

01:21:35.980 But once they rupture, the mortality is incredibly high.

01:21:39.960 The mortality of a ruptured aneurysm is over 93 to 95%. So most people don't make it. Whereas when

01:21:45.740 you do find them earlier, there's all sorts of options, such as coiling, where you can actually

01:21:49.020 treat it or clipping. And so that's actually one of the real powers of being able to kind of see what's

01:21:54.240 going on without any injection or anything like that. You can see the exquisite detail of the arteries.

01:21:59.000 And if there is a problem and we've found a fair number of people with them,

01:22:02.680 you can actually save their lives.

01:22:04.400 What is the frequency? Admittedly, you have a, you could argue a somewhat biased population

01:22:09.560 because they're more health conscious. Obviously, anyone who can afford to just pay for an MRI out

01:22:15.240 of pocket is going to have a socioeconomic advantage. But if you argued that that's still

01:22:19.440 a reasonable cross-section of the population from a genetic standpoint, which is what we're basically

01:22:23.400 asking, what is the prevalence you find of aneurysm in the brain?

01:22:27.560 So when we actually scanned a thousand people, we actually found eight intracranial brain aneurysms.

01:22:32.760 So 0.8%.

01:22:33.880 That's higher than I would have guessed. Does the literature support that?

01:22:37.720 The literature is actually a little bit less. And the question is, is that because you don't

01:22:41.680 find them? Because basically people have passed away and we don't know what happens to the elderly

01:22:46.900 because if they pass away, it's natural causes.

01:22:49.620 Yeah. We're not doing the autopsies. Now there are other aneurysms that are not quite as lethal,

01:22:54.200 but are really bad. And the two that I remember from residency were splenic artery aneurysms and

01:22:59.960 popliteal artery aneurysms. Do you see those? And if so, at what frequency?

01:23:04.680 We found only, I think, two splenic artery aneurysms, but they are particularly deadly.

01:23:09.500 And now the popliteal arteries, haven't seen many of those. No, haven't seen those. And then those

01:23:14.220 should be actually easier.

01:23:15.360 You can palpate those on a thin enough individual.

01:23:17.760 I was going to say, those are actually easier to feel and to see and to look at,

01:23:21.380 but we actually haven't found any of those. I'm kind of still shocked. That's a frightening

01:23:25.220 statistic, Raj, to think that almost 1% of the population has an aneurysm in their brain.

01:23:30.760 And one of the things that we actually find though, and this may be showing the genetic

01:23:34.040 component to it, is that when you find it in one person in the family, next thing you know,

01:23:38.740 all their extended family's coming in, right? They want to know what's going on.

01:23:42.060 We looked into this three years ago. I have a patient, a young woman, probably in her late 30s.

01:23:47.280 Her mother died very young when she was very young. The patient was very young and the mother

01:23:52.320 herself was quite young from an aneurysm. It was not in the brain, but I'll, for the sake of trying

01:23:57.500 to protect her confidentiality, I'll refrain from saying what part of the body it was.

01:24:01.700 And then when we dug into her family history further, we found another person who had died

01:24:05.840 of an aneurysm in yet a different part of the body. So not aortic where you normally see that

01:24:10.480 linked to atherosclerosis, but a different major vessel. And our team did a bit of work on this and

01:24:16.920 actually found evidence that there really was potentially a genetic component here.

01:24:21.460 And so we petitioned her insurance company to pay for an MRA, a magnetic resonance angiography,

01:24:27.920 which is basically what we're talking about here. And they declined it, which really irked me. And

01:24:32.320 we fought with her insurance company for six months to get this paid and they denied it and they denied

01:24:38.060 it and they denied it. And finally the woman just paid out of pocket for it. And I was blown away at

01:24:43.260 how much it costs. Do you want to take a guess at how much it costs to get an MRA in the United

01:24:47.800 States? I've seen some interesting pricing. It was $9,000. Wow. Wow. That's a, that's

01:24:55.620 unbelievable. Yeah. And it was negative. So we were happy and she was obviously fortunate enough to be

01:25:00.620 able to afford that. But it upset me that we couldn't make a case to the insurance company that

01:25:03.960 two people, one first degree, one second degree related to this woman have died from an aneurysm young.

01:25:11.040 And they were like, yeah, that's cool. Wow. That's a tragedy. That's so let's come back to now

01:25:17.840 what you do here, Raj, because I've been around a lot of MRI. And if we bring this story back full

01:25:22.680 circle four years ago, or whenever we were introduced, the question that was asked of me

01:25:27.360 was, Hey, is this thing kind of cool or what? And I came up here and I spent the full day with you. And

01:25:33.160 I thought, yeah, boy, this is really cool. But so many things that you were doing just seemed

01:25:38.080 counter to what we were seeing in the big shot hospitals. And for starters, you were using a

01:25:44.420 puny magnet, right? So one of the things hospitals love to brag about is the size of their magnet,

01:25:50.720 right? I'm going to just try to not to name any hospitals and offend everybody, but, you know,

01:25:55.360 pick your favorite institution. We've got the newest four Tesla magnet. I probably was in the top five

01:26:02.400 questions I asked you. Oh, so what size magnet are you using? And you said, we're using a 1.5 Tesla

01:26:07.460 magnet. And I said, Oh, that's interesting. That seems a little bit JV. So explain why you use a

01:26:14.960 magnet that is not at the peak of what you're capable of just from a technology standpoint.

01:26:20.600 Well, it really sort of depends on like basically how you tune a magnet. So I actually kind of view a

01:26:24.840 magnet as very much like a smartphone. If you actually build the right hardware, you can actually

01:26:28.900 really get it to sing and do an amazing job. And so with a 3T magnet, one of the things is you can

01:26:34.660 get some pretty exquisite imaging as you see quite commonly. And really there's a lot of bragging rights

01:26:39.060 associated with it. We're not much of braggers up here. We just actually want to do the best imaging

01:26:43.580 we can and actually tune the machine as optimally as possible. So with a 3 Tesla magnet, one of the

01:26:49.860 things that actually commonly happens is that when you look at the wavelength of a 3 Tesla, because

01:26:53.760 remember it's electromagnetic fields actually go hand in hand, and those are actually

01:26:58.200 wavelengths. And so the 3 Tesla wavelength is roughly 15 centimeters, so it's the width of your

01:27:03.840 head. 1.5 is roughly 30 centimeters, so it's the width of most people at your shoulders. So as a

01:27:10.060 result, you actually start to get a lot more penetration with a lower field magnet. And so that

01:27:14.960 way you're actually able to see things quite well when you're particularly having everything tuned to

01:27:19.880 that particular wavelength. And quite commonly, when people are purchasing machines, they just don't know

01:27:26.000 the physics of how the static magnetic field, the gradient magnetic fields, the coils. And there's

