Phillip Koo: Hi, this is Phillip Koo, and welcome back to the Imaging Center of Excellence at UroToday. We're very fortunate to have with us, Dr. Peter Choyke, who's a senior investigator at the Molecular Imaging Branch at the NIH. Welcome Peter.

Peter Choyke: Nice to be here.

Phillip Koo: Now, we're going to talk about prostate MRI, and I know you're going to be speaking at the AdMetech meeting in September being held virtually. But, wanted to talk about prostate MRI. Years ago it was more of a technology that we used post-biopsy after a diagnosis. It seems like now the pre-biopsy type space is where it's infiltrating into, and it's really rapidly growing. So for all the listeners out there, can you just level-set for us where we stand today?

Peter Choyke: Sure. Yeah, you're totally right. It was introduced initially as a staging tool for people who were already diagnosed with prostate cancer. We actually had a lot of trouble actually seeing the cancers back in the 90s on that generation of MRI equipment. Several studies really showed that it wasn't all that good at staging. But the real revolution started when there was this idea of the index lesion that you didn't have to find every single prostate cancer, but you really needed to find the most aggressive one. Because that was the one that dictated the biology. Once that philosophy came in, then the idea that MRI could pick up some of the lesions, particularly the index lesion, started to take hold. The diffusion-weighted imaging in the mid-2000s really was a key development. Because at that point, you had another tool beside T2-weighted imaging and DCE to pick up cancers, and that really changed things.

So it was still just a diagnostic tool and people were starting to do biopsies inside the MRI gantry. We did that. Very cumbersome. Lots of people involved, at least when we did it. And, it was very uncomfortable for the patient plus the urologist did not like it. Let's be honest. At the time we were working on the idea of fusion biopsy, where you could do the MRI in the MRI suite, but do the biopsy and the ultrasound suite, wherever that might be, and fuse the MRI to the ultrasound. We developed that with Brad Wood and Peter Pinto here at NIH, and that worked. Because then you could direct the needles into the lesion. And then it became worth it to get an MRI. So that's really what stimulated the great growth. I would say 2013 AUA, was when Peter Scardino gave his presidential address to the urology community and said this is the way of the future. Almost overnight, I would say, people started doing it. We started getting feedback. Other companies started getting into the field. It really became something.

Phillip Koo: That's great. Now we're seeing a lot of use of this pre-biopsy, and it's really flourishing. And now discussions about using MRI before focal therapy. But it's interesting, maybe one of the limitations though is MRI doesn't pick up a lot of the other lesions, it really just focused on the index lesions. How do we reconcile that?

Peter Choyke: Well, I think the observation is that there are, if you look at a radical prostatectomy specimen from your average patient, you'll find a big lesion with a Gleason grade four, or a Gleason group four or three, and then you'll find a lot of little islands of tumor scattered around. That's just the way it is, it's a multifocal. But the thing that needs treatment is the big lesion. The other lesions don't really require treatment. That's probably the next revolution when people started doing active surveillance in patients instead of treating everything. So, it's almost as if it's a good thing that the MRI can't pick up these small, benign-acting tumors, that clears the picture for what actually needs to be done for the larger, aggressive tumor. So that needs to be biopsied. Depending on what that shows, the patient should go on for definitive treatment of some sort.

Phillip Koo: That's great. So MRI could really be used as that tool that identifies those significant tumors, and the fact that it can't detect everything might be a good thing. That leads to this space of potentially having MRI be used as a screening tool or a supplement to PSA screening or whatnot. How do you envision that might take form in the years to come?

Peter Choyke: Yeah. Just to finish the last thought. It isn't true that MRI picks up every significant lesion. There are significant lesions that MRI doesn't pick up. We've looked at those and they tend to be these sparse tumors that, they're aggressive, but their islands of tumors separated by normal tissue, or at least non-neoplastic tissue, and the overall density is not sufficient to pick it up on an MRI. That is a limitation of MRI. So far, it seems as if patients in that category do very well and they don't have the same urgency to treat as a patient who has a well-defined lesion. But it is concerning that the MRI is missing something that we call significant. So that is a recognized and controversial limitation.

But in terms of screening, obviously, for the majority of men, PSA is still the gatekeeper to activity, DRE and PSA. If those are elevated or there's a bump on the prostate, then it's very common now to get an MRI, not so much as a screen, so much as helping the biopsy direct. Now, in high-risk populations, people are starting to think about using MRI in a more screening mode. That is, regardless of what the PSA is, get an MRI. That's usually in a younger population who has some kind of genetic risk of cancer. For instance, we're doing a trial in patients with DNA damage repair gene defects. So we're recruiting from all over the country. They're not that common, but they exist. And we're scanning almost like active surveillance, year on year, and it's really remarkable how often we're starting to see positives on the MRI, even when the PSA is not elevated. Some people, including myself, feel that that may be the best use of resources.

