Dominik Deniffel: Thank you for having us.
Thomas Herrmann: Thank you.
Dominik Deniffel: Thank you, Dr. Klaassen, for your interest in our work, and I'm really happy for the opportunity to discuss and share this with the urology community. So in our recently published work, we established the first empirical learning benchmarks for prostate MRI interpretation. And importantly, we established these for both urology and ... Or, I should have started with radiology and urology trainees. So when you see this topic, your first instinct might be, "Wait, this hasn't just been done before?" And that was my exact reaction when I started working on this project. But as it turns out, we are looking at a major gap in literature here. So what's our motivation for all this? So in radiology, there are some certification frameworks in place in some countries, but all those training standards are based on expert consensus. There is no real data. Nobody can point to hard evidence and say, you need exactly this many cases to reach competency. And then another open question is, when do we start learning prostate MRI? In many institutions, including ours, by the way, it tends to be pushed to fellowship. Now, when we look at urology, I mean, prostate MRI is literally everywhere. It's baked into all the major guidelines.
And the reality is you are already reviewing these images whether or not you're formally signing the report. Yet only 17 to 23% of urology residents report any formal training. Now, urologists take these high-stakes clinical decisions based on images you are not properly trained to interpret. Now that's an obvious problem. So for our study, we recruited 14 prostate MRI-naive trainees. We had 10 from radiology and four from urology. So now on the surface, that does sound like a small group, and that's exactly why the statistics matter so much here. And I promise I explain more in the next slides. So each participant had to read 200 multiparametric MRI cases. We tried to maintain a balanced mix across PI-RADS categories. And for every single case, every participant had to assign a PI-RADS score, localize the index lesion, then also rate the image quality using the PI-QUAL scoring system and grade for extraprostatic extension. And we also recorded the readout time, and all of this happened on an interactive learning platform that provided real-time feedback after each case. So for the reference standard, each case was read by two expert uro-radiologists. And for the analysis, we pulled all 200 cases per trainee into one single regression model. So that gives us 2,800 data points per outcome. So why does this matter? So most prior studies, my mouse here, use a much simpler approach when assessing learning.
I'm not only talking about prostate MR learning, but in general learning, they measure a baseline on say 20 cases, then another 20 cases after some learning intervention or just some time passed. And then in the end, that gives us only two snapshots. And here's the advantage of our approach. And I get generally excited when talking about statistics. So we are statistically modeling and visualizing the entire learning trajectory across those 200 cases, and that gives us so much more insight than just two arbitrary time points. We see all the trends, we see the learning speed, we see, is there a plateau? We can visually compare the curves between groups. So this is what we found. Prior radiology experience, first of all, did not impact the learning trajectories, at least not significantly. And both groups improved rapidly in the first 50 cases. So urologists jumped nearly 30% in accuracy and radiologists about 24%. And then both hit a plateau, urologists at about 69 cases, and radiologists at 75 cases. So this is our magic number, the training benchmark. And at the end of training, urologists reached 81% agreement and radiologists, 77% agreement, and no significant difference here. And here's the beauty of our statistical approach. Again, just this confidence that does not come from a P-value alone. So look at the learning curve side by side. The two groups are pretty much the same across all of those 200 cases.
Now, the secondary outcomes show the same pattern across all metrics. We have similar learning trajectories, no significant difference between the specialties at the end of training, whether it's assessment of image quality using the PI-QUAL, the sharp drop of readout and reporting time or the grading of extraprostatic extension, the two groups performed equally well. So to wrap it up, structured feedback-based training works, and both radiology and urology trainees reached the same level of competency after training. So our key learning benchmark we found was a performance plateau after about 75 cases, that's the minimum number of cases we should mandate for training programs. And interestingly, prior general radiology experience was almost irrelevant. Everyone ended up learning at the same pace. So this gives us an empirical benchmark for those certification frameworks that already exist in radiology. 75 cases should be the standard, and this training should take place early in residency for the radiologists. For urology programs, on the other hand, this supports integration of formal prostate MRI training into residency curriculum, which is not the case yet in most institutions. And I believe, and that's my personal opinion, all this multidisciplinary training will finally strengthen multidisciplinary cancer care. So this is where you can find the paper in full length, it's open access. And I'll hand it over to you, Dr. Klaassen, for further discussion.
Zachary Klaassen: Wonderful presentation and great data. This is really important. We were just talking offline. We did a series probably about two or three years ago in UroToday, beautiful review papers in European Urology, looking at how to do MRIs, how to read them, how you guys look at them. It's very complex stuff. And so I think to be able to break it down, I love the fact that hopefully our listeners can take a take-home number of 75 cases. You guys very nicely showed that that's where that plateau is. So my first question really is, what's next in terms of getting this implemented into training programs? You mentioned there's differences across countries, Europe versus Australia, USA, Canada, et cetera. What's the next step to take this great data and what you've shown and put that into a training program curriculum?
Dominik Deniffel: I mean, first path forward is to start integrating, start integrating that really strictly into existing residency curricula in both specialties. So on an institutional level, we're starting simple. I mean, if you want to implement something like we did in our study, that's actually straightforward. You start by teaming up with your local radiology department. You have them collect 100 cases. They can be stored on your local PACS, separate folder, anonymized. And then I guess the least fun way to go through those cases would be with an Excel sheet, but I think it does the job. I mean, what you need is some sort of feedback and going through those cases. But otherwise, I mean, setting up a web-based learning platform as we did for our study is also straightforward. It's become so accessible with LLMs these days, so that shouldn't be a barrier for most institutions either. And then on an international level, I was talking about the certification frameworks that are already in place, at least in radiology, we provided an actual empirical number for those existing frameworks. And I don't know, in case urology community comes up with the idea to also set up those frameworks in their societies, turns out they can possibly use exactly the same benchmarks.