01:27:32.160 roughly about 150 parameters per T1, T2 fat saturation sequence that you can actually adjust to make it

01:27:39.140 work the way you need it. And most of the time, what commonly happens almost everywhere is that you'll

01:27:44.960 have the vendor will actually come in and set the standard parameters. And from there on, it's kind of

01:27:50.520 like, let's make it as simple as possible, push a button. And that's kind of not what I do. We don't

01:27:55.640 want to just push buttons. We actually want to understand how it all works, how it can be optimized to

01:28:00.580 the person that's coming in, and how the entire system from front to back is optimized so we're

01:28:05.100 maximizing our signal-to-noise. And when you maximize your signal-to-noise, that's when you actually

01:28:09.920 really get a lot of speed and detail. And that's where one of the real values of being able to talk

01:28:16.020 engineering or physics with the MRI physicist, and also then put on your clinical hat and kind of

01:28:20.960 saying, well, this is what I want to see, this is the level of detail I need, and the level of resolution

01:28:25.220 that you can really tune the machine to where you want it to be.

01:28:28.620 To me, of course, that was the analogy you came up with. The first one that jumps to my mind is looking at

01:28:32.840 sort of the heyday of F1 in the late 80s and early 90s when the cars had become monsters. So if you look

01:28:41.520 at the McLaren MP44, which is regarded as either the greatest or second greatest Formula One car in

01:28:48.600 history, the only other car that could probably rival it would be the 1993 Williams. It had a 1.5

01:28:54.580 liter engine. So for any gearhead listening, that's a really tiny engine. Like your Prius probably has a

01:28:59.440 bigger engine than a 1.5 liter engine. And yet it produced 1,200 horsepower redlining at something

01:29:05.720 like 15,000 RPM. And I just remember being a boy, being so obsessed with that. Like, how could that

01:29:14.660 possibly happen? And I remember looking at my dad's station wagon, which had like a five liter Chevy

01:29:21.140 block in it and thinking, how is this thing so inferior? But in the end, like you can engineer, I mean, I

01:29:28.520 think the technical term is you can engineer the shit out of anything, right? That would be.

01:29:32.640 Exactly. And that's exactly what it is. And it's very analogous to the Formula One. And a lot of

01:29:37.660 people kind of, they're more into the bragging rights of like the bigger, the bigger, the bigger.

01:29:41.800 It's not always bigger. It's kind of like, if you really understand what you're doing and want to

01:29:45.680 get underneath the hood, you can take that 1.5 liter engine, you can put the turbochargers on it. You can

01:29:51.320 put, you know, like the multiple valves and everything to actually get the torque and horsepower you want

01:29:57.120 out of it. But most people don't think that way. They think bigger is better.

01:30:00.760 How did you do all this tinkering? Because you basically have here in Vancouver, a piece of

01:30:08.740 hardware that doesn't exist anywhere else in the world. And you've layered on top of that software

01:30:13.760 that is now also in a league of its own. And that's even more complicated. I just am interested in this

01:30:19.480 hardware. I mean, one of the things that I send a number of patients here, and usually the first

01:30:24.200 question they ask is, why are you sending me to Vancouver? Like I live in fill in the blank,

01:30:28.860 some city in the United States, we do everything the best. There's no way, Peter, you're telling me

01:30:34.760 there's an MRI in Vancouver. In fact, I told my parents the other day, I was coming to Vancouver

01:30:39.420 to get the MRI and they live, you could almost hear them look at me through the phone. Like,

01:30:43.860 why are you coming to Canada? So, I mean, not to toot your horn too much, but you're doing something

01:30:50.480 very different, Raj. So, how did you go through this process of tinkering with the 150 variables

01:30:58.260 at your disposal to come up with this super custom pimped out hardware that doesn't resemble

01:31:05.700 anything else on the planet? So, part of it is it's like the biggest problem with me is that I'm an

01:31:11.260 engineer in medicine. And so, when I kind of go through, it's like I say, well, what do I want to know?

01:31:16.980 What do I want to see? And how do I make it work? And so, I kind of start at sort of the back end and

01:31:21.220 say, okay, what I want to know is exactly what's going on everywhere in the body and what are the

01:31:26.160 different sequences that are going to get me there? And then I kind of work backwards. And then this is

01:31:30.820 where you put your engineering or physics hat on and you actually wind up starting to talk to like

01:31:36.340 the MRI physicists who are, there's a plethora of them, they're almost everywhere. And they love to

01:31:41.660 talk to doctors. But just most of the time, doctors can't talk to them or vice versa. And

01:31:46.900 that's when you start to really kind of get under the hood and really kind of understand how to make

01:31:50.540 this work. And then you travel around to different academic centers or different places and you wind

01:31:55.840 up spending time with the people who actually really understand the hardware. And those are mainly

01:32:00.960 the physicists, very similar to a mechanic. It's like, if you want to figure out how to make your

01:32:05.900 five liter behave like a thousand horsepower, you talk to the mechanics who know how to do it.

01:32:10.720 But don't try it yourself, right? They've already been there. They've seen it. They've worked on

01:32:14.480 things forever. And that's actually how a lot of it happens. And so we would actually be having some

01:32:19.460 of the top MRI physicists sort of saying, hey, Raj, can you test this? We wrote this sequence.

01:32:24.300 And a sequence is basically like a filter, like a T1 or a T2 or a fat saturation. Can you try it out

01:32:29.960 and kind of see how it works? So we'd go and try it out, usually on me as a subject, because I also know

01:32:34.760 what I'm looking for. And then we'd have a feedback loop, one or two of my MRI technologists here who would

01:32:39.520 actually sort of push the buttons and run the machine, and we'd see what it would give me.

01:32:43.720 And I knew from all the other imaging training that I would have, what I would be looking for.

01:32:47.980 So I'd put my nuclear medicine hat on and say, okay, from a functional point of view,

01:32:51.620 this is what I want to see. Then I put my radiology hat and say, this is what I want to

01:32:55.780 see from an imaging perspective. And let's define that a little bit. I mean,

01:32:58.940 objective is important. So when you went about this process, what were you optimizing for? Were you

01:33:03.840 interested primarily in detection of cancer? Or what was the problem you wanted to solve clinically?