Because, men who present with metastatic disease, obviously their tumor started much earlier. And if you were going to save them, you really needed to start treating them in their 40s and 50s, not in their 60s and 70s. It's too late for those men. So if we're really going to do something, we need to start earlier, but you can't do that in every man. So, select people who have a higher risk, genetically, we're wondering about it in the African American community, because the baseline risk is 1.8 fold the general population, in that population. There are particular issues of MRI in that setting because of access to care. But it's at least something to be considered.

Phillip Koo: That's interesting, because it seems similar, some parallels to breast MR in women who are high-risk, so I think there's some maybe lessons we can learn from that space.

Peter Choyke: Yeah, that's a good point.

Phillip Koo: It's interesting. Quality with prostate MRI is still an issue and there's still a lot of concerns with interrater variability, protocols, all these questions, despite a lot of great efforts that have been done through [inaudible] and whatnot. Can you help us understand the issue and then how we make some meaningful improvements in this space? Because, I think we've gotten better, but there's still a long way to go.

Peter Choyke: I agree with you completely. As a referral center, we see a lot of outside scans. The decision is whether we should repeat the scan or not. To me, that is the acid test for whether something is of sufficient quality. Because if you've gone through a test that may have been billed out at $1,000 or $2,000 and nobody can use it, that's a fail. That's about 30 to 40% of the cases that we see coming through the door and that's concerning. So, what's at stake? What's the cause? By now the machines are pretty good, but they are not always being used correctly. An application specialist needs to come in from that particular vendor and tune the protocol up to make it better.

The next issue is that, sometimes people aren't that familiar with interpreting prostate MRIs. They don't do it all the time and you can tell, right? It's just not correct or a lesion is missed. So, one of the things we've been thinking about, there are many ways to approach this problem. One is education. Just get out there, show what it should look like. The American College of Radiology has this onsite. Before COVID it was onsite. Now, I think they're going to start thinking about returning where they showed people 120 cases. Model cases is what it looks like. This is how you interpret it, and they're onsite mentors. So, education, that's one thing. The other thing is the manufacturers are getting into the game. That is to say, they recognize this as a limitation. Even though they're competing against each other, they have an interest in people not saying that the MRIs are terrible. So, they're trying to make their protocols more set, more equivalent to each other.

What I really like is artificial intelligence. With AI, one of the AI's we've developed is one that has looked at a panel of MRIs and judged whether they're inadequate or adequate. So quite objectively you can put your MRI into this device and anonymously see whether your scan passed muster. And if it doesn't, it can give you tips on what is not right with your scan. That'd be sort of a self-teaching tool that I think would be very useful. And the other thing is AI-assisted interpretations, which are getting really pretty good now, like in the 90%. What I say is, it's like having my colleague, Barris Turkbay, who would read my MRI if I had one, it's like having him on your shoulder while you're reading. It's just fantastic because, you could read it first, and then ask the AI, "What do you think?" Sometimes you'll disagree. Sometimes it'll pick up things that you're not seeing. I think that's really good. I know that several of these are out there. There may even be six or seven of these tools out. I haven't looked at them in-depth to say whether they're good or not, but it's certainly the way in the future.

Phillip Koo: Yeah, I'd agree. I think these AI-assisted tools, whether it's prostate MR, or a lot of other areas, really have the potential to improve quality. I think it really changes the scale of what we can do as well and maybe increasing the volume. Maybe somehow we could bring down the price, becomes a tool that's more accessible.

Peter Choyke: The one thing I'm pretty sure shouldn't happen is that, AI-only interpretations. That's a recipe for disaster. Because right now the false positive rate is still pretty high. If you just put that out as a reading, there'd be a lot of unnecessary targeting and things. So, not yet. But assisted, I think we're there. We're ready for that.

Phillip Koo: All right. We're going to reemphasize that point. AI-only tools do not have a place today. Any urologists, medical oncologists, regulatory people who are listening to this video, please, that's the one take-home message. Among many messages.

Thank you very for joining us, Pete, always a pleasure.

Peter Choyke: Thank you.

Phillip Koo: And thank you for your service to the country, on so many levels.

Peter Choyke: Oh, thank you. I appreciate it. Thanks for the opportunity to talk.