Zachary Klaassen: No, that's great. I think it's well said. Professor Herrmann, from a urology standpoint, we're taught at early age if we're taking out kidneys, we look at the masses doing a partial nephrectomy, and then we switch over to MRIs and it's a little scary when we first start looking at these. If we had an integrated curriculum where we could say to our graduates, you now are proficient to look at MRIs, we're using this for TP biopsies, we're using this for focal therapy, we're using this for surveillance and active surveillance, et cetera. All these things we're using it for, how would that give confidence to that graduating urology resident? How would they use that in their clinical practice?
Thomas Herrmann: Well, I would say it's not a question whether we should do it, but when we should do it. Because I think after removal of digital rectal examination from the, let's say, panel of examinations with regard to early prostate cancer detection, it's paramount for the urologists to remain in the credibility state when they can interpret properly MRIs. But I don't want to be the party pooper here, but it will take some time to make it an obligatory, integral part of residency training. It will be easier, as Dominik said, to implement it in fellowship training because that's easier to change. I just did the re-accreditation of the Swiss residency program last year, and we have a formal board which takes about effectively two to three years and then a recertification process. So we are within the timeframe in Switzerland, at least of five years to make it an integral part of it. Institution, it's very easy. On the urological body size, for example, European Urology, I am part of European Urology section. So that is the place where you can implement this very easy over the bridge of image-guided therapy or courses. But breaking it down, I think there's no doubt that we need to integrate this to be competent and to be competitive in a way that we can understand each other to talk about this very promising tool because probably in future, we will only have image-based decision-making with regard to prostate cancer treatment without biopsy.
So that is the future, and AI is coming in, and so somebody needs to control the AI. And why I was in this nice project that Dominik signed up here, because I'm very prone into residency training, and this also takes a big share. I'm editor of the World Journal of Urology since two years, and we have encouraged special issue collections for residency training, because I think that is the most important thing to do, to stay competent, competent and to stay reliable and to maintain in the credibility frame to be competent to treat oncologic patients because who knows where the cancer is, is allowed to treat it.
Zachary Klaassen: That's right. Very well said. Dominik, I'll pose the same question too from a radiology training perspective. You mentioned this has typically historically been in fellowship, but moving it into the actual residency training, what's the benefit? And it seems like an obvious question, maybe just expand on that from a radiology graduate who may or may not go into fellowship. Just having that time point as a resident versus a fellow and what that means going into their practice, having that-
Dominik Deniffel: So far, I guess the assumption everywhere was prostate MRI is so complex. You have to deal with all the different sequences. You need to know MR physics, you need to know the artifacts. And well, as it turns out, based on our study, it's not that important, at least if you just focus on the prostate itself. So if it's just about rating those prostatic lesions, it's not about the whole pelvic exams. We did not assess that. So keep that in mind, please. But if it's just about the prostate, obviously all that general radiology knowledge does not help so much. It's such a specific and small organ after all that all that general radiology knowledge did not help much. So there's no reason to wait until later stages of residency, in my opinion. So we have this training program in place right now. So we try to embed it into our training curriculum early in residency, have those residents read those, well, probably a little bit more than 75 cases, at least 100, but above the benchmark. And then it's like a jumpstart to start off their prostate MRI reporting career. But what we do not know yet or do not know much about is long-term skill retention. So there is no data out there right now. So I mean, we would expect if a resident early in his residency just goes through this training program and has no exposure whatsoever on other cases, then obviously that skill level will drop. And some experts in our field actually advocate for maintaining for at least for expert-level uro-radiologists to maintain a high case level per year, like some experts say 200 cases, which sounds like a reasonable number to me.
But again, we're in this expert opinion area again, so that's an interesting avenue for further research for us. So what does it take to keep that skill level, which is obviously important with regards to when do you start your training?
Zachary Klaassen: Yeah, well said. Gentlemen, excellent conversation. Maybe just a quick wrap up from each of you. I'll start with Professor Herrmann.
Thomas Herrmann: Yeah, my wrap up would be, as you said, so it is necessary. We did a nice start. We have to have longitudinal data with regard to the skillset as Dominik just said, but this I think is a good start and research like this will encourage the bodies that provide the residency program to make it an integral part. And until then, it's bottom-up and not top-down. And I think this is a very nice paper and I'm very proud that we could provide it to the urological community to spark more interest into this training.
Zachary Klaassen: Absolutely. Dominik, give the final say.
Dominik Deniffel: Well, so I think the most important takeaway for the urology community is this. So we finally answered the question that has been lingering for years. Can we actually train urologists to properly interpret prostate MRI? And the answer is definitely yes. And the number of cases required is 75. And then another question we could answer is, does it take any general radiology experience to train prostate MRI? And it turns out not so much. So I think everything is aligned now to implement this into curriculum. And this is not about shifting any professional boundaries. So this is about creating a common language and improve interdisciplinary communication and ultimately improve patient care.
Thomas Herrmann: Urologists are trainable.
Dominik Deniffel: Yes.
Thomas Herrmann: What a nice summary. Nice summary.
Zachary Klaassen: Great summary. Well said.
Dominik Deniffel: Absolutely.
Zachary Klaassen: Gentlemen, thank you again for your time. Congratulations on the great work and for joining us on UroToday. Dominik Deniffel: Thank you for having us.
Thomas Herrmann: Yes thank you.
Dominik Deniffel: Bye.
Thomas Herrmann: Bye-bye.