01:33:08.580 The first thing I wanted to do is, so when we talk about the nuclear medicine, so when we're

01:33:12.720 injecting somebody with radioactivity, we want to basically optimize that dose that we're giving

01:33:17.280 to somebody. So that means we want to cover everything from head to foot. We don't want

01:33:20.780 to just look at an individual body part, and we want to see how it all works. Then we go to

01:33:25.060 radiology, typically we're just doing a snapshot of an individual part, like a head or a neck or a

01:33:29.860 torso. And so I kind of thought, well, of all these MRI machines that are out there, all they're doing

01:33:34.200 is looking at individual body parts. And I thought, if I customize this hardware with a few things that

01:33:40.240 might allow me to move people around while they're on the table, while they're laying there, will I

01:33:45.740 maybe be able to connect the head with the neck, with the chest, with the abdomen? And so I actually,

01:33:51.100 when I bought the hardware myself, I kind of put together probably about 50 options that the vendors

01:33:57.620 kind of said, you're crazy. We've never seen any of this stuff. Just with the thought that, you know,

01:34:01.620 I think this might work. This is where I was kind of looking at all the different options and kind of

01:34:05.600 knew what they could possibly do. And I thought, if I build this hardware this way, will it work?

01:34:13.080 And it was a complete guess, but an educated guess. And then once I put that hardware together, then we

01:34:18.600 started to test and test and did more and more. Then we would start talking to more physicists. The

01:34:24.560 vendors themselves actually have a lot of good resources with very technologically capable people.

01:34:30.380 And when I started to put this together, then I kind of thought, okay, what would I want if I'm a

01:34:35.760 patient? What would I want to know? Well, number one, I'd want to know that my brain's okay. I'd want

01:34:40.440 to know that the arteries in my brain are okay. They're not going to rupture. And then I basically

01:34:44.680 want to say, whenever I go into any test or anybody goes into the doctor and gets any kind of imaging

01:34:49.440 test of any kind, the first question out of their mind is, do I have cancer? Yes or no. And in nuclear

01:34:54.880 medicine, most of the tests are binary. We can actually answer that with a yes or no. In radiology,

01:34:59.820 it's not so clear. It's kind of like, maybe. And you're actually kind of playing more statistics,

01:35:04.700 like probably not. And I thought, how do I marry these two together? And this is where an MRI becomes

01:35:09.940 a beautiful machine. And the fact that it actually allows you to take that yes or no binary answer of

01:35:14.800 functional nuclear medicine and combine it with the anatomic localization and understanding of

01:35:20.060 tissue types that radiology has. And so I merged those two together on the one machine.

01:35:25.460 And what is the functional arm that you've brought into it?

01:35:29.360 Yeah. So the functional arm is actually an area that's actually growing a lot in academia. It's

01:35:33.220 actually called DWI or DWIBS, which stands for diffusion weighted imaging with background subtraction.

01:35:38.760 And so what we're doing with DWIBS is that we're actually looking at water at two points in time.

01:35:44.560 And so we're actually looking at it within about 60 microseconds of water motion. And so by doing that,

01:35:50.340 what happens is that you first look at water at one point in time, and then you look at it at the

01:35:54.960 second point in time. And if it hasn't moved, it's because it's not allowed to move. It's effectively

01:35:59.100 trapped between walls. And so that could be because of the fact there's a tight cell membrane,

01:36:04.580 or it could be because components of a cell are preventing that water from moving.

01:36:09.120 What's the time gap between those two samples?

01:36:11.500 Six microseconds.

01:36:12.700 Wow. And so when you see that something isn't moving, when you see that water is being forced to be

01:36:19.200 stationary, as opposed to moving according to the stochastic prediction you have,

01:36:24.500 what do you infer clinically about that tissue?

01:36:27.240 So as soon as you start to see that water is being prevented from moving, that means that there's

01:36:31.180 basically going to be a high density of cells there. So a big cluster of cells. And so it's very

01:36:36.940 much like, I actually call it the lump detector. So it's basically like they tell women for breast

01:36:41.100 cancer, feel for a lump. And basically when you're feeling for a lump, that's hard spot. And so the reason

01:36:46.000 it's hard is because you have this increased cellular density. And so with that increased

01:36:50.140 cellular density, that's where water is restricted from moving. And that's what DWI or diffusion

01:36:55.260 weighted imaging does. And why it's called DWI is because the fixed law of diffusion basically means

01:37:01.080 that the normal diffusion of water, allowing it to move a lot, is being completely restricted.

01:37:05.100 And it just doesn't have the ability to move.

01:37:07.180 You know, I usually tell patients about this part of the MRI, telling them something that I think many

01:37:12.320 people outside of medicine would find surprising. When you do what's called a laparotomy, when you

01:37:17.280 open up a person's abdomen, and let's say they have colon cancer. So the colonoscopy has confirmed

01:37:22.580 that, of course, you have a biopsy. So now you're going in to do an operation to remove their colon,

01:37:27.260 but you still have another step along the way, which is you have to complete what's called staging.

01:37:30.740 You have to ask the question, has this cancer spread? So usually the first thing you're doing is

01:37:36.460 you're running your hand along parts of the abdomen. You can't even visualize the entire surface of

01:37:41.560 the liver, even behind the liver, something you won't see. It's actually unmistakable what cancer

01:37:45.920 feels like because it is so in contrast to what normal tissue feels like. And even the colon,

01:37:53.720 before you cut it out, if you just reach in and pick up a piece of the ascending colon,

01:37:58.580 it's not remotely subtle where the cancer is. It's entirely obvious just based on the firmness of

01:38:05.280 the tissue. And even when I talk to surgeons who operate on parts of the body that I didn't operate

01:38:10.220 on, such as the prostate or things like that, it's the same thing. And the example you give

01:38:14.420 with breast is perfect. And so it really is this amazing ability to pair exquisite anatomic detail

01:38:22.200 at resolutions that we'll discuss, but basically now approaching one by one by one millimeter

01:38:27.720 resolution anatomically with now this functional property of firmness. Is that an accurate statement?

01:38:34.940 It definitely is. So we're actually combining the anatomic and functional. And that's where just like

01:38:39.580 the PET CT where that famous one plus one equals three is exactly what we're doing. And the beauty

01:38:45.160 of it is there's absolutely no radiation. So there's no risk. So the risk, it seems, because one of the

01:38:51.280 things I do explain to patients is who should consider doing this, right? And my view with cancer

01:38:56.620 screening is it's just a very personalized decision. I don't think it is for everybody to do the kind of

01:39:01.960 stuff that I do or that a number of my patients who you've now scanned have done over the past three or

01:39:06.900 four years. And when people say, is there a harm of doing this outside of the probably rare, rare event

01:39:13.740 of getting a migraine headache triggered by the magnet, I say there is a harm, the harm of a false

01:39:18.060 positive. The harm is that we see something that turns out in the long run to not be cancer, but in

01:39:24.360 the process of going down the advanced diagnostic pathway to get there, you are either physically harmed

01:39:32.680 by something we do subsequently, for example, another biopsy or a biopsy or a subsequent biopsy.

01:39:38.840 And of course, the emotional toll it takes on you to see a shadow in this part of your body and have

01:39:46.040 to sit there and have a discussion about what it could be, what it's probably a cyst, but it might be

01:39:51.160 a tumor. We probably need to do a follow. I mean, to me, this gets back to where we were a while ago on

01:39:57.260 the discussion of sensitivity and specificity, right? So if somebody came along and said, I have

01:40:01.880 a test that is a hundred percent sensitive, but only 50% specific, I'd throw it in the waste

01:40:09.560 basket. I mean, it would serve virtually no purpose for reasons I could walk the listener through in

01:40:14.660 terms of positive and predictive value, negative predictive value. So when you think about the

01:40:20.100 sensitivity and specificity of the technology that you've developed here, which again, we haven't even

01:40:25.340 really done justice to getting into, we've talked a little bit about the hardware. We haven't even

01:40:28.880 got to the software. I'd love to talk about that a little bit. Do you think about sensitivity and

01:40:32.940 specificity by tissue type, by cancer type? How do you, in your mind, wrap your head around that?

01:40:39.940 So typically the way I kind of look at it is really almost organ by organ. When we're actually kind of

01:40:44.380 going through, for example, in the liver, basically the simple thing we want to kind of know,

01:40:48.580 is there a problem? Yes or no. Like really that's the simplicity that the average person,

01:40:53.580 and I put myself in that category wants to know, do I have a problem to worry about?

01:40:58.060 And that's where I kind of say, by combining this functional as well as an anatomic imaging

01:41:01.480 together, we're actually really able to nail that down. So in our thousand people that we did,

01:41:07.300 the fascinating thing about this is all these people, we actually followed them up and actually

01:41:10.920 talked to them and kind of found out what happened. And so we actually had two false positives. So these

01:41:15.480 were two people where we kind of thought, okay, there's a problem here that you need to

01:41:18.660 get addressed further by their further imaging and see what's going on. And of those two false

01:41:24.300 positives, one was a male with asymmetric breast tissue. So he actually just had one-sided breast

01:41:30.200 tissue. We didn't know what it was. It's like-

01:41:32.480 So meaning you, he came out of the MRI and you believed he had breast cancer, which

01:41:36.140 most listeners might think is odd, but it turns out men can get breast cancer. It's just

01:41:40.360 virtually and exceedingly rare.

01:41:42.440 Exactly.

01:41:43.040 But that's basically the DWI showed a difference in density between-

01:41:46.660 One breast and the other. And so we're like, why is that? Most of the time when

01:41:50.600 men actually have gynecomastia or they have breast tissue, it's usually bilateral because

01:41:54.640 it's hormonal. But to have unilateral is somewhat odd. And so as a result, we sent this person

01:42:00.640 to an ultrasound, which is a commonly done, and they too had no idea. And so they actually

01:42:05.720 also made the call to biopsy and it came back as normal breast glandular tissue in a male.

01:42:10.080 So let's talk just about the harm there. So emotionally, that man probably spent a series of,

01:42:15.360 well, if it was in Canada a year, if it was in the United States a week, being stressed

01:42:19.900 out about this, or I just can't help but take digs at your healthcare system as you take digs

01:42:24.460 at ours. No, I'm totally teasing. So there's a legitimate emotional strain here, which I

01:42:28.940 don't think anybody who's known someone who's gone through that or who's gone through that

01:42:32.320 themselves, you just can't deny this. And I've not lived it personally, but I've seen

01:42:35.700 it and it's very difficult. And then secondly, he had to get a procedure. He had to get a needle

01:42:40.800 stuck into his breast tissue. And look, that's on the scale of procedures, that's still pretty

01:42:45.340 minor, but it's not trivial. So what was the second case?

01:42:49.780 And so the second case was actually a woman who basically had a seatbelt injury to her

01:42:54.060 breast and actually wound up having an unusual scar that had actually trapped fluid in it.

01:42:58.420 And so this person actually should have, but never did actually have any mammograms because

01:43:02.660 that would have led to the same conclusion that we don't know what this is. So that was the

01:43:06.920 second case where it's kind of like, there's something unusual going on in this breast. We

01:43:10.040 don't know what it is, but we didn't actually have any proper history from her. And so we-

01:43:15.220 How old was she?

01:43:15.880 She would have been late fifties.

01:43:17.880 A woman in her late fifties had never had a mammogram?

01:43:20.300 Surprise. It's out there.

01:43:21.720 Even in Canada?

01:43:22.660 Even in Canada. Yeah.

01:43:24.160 That in and of itself is very hard to believe in a country where-

01:43:27.060 It's free. Yeah.

01:43:27.880 Yeah. And then what would bring that patient in to get a whole body MRI?

01:43:32.340 The person actually kind of felt that I just want to know where I'm at. And you know,

01:43:36.140 it's like, I don't believe in the radiation from mammography. And I keep trying to tell

01:43:39.940 them, look, it's so minimal that the benefit outweighs the risks. But they're like, I want

01:43:44.400 the MRI instead because number one, patients kind of know that it's the most detailed exam

01:43:48.240 you can get. And as well, there's no radiation. So we actually had the patient and it's kind

01:43:53.380 of like, yes, there's this, I don't really know what this is going on here. So we sent

01:43:57.660 them off to a facility that does nothing but women's imaging. And they too didn't know what

01:44:02.620 it was. And so they stuck a needle in it and it actually came back as a trap.

01:44:06.620 Something fibrous.

01:44:07.780 Exactly. It was actually trapped scar tissue. And at that point when we'd spoken to a woman

01:44:11.540 afterwards, we're like, did you ever have trauma? Like, why would you have scar to that

01:44:15.860 area? And it's like, oh yeah, I was in a bad car accident. And that was it. It was

01:44:18.860 a seatbelt scar.

01:44:19.580 Those are two pretty interesting false negatives, right? Both breast men, women. I would have

01:44:25.320 always maybe guessed or been most concerned about the false positives that occur deeper in

01:44:31.180 the body. I always tell patients, the call I'm always most afraid of getting is the little

01:44:36.240 shadow in the pancreas where you just don't know, is this an adenocarcinoma of the pancreas?

01:44:41.340 Which of course is something that if you're lucky, you can resect it. And if you're even

01:44:45.960 luckier, you can survive it. And I guess that's the only shot you're going to have at surviving

01:44:50.300 pancreatic cancer is an incidental finding. Really don't think anybody that presents with

01:44:54.180 pancreatic cancer is going to survive it, at least not as an adenocarcinoma. But then you

01:44:58.840 worry about, well, what if it's something that turns out not to be cancer? And you hear

01:45:02.920 these horror stories. There's a very famous one, I believe at Stanford several years ago

01:45:06.060 where a woman went to a sort of drive by CT clinic, got a CT scan, showed something in

01:45:11.760 the pancreas. She ended up going to get a biopsy. And I believe it was an ERCP. I did

01:45:17.260 biopsy. One complication led to another, led to another. She died of sepsis. And by the way,

01:45:22.240 it turned out she didn't have pancreatic cancer. I don't know that firsthand. So that might

01:45:26.120 be a bit of a wives' tale stretch, but certainly that's a very popular story in the Bay Area.

01:45:31.280 But it's the cautionary tale, right?

01:45:32.880 Definitely it is. And that's kind of where the real value of actually having MRI as opposed

01:45:36.520 to CT comes in. And the fact that when we're looking at organs, particularly like the pancreas

01:45:41.040 or any of the visceral solid organs, we're looking at about seven different filters, looking at

01:45:45.720 different ways, top to bottom, front to back, to really be able to see what's going on. And

01:45:49.940 that's where what we call contrast density becomes really important in the fact that

01:45:54.100 when we're actually looking at the pancreas in particular, we can actually pick out the

01:45:58.920 pancreatic duct as well as a bile duct and be able to see that just standing out against

01:46:02.740 the rest of the organ. And one of the first, most common things that pancreatic cancer likes

01:46:06.860 to do is to actually start to block that duct. And so that's why when ERCPs are done, they're

01:46:11.700 actually looking for cells of pancreatic cancer. And an ERCP is basically when they go down

01:46:16.760 into the mouth and the esophagus, and they actually go and take a trace of fluid from the pancreatic

01:46:22.380 bile duct. So one of the other things that kind of amazes me when we go through these

01:46:27.700 images here is when we've talked a little bit about the hardware, though, actually, I

01:46:31.620 kind of want to come back and ask more hardware questions, but it's almost like you've created

01:46:34.900 your own software now as well. I had never seen this. Is it commercially available to have

01:46:40.380 that rotating diffusion weighted image map? Is that a commercially available piece of

01:46:45.980 software? Or did you guys make that?

01:46:48.300 We actually built that as a display tool. And that's actually effectively taking a page

01:46:53.520 out of the nuclear medicine positron emission tomography or PET CT handbook. And the reason

01:46:58.460 why we actually put it together is because it actually allows you an effectively a pretty

01:47:02.580 efficient viewing to be able to see what's going on through the entire body. It's almost

01:47:06.640 like making a transparent person. And where basically any of the black spots that stand out

01:47:11.140 would be the hard spots or the firm areas.

01:47:14.140 And the one that we just looked at earlier, I remember you saying that, if I understood

01:47:20.660 you correctly, one of the advantages of using a quote unquote low power magnet like you're

01:47:25.300 using is you don't have any of the gaps in the spine. You've got this, everything that

01:47:30.080 you showed on that rotating diffusion weighted image, you had the dark brain, obviously full

01:47:35.200 of firm fluid. And then you have this dark, beautiful tail coming out of it, which is the spinal

01:47:41.020 fluid, but it was perfectly smooth.

01:47:43.020 Right. And that's actually one of the important things because magnets actually have a lot

01:47:47.120 of homogeneity problems, we call them. And the fact that you want it to be perfect so

01:47:51.880 that the field in between the top and the bottom and the left and the right are identical.

01:47:56.360 And as soon as you put a person in there that comes in various sizes and shapes, they actually

01:48:00.360 distort that magnetic field. And the sweet spot for the magnetic field is perfectly in the

01:48:05.180 center. And so when we put all these protocols and built all these things, we actually built

01:48:09.160 it for different body shapes and we can actually go and tune it for all these body shapes so that

01:48:13.980 the goal is that no matter what, when we're doing these rotating images, that it looks

01:48:18.000 like a normal person, not a segmented piece of a person.

01:48:21.220 Yeah. It's funny you say that you took a playbook out of the PET CT. That's exactly what it looks

01:48:24.920 like. It looks just like you're looking at the FDG PET juxtaposed with the CT.

01:48:30.020 Right. And that's the entire purpose. And so that's where I talk about the MRI being this tool that can

01:48:34.960 actually combine functional imaging of, in this case, DWI or DWI, diffusion-weighted imaging,

01:48:39.700 and the MRI being the equivalent to the CT, which is far more tissue weightings in detail.

01:48:46.340 It's where the one plus one equals three.

01:48:48.660 Now, if the radiation didn't bother you of a whole body PET CT, and of course it should bother

01:48:54.280 you because a whole body PET CT would be more than 50% of your, it'd be probably close to 80%

01:49:00.820 of your annual allotment of radiation, right? If not more. It'd be quite high. Yeah. And depending

01:49:05.700 on if you live at altitude or where you live, right? So if you're at sea level, you'd probably

01:49:10.340 be allowed one a year maximum, if not one every two years. Yeah. So putting aside that issue,

01:49:17.940 which is not trivial, what advantage do you think that the MRI with DWI has over the PET CT? And where

01:49:26.840 does the PET CT have an advantage? So if you go either by histology, by tissue, by tumor type?

01:49:32.260 With the PET CT with radioactive glucose, one of the areas that actually doesn't work very well at

01:49:37.160 all is the brain. And the MRI is always known as like the best image of the brain. The PET CT,

01:49:43.220 you can actually miss things in the brain because what you're looking for with the radioactive glucose

01:49:47.020 is areas of increased glucose utilization. And the brain in particular is a glucose,

01:49:53.160 that's all it can use unless you're in ketosis. And then the other problem is that the glucose is

01:49:58.860 then excreted by the kidneys. And so the kidneys now become difficult to see because they're actually

01:50:03.940 full of glucose. And as is the bladder, you can't see a thing in the bladder because it's full of the

01:50:09.380 accumulated glucose. As well in the prostate, prostate is very, very poorly perfused. And as a result,

01:50:15.480 it doesn't get a lot of glucose coming to it. And so PET is actually almost entirely useless with

01:50:20.900 FDG and looking at the prostate. Diffusion weighted image of the prostate coupled with

01:50:27.100 the more advanced molecular tests, the 4K as an example of a blood test, in my mind have

01:50:33.220 totally revolutionized the way we think about prostate cancer. So we now have a blood test

01:50:38.520 that produces much better resolution than just a PSA. But more importantly, we have this MRI.

01:50:45.340 And even in a practice as small as mine, I have had two patients for whom PSA is high,

01:50:54.180 4K comes back high. So these are patients who now have a 20% chance, maybe 16, 18, 20% chance

01:51:04.160 of having cancer in their prostate or having metastatic cancer over the next two decades.

01:51:08.940 That's basically what the high 4K tells you. In the olden days, we would have just biopsied them.

01:51:13.620 And now we run them through this MRI. And the answer is, nope, that's totally fine.

01:51:19.560 Right. And then that's actually where MRI becomes very, very powerful, particularly with the DWI. And

01:51:23.660 I think in many countries, it's now coming to US and becoming more popular. But in Australia,

01:51:29.520 in Europe, particularly UK, Scandinavia, it's actually the de facto standard. Basically,

01:51:34.660 almost all men are actually getting screened with MRI.

01:51:37.800 Wait, wait, wait. Did you say in Europe and Australia?

01:51:39.840 Yeah. They're actually doing it with a DWI. How is that even possible that countries with

01:51:45.400 single-payer socialized medicine could use an MRI for a screening tool? I mean, that would be

01:51:50.900 unheard of in Canada, wouldn't it?

01:51:52.660 Well, we just don't have the access to the number of machines here. But I think one of the things

01:51:56.520 that people are actually finding with screening for the prostate is that all men are either going

01:52:00.680 to die with or from prostate cancer. And so you really want to be able to separate those out.

01:52:04.520 And up until MRI with DWI came into effect, there was no real way to do that. So you'd be doing PSA

01:52:11.080 or 4K. And all that would do was say, yeah, there's an increased risk of something going on,

01:52:16.320 but is it going on? So PSA in particular, it can actually be elevated for three reasons. One,

01:52:21.220 prostate cancer, the other one, inflammation or prostatitis, and then the third being enlarged

01:52:26.080 prostate. And so if it was up for any of those causes, then you would actually go and take a

01:52:30.020 biopsy, which actually comes with risks. But the whole idea is that all the staging was based on

01:52:35.340 that tissue sample. Whereas what's happening very similar to the breast is kind of like a lot of

01:52:40.560 people are saying, well, treating this is like a big deal. I don't want the biopsy and people get

01:52:45.320 a choice of what they want to do. And a lot of men are saying, look, I want to see what's going on.

01:52:50.000 I want to see if there's going to be a change. And if it actually starts to grow or something's

01:52:54.100 growing in the prostate at an accelerated rate, that's when I want to deal with it. Whereas if it's

01:52:58.200 actually just there and holding still and not changing much, I'm not going to worry about it

01:53:02.160 because something else may take me first. Is DWI going to have the same effect on breast

01:53:06.860 cancer? In other words, if you could put resources aside for a moment, if a woman could have a

01:53:12.260 mammogram and a DWI MRI, and it's important to point out that you can't eliminate mammography

01:53:18.960 because we're going to come to this, but MRI has its own blind spots and small calcifications would be

01:53:24.460 a blind spot. But with those two tests, mammogram and DWI MRI done the way you guys are doing it,

01:53:32.500 not just off the shelf, are you going to miss breast cancer in those situations?

01:53:36.840 Pretty unlikely. And actually there's been a couple of really nice big studies that have

01:53:40.240 finally started to come out from groups at UCSF and as well at Sloan Kettering,

01:53:44.600 Memorial Sloan Kettering in New York that have actually shown that.

01:53:47.160 You did your fellowship at Memorial, didn't you?

01:53:49.240 No, I was actually there for quite a while actually doing PET-CT.

01:53:51.980 Okay. Yeah. Yeah. Amazing institution. Absolutely amazing. But what's actually come

01:53:56.540 out of the MRI group is that basically if you actually use DWI with MRI, you actually are as

01:54:02.620 sensitive as giving a contrast injection breast MRI. Wow.

01:54:06.700 But that's diffusion done right. And the problem is out of the box, it's not always done right.

01:54:11.040 Yeah. That's the challenge I think for the patient, right? Is the patients, look, you kind of need a PhD

01:54:16.480 in physics. Let's be honest to really understand the nuances of MRI. I consider myself pretty

01:54:22.720 technically smart when it comes to the physics of this stuff. And I would not feel competent to

01:54:28.920 try to differentiate or even parse out the differences between scanners. In fact,

01:54:35.300 anytime I'm sending my patients to a scanner, I sort of have to rely on other people to help me. I have to

01:54:40.180 reach out to experts and say, Hey, my patient has to get this scan done in New York or in San Francisco

01:54:45.560 or LA. Is this the best we got? If they're not going to get on a plane and go to someplace where

01:54:49.880 I know exactly what they're going to get. So that's a significant challenge, right? Because there's

01:54:54.360 people are going to listen to this and think, okay, well, as long as it's diffusion weighted imaging

01:54:58.180 MRI, it's perfect. Yeah. And that's actually the biggest problem with MRI. Like earlier on,

01:55:02.300 we talked about CT having units of Hounsfield, like Hounsfield units to actually sort of calibrate

01:55:06.580 and standardize them. Unfortunately, MRI has no standardization whatsoever. And there's actually

01:55:12.380 a movement called Kiba or Quantitative Imaging Biomarkers Alliance. It's a component of the

01:55:17.460 Radiology Society of North America that's actually really trying to push to standardize the amount

01:55:22.460 of signal to noise coming off of MRI machines with the goal that if you actually get a scan at one site

01:55:27.580 or another site, the image quality is the same. Right now, it's really not that at all. And it really

01:55:34.680 is sort of caveat emptor lookout. So right now, if somebody goes to Shreveport and gets a 256 slice

01:55:43.480 CT scanner on a Siemens, pick your favorite model, and they do the same thing in Seattle,

01:55:50.260 you can share those data across radiologists and it can be made to look identical. Your acquisition is

01:55:57.000 the same. Relatively. So what I mean by that is that the actual amount of signal on your film or on your

01:56:02.060 screen, water is going to be zero. And so the Hounsfield unit is always going to be exactly

01:56:06.960 calibrated. So what's the opposition? I mean, that seems like a no-brainer. What is the opposition

01:56:11.020 to doing this with MR? It actually relies on the vendors coming together. And a few years ago,

01:56:16.080 we actually spoke with a bunch of the vendors and saying, you know, look, why don't you guys do this,

01:56:19.340 particularly for a diffusion where it's actually so important because this is such a new and powerful

01:56:23.860 sequence. And they all kind of came back and said, well, you guys write a white paper and then

01:56:28.860 we'll implement what you said. And this is fortunately starting to move forward with this

01:56:33.400 organization, Kiba, out of RSNA. And they're doing it organ by organ. So there is a bit of a

01:56:38.080 standardization for prostate, for liver, and as well, breast is now coming out. And that was actually,

01:56:44.860 the breast was led by a group out of University of Washington. And the overall leader for this is a

01:56:49.980 person named Michael Boss out of, now he's the American College of Radiology. And this needs to

01:56:54.700 move forward because otherwise people have no idea what they're getting.

01:56:57.880 And it's really sad because even if you walk down the streets, for example, in a place like New York,

01:57:02.420 I used to live literally 20 feet from a MRI shop. And they had all these sort of images and what I

01:57:10.540 consider just sort of bogus propaganda all over their window. Like why go down to Memorial and get

01:57:15.420 your MRI there when you can come here and do a standing MRI? It's comfortable, it's fast,

01:57:20.280 and blah, blah, blah, blah. And I'm thinking to myself, and it sucks. So, and again, I don't know

01:57:24.960 that that exists in Canada, but it's certainly in the United States, there's a bit of a cottage

01:57:27.940 industry around one-stop shop scanners that I think patients just don't know what they're getting

01:57:33.820 into, right? Right. And I don't know how well it works in the United States, but it really is.

01:57:38.140 It's like because of this lack of standardization in the field of MRI. And it's really unfortunate

01:57:42.520 because this is, of all the imaging tools, it's the most powerful, but it really does need

01:57:47.700 stability for standardization. And it may also sort of be the fact that in order to make it

01:57:52.260 standardized, you need to get the physicists together with the radiologists who are basically

01:57:56.820 the eye of the final image. And quite often that doesn't happen because of the language barrier.

01:58:02.740 Right. And of course it's not, it's the language of physics is the language, you know.

01:58:06.480 Going back to the hardware for a second, where do you see things evolving? In other words,

01:58:10.520 if you had to project where you think you, with the right technology, would like to see this in

01:58:16.480 five or 10 years, what could make this better?

01:58:18.400 What I'd actually like to do, and when you see sort of what we're doing, the speed with

01:58:21.820 which we do and that the detail and resolution that we're able to acquire in about 55 minutes

01:58:26.420 is really unprecedented anywhere. But I know from a physics point of view that I could speed

01:58:30.440 this up further. Right now, everything we're doing is perfectly within FDA specifications.

01:58:34.640 There's nothing outside of the box, but I really would like to push this and go outside of the box.

01:58:39.800 And what that really requires is a lot more computational horsepower.

01:58:43.220 It's really difficult to do all that computation within a single CPU machine on site. Whereas I

01:58:49.560 expect that in the future as more as long computers get faster and faster, you can actually do a lot

01:58:54.640 of this stuff computationally and make it much faster. And the goal would really be to actually

01:58:59.240 have these scans that we're doing, you know, under half an hour, even faster. And it can be done

01:59:04.180 from a physics point of view. It's not a technological barrier.

01:59:06.560 Even with the magnet that you have, even at 1.5. Yeah.

01:59:10.420 Now I had my scan today, a friend had a scan today. And one of the things that people always

01:59:14.800 talk about when they're done these scans is, my God, it gets hot in there. What is it about a

01:59:19.120 whole body MRI, even one that's done in as short a period of time as 55 minutes, that makes it so

01:59:24.760 uncomfortable by the end in terms of body temperature?

01:59:27.280 It really has to do with the amount of energy that's being absorbed. So what we're doing with the

01:59:30.860 electromagnetic field is you're actually putting in radio frequency, right? And that radio

01:59:35.660 frequency is always also coming out. And that radio frequency is the same thing as a cell phone

01:59:40.620 would get. That's called SAR or specific absorption ratio. And the hydrogen ions are basically moving

01:59:48.200 around and that's basically effectively heating you up. So it's not quite like a microwave, but you can

01:59:53.120 actually think about it as a microwave. The wavelength is exactly very similar to an AM radio. And so your

02:00:00.260 cell phone is typically far more powerful. So as an example, I went and did the calculation a while

02:00:05.100 ago on how much SAR we're actually putting into a whole body prenuvo scan. And it's equivalent to

02:00:11.060 talking on a cell phone for about four hours. Interesting. But it concentrates it across your

02:00:15.920 whole body. Right. Yeah. It's funny. Every time I get an MRI or a whole body, the first 30 minutes,

02:00:22.260 I'm like, yeah, it's not so bad. And then the last 10 minutes, I'm like, get me out of here.

02:00:27.180 Right. And we've actually done it that way on purpose because we kind of know that the way we kind of run

02:00:31.520 our sequences or filters together is that we've actually kind of want to make, get as much

02:00:35.760 information in such a way as possible. Yeah. You do the head first and that's not nearly as much

02:00:40.080 as doing the abdomen and thighs or probably where a ton of that heat gets generated, right? And it's

02:00:44.200 sort of knees to abdomen. Right. And so when you look at basically the overall blood flow, which is

02:00:48.660 what's cools your body, well, the brain takes 20% of your cardiac output. So it's basically like this

02:00:53.360 big cool, it's this big heat sink, right? It's a, just cools everything away. Whereas when you get

02:00:58.320 down and lower it to the legs, which you're all muscle, it's going to heat that up. And so that's

02:01:02.560 why we figure out how to orient these and what organization to make, what plan of sequences to

02:01:07.460 make it not as uncomfortable. Commercially, what's the best off the shelf scanner that could come

02:01:12.820 close to producing the resolution you're producing, which is it fully isotropic? It's isotropic in the

02:01:19.540 brain, but in the rest of the body, we're actually doing more conventional clinical images.

02:01:23.720 Tell us what isotropic is. I'm sorry. I should have clarified that.

02:01:26.100 Sure. So what isotropic basically means is that we're basically slicing you in cubes. So a

02:01:30.340 one by one by one millimeter cube, for example. And then the power of actually doing that one by one

02:01:35.300 by one is that you can now look at things again in three dimensions in any direction you want.

02:01:40.040 The detail and resolution is perfect. And so that's what isotropic means. Whereas MRI typically can't be

02:01:47.060 done isotropically just because of time. You want to cover as much as possible, but you want what we

02:01:52.680 call in-plane resolution, which is how you orient the first gradient to have the highest amount of

02:01:57.640 detail. And then you typically will take a perpendicular view. And that's why when you

02:02:01.960 do these rotations off your sagittal plane, the resolution deteriorates wildly in conventional scans,

02:02:08.640 right? Right. Exactly. And so the diffusions that we're doing are done isotropic, but unconventional,

02:02:14.020 not often. So that's why when you rotate, and hopefully in the show notes, we'll be able to

02:02:18.340 get some videos. Like we'll literally just show what it looks like. Cause I know this is a

02:02:22.540 kind of a difficult discussion to have without being able to kind of picture for us. It's easy,

02:02:27.340 but I think we want to make sure that the listener can see this. That's why you don't see any distortion

02:02:31.440 when you rotate the DWI. So is there a commercially available scanner that can do that?

02:02:35.640 Not yet. Someday, not yet.

02:02:37.620 And so basically your goal was just to be, I don't know how to put a timestamp on it. How far are you

02:02:42.040 ahead of what's happening conventionally? I mean, four years ago, you were doing things that I still

02:02:47.500 don't see any scanner doing in the country. And I see the best scanners.

02:02:52.460 Thank you. But, uh, well, for example, like when I have a patient that went to one of the most famous

02:02:58.400 hospitals in the country and had a dedicated prostate CT scan, it took 40 minutes just to do

02:03:04.940 the prostate. Dedicated prostate MRI. MRI, not CT. Yeah. So he had dedicated prostate MRI, took 40 minutes.

02:03:11.280 It was on a three or four Tesla magnet. And so he spent two thirds of the time to get a slightly

02:03:18.520 inferior image by resolution, especially on the DWI was far inferior to what you were doing on the

02:03:25.120 whole body. It all comes down to basic engineering of signal to noise. If you can make all the hardware

02:03:31.160 that you have really sing, your signal to noise is so much better. And then you can basically dial it

02:03:37.000 to effectively where you want. So if you need more and more signal, depending on your machines,

02:03:41.000 I guess, horsepower and coil configuration just takes time. It's always a balance between time

02:03:46.260 and signal to noise. So where is machine learning going to come into the fold here? You actually

02:03:52.500 mentioned something to me recently that I had never thought of, which is it's really hard to throw a

02:03:59.500 whole body image at a machine and have it solve even the simplest problem that we take for granted

02:04:05.340 as a human, which is which one's the liver, which one's the kidney. And whereas if you weren't doing

02:04:11.280 a whole body, if you were just doing a liver image, that's an easier problem to hand the machine

02:04:16.060 because it already knows it's looking at liver. So, I mean, how far are we away from a machine being

02:04:21.460 able to help you do this? I think that there's actually a lot of tools that actually still need

02:04:26.020 to be written. So for example, part of that is to really kind of look at organs and isolate organs.

02:04:31.400 Conventionally, most imaging is done by body part. So head, neck, chest, abdomen, pelvis,

02:04:37.260 and not connected together. And so when you actually give a machine, okay, here's your brain,

02:04:42.700 it's actually able to kind of go through that relatively efficiently. But part of it really

02:04:47.120 comes down to building the tools to actually analyze whole body, like even the software to

02:04:51.620 view the whole body is exquisitely complex. But the difference is that when you actually start

02:04:56.640 to build these tools, it can actually start to help us narrow down what's going on.

02:05:00.680 And the goal is to really sort of have machines make radiologists more efficient. And also more

02:05:05.480 importantly, we never want to miss anything, right? And so, you know, we can't have a second

02:05:09.520 reader all the time, but if you have a machine being a second reader, you wind up actually training

02:05:13.820 that machine as you go along. And I think most people would be comfortable with that. And that's

02:05:17.480 what's actually done in mammography right now.

02:05:19.720 And so for, I mean, mammography obviously is like the tip of the iceberg, but you got to start

02:05:23.720 somewhere, right? It's very, I don't want to minimize because I couldn't read a mammogram

02:05:27.780 to save my life, but it's relative to what we're talking about. It's much simpler. It seems that

02:05:32.840 where the machine might first be able to make a dent in what you're doing is not the patient who gets

02:05:37.800 their first scan, but the one who gets the repeat scan. That seems to be paired T-test seems to be an

02:05:43.240 easier problem to solve.

02:05:44.980 And an important problem to solve because a lot of times for us as radiologists, like those studies,

02:05:49.660 we still do things the same way when we review them. Whereas when you actually go and you take

02:05:54.620 like a paired T-test, like you said, and you actually do a subtraction where you're looking

02:05:57.760 for the difference or that delta, it can actually stand out and become very, very simple and very

02:06:02.400 obvious. And once that subtraction is done, and I think that's where machine learning will actually

02:06:07.120 really help make us more efficient. And at the end of the day, we just don't want to miss anything.

02:06:11.700 Well, Raj, I've monopolized more of your time today, and that means that there've been

02:06:15.920 probably three or four fewer patients that have been scanned today. So I really appreciate you

02:06:20.520 taking the time. And I appreciate just all the time you've spent over the last three or four

02:06:24.160 years educating me. You're incredibly generous with your insights. And I constantly lob questions

02:06:29.620 at you and you've always got all the time in the world for me and by extension, my patients and all

02:06:34.760 the people that I hope to sort of try to educate with this. So thank you for the amazing work you're

02:06:39.020 doing here. And then also just for your generosity.

02:06:41.420 Oh, thanks. It's a pleasure. Love it. You need to visit more often.

02:06:44.260 Yeah. Next time you come down to California, we have Uber.

02:06:50.120 Thank you for listening to this week's episode of The Drive. If you're interested in diving deeper

02:06:54.360 into any topics we discuss, we've created a membership program that allows us to bring you

02:06:58.760 more in-depth, exclusive content without relying on paid ads. It's our goal to ensure members get

02:07:04.260 back much more than the price of the subscription. Now to that end, membership benefits include a bunch

02:07:09.760 of things. One, totally kick-ass comprehensive podcast show notes that detail every topic,

02:07:15.180 paper, person, thing we discuss on each episode. The word on the street is nobody's show notes rival

02:07:20.520 these. Monthly AMA episodes or ask me anything episodes, hearing these episodes completely.

02:07:26.640 Access to our private podcast feed that allows you to hear everything without having to listen to

02:07:31.920 spiels like this. The Qualies, which are a super short podcast that we release every Tuesday through

02:07:37.520 Friday, highlighting the best questions, topics, and tactics discussed on previous episodes of The

02:07:42.260 Drive. This is a great way to catch up on previous episodes without having to go back and necessarily

02:07:47.420 listen to everyone. Steep discounts on products that I believe in, but for which I'm not getting paid

02:07:53.360 to endorse. And a whole bunch of other benefits that we continue to trickle in as time goes on.

02:07:58.300 If you want to learn more and access these member-only benefits, you can head over to

02:08:01.900 peteratiamd.com forward slash subscribe. You can find me on Twitter, Instagram, and Facebook,

02:08:08.940 all with the ID peteratiamd. You can also leave us a review on Apple Podcasts or whatever podcast

02:08:15.440 player you listen on. This podcast is for general informational purposes only and does not constitute

02:08:21.020 the practice of medicine, nursing, or other professional healthcare services, including the giving of

02:08:26.500 medical advice. No doctor-patient relationship is formed. The use of this information and the

02:08:32.220 materials linked to this podcast is at the user's own risk. The content on this podcast is not intended

02:08:38.400 to be a substitute for professional medical advice, diagnosis, or treatment. Users should not disregard

02:08:44.880 or delay in obtaining medical advice from any medical condition they have, and they should seek

02:08:50.860 the assistance of their healthcare professionals for any such conditions. Finally, I take conflicts

02:08:56.580 of interest very seriously. For all of my disclosures and the companies I invest in or advise, please

02:09:02.620 visit peteratiamd.com forward slash about where I keep an up-to-date and active list of such companies.

02:09:32.620 Thank you.

02:09:39.340 Thank you.

02:09:40.340 Thank you.

02:09:48.560 Thank you.

The Peter Attia Drive - July 03, 2023

Cancer screening with full-body MRI scans and a seminar on the field of radiology ｜ Rajpaul Attariwala, M.D., Ph.D. (#61 rebroadcast)

Episode Stats

Length

Words per Minute

Word Count

Sentence Count

Misogynist Sentences

Hate Speech Sentences

Summary

Transcript