Jelle Barentsz: Okay. Well, the question is VI-RADS is not enough, why? Why is image quality the key to reliable bladder cancer staging? VI-RADS is just like BI-RADS, a standardization method for multiparametric MRI for bladder cancer reporting. Why is that important? It has been shown that this improves staging and consistent communication in the multidisciplinary team, but the performance depends on consistent high-quality image acquisition. And this is even a larger challenge than with prostate cancer. The bladder is a rounded structure with only a very thin wall and to make a consistent good image quality in all the areas is quite a challenge. We have seen that the current multiparametric MRI quality of bladder cancer is varying widely. And just like with prostate cancer, this is a threatening of the distinction between the things we want to see. In prostate, it is between an insignificant and a significant cancer. In VI-RADS, you want to discriminate the non-muscle-invasive tumor from the muscle-invasive tumor.
That was why VI-RADS was built, but this requires a good image. It sounds like kicking an open door. Just like with PI-RADS, the PI-QUAL classification, the VI-RADS quality score called VI-QUAL, has three categories. Number one, inadequate, two, adequate, and three, the best, that's optimal. And it has a per-sequence scoring of T2, diffusion and dynamic contrast. This enables targeted protocol changes and technologist training. You just can evaluate where is the weakest part of this image. So you can say to the technician, you have to pay more attention to the T2 or diffusion. You have to increase the signal-to-noise and decrease the artifacts usually from air in the rectum. The same is for contrast. If you do this, this will increase the image quality, the IQ, and the confidence in urologists and oncologists. What are the challenges? The experience gaps. Well, we have seen that the inter-reader reliability, for that we need subjective... Now is based on subjective criteria and we need training. We need to have a tool to increase it with objective criteria. For example, having a lexicon and preferably, like with prostate cancer, a reader certification, we have the ESUR prostate MRI certificate that focuses not only on the recognition of the things you want to see, non-muscle versus muscle-invasive bladder cancer, but it also focuses on the image quality.
And usually we don't look at image quality with attention. We've seen that less experienced read this overestimate quality. They think, "Oh, this is good." But it's actually the experts that are critical. We need in the future to validate the reproducibility. Thus far, we do not know what VI-QUAL is in one center and what the quality score is in another center. In prostate cancer, we are a little bit ahead and we've seen that there are now reports that show that the inter-reader variability of PI-QUAL is equal to PI-RADS. So in the near future, we will learn what is the reproducibility and we need to have correlation with diagnostic performance. We need to have prospective data showing that higher image quality scores improve the VI-RADS's accuracy, sensitivity, specificity, and predictive values. We've seen that link in prostate cancer, the link between good image quality, high PI-QUAL, and good scores of the PI-RADS. What do we have more? We have technical and implementation issues. We have the dependence on scanner technology and vendor differences and also patient factors.
What is the motion of the bladder? What is the motion of bowels, air in the rectum? What about filling of the bladder? Do you want to have a full bladder, a moderately distended bladder, or an empty bladder? I prefer a moderately extended bladder, but if the patient is drinking a lot of coffee before the MRI, at the end of your exam, it's a very, very full bladder. The risk of a two-tier system is disadvantaging for non-expert centers, so we need to have a three-tier system. Also, the added workload for radiologists must be justified by clear clinical benefit. At this moment, we are facing a lot of work at the clinical practice and something in addition, we don't like it, but this is important. You need to look at your image quality. It's just only costing a few minutes of your time. That's my experience, but I need to convince you guys in clinical practice, if you have 25 MRIs and if it takes you two minutes per MRI, you need to get that 50 minutes in total and you need to, well, get that from your time. And it's my task and the task of the expert group to convince you that the 50 minutes is crucial.
You need to have image quality control. The clinical impact, the outlook, what can we achieve with the quality score? And this is important. It will standardize image quality in routine care and also in multicenter trials so we can compare apples with apples. It will reduce the diagnostic uncertainty. It will decrease the unnecessary repeat removals of the bladder cancer and inappropriate use of the neoadjuvant therapy. So it will increase also the clinical factor. Good quality gives good imaging, good imaging gives good information, good information gives better clinical care. Also, this is very important for the AI development. Only high-quality curated data sets will be the feed of reliable models. If we have a poor dataset, it's garbage in, garbage out, and we do not want to do that. And we are now starting with AI. So we need also to look at image quality before we feed in the images into the AI training sets. Also, just like the prostate, there may be a possibility that AI helps us to score VI-QUAL, so AI-assisted image quality control. What are the future priorities? Well, we need to have prospective validation in diverse real-world settings. So we need to start to work on trials, short trials that are evaluating VI-QUAL and VI-RADS in relation to the clinical outcome.
We also need to have broad structured education and implementation programs. If you now look at, for example, prostate cancer in my country, 30% of the centers that were tested, and that was just along the country, 30% of the MRIs of the multiparametric MRIs were inadequate. PI-QUAL 1, 50% of the biparametric MRIs were inadequate. And we have to be very careful that we will not have this with the bladder, that VI-QUAL training should be part of the education and implementation programs today. If successful, VI-QUAL-like scoring could enable consistent, reliable, and equitable bladder multiparametric staging for all patients. A last word, I had a discussion in the editorial board of European Urology, and we are considering in asking the authors that submit prospective studies to ask whether they have been doing image quality control in prostate cancer. And I think in the near future, this is also possible with bladder. So it will be asked, did you perform any form of image quality control? And for that, VI-QUAL is an important tool. So the future of bladder cancer is bright with VI-QUAL followed by VI-RADS. Thank you for your attention.
Ashish Kamat: Great. Thank you so much, Professor Barentsz. And yes, I remember the discussion we had during the editorial board at EMUC. I think it was very enlightening. I think it's very, very important. If you could outline for our audience, because of course most of us listening in are going to be in the urology field, but there's going to be a lot of radiologists and young aspiring radiology residents and fellows. Share with us briefly, of course, some of your tips in how people can go about getting this education for the VI-QUAL, [inaudible 00:11:52] to be approaching education and training for VI-QUAL.
Jelle Barentsz: Yes. I think it needs to be integrated in all the courses that we have at the AUA, at the RSNA, at the SAR, at the ECR, at the EAU. We need to have specific courses that focus on image quality. It can be just a small part, but our clinicians and the radiologists, we need to know that there is a tool that is able to measure quality and that tool needs to be presented in a comprehensive way. So you can have a short refresher course of 20 minutes dealing about PI-QUAL and VI-QUAL, that is simple. In addition, there are fortunately very nice workshops on PI-RADS and on VI-RADS. I think in that workshop, and it has already been implemented in a little bit, in those workshops, it should start with VI-QUAL and with BI-QUAL. For example, in the prostate MRI examination for the certification, I participated, I did not participate in making it, but I was a participant. So I did the test to see whether I would pass the prostate certificate and fortunately, I did. One part was already on image quality. So the question was, what quality has this image? And you were asked to score. So the other advice I have to my colleagues in the clinic is that the radiologists, we should score every exam. So you have to put that in your template of your reporting.
It starts with clinical information, then it starts with image quality. Image quality: T2 diffusion DCE 4-4-2. Results: VI-QUAL 3, good quality. And then in brackets, why the quality is not good enough so that everybody can see what your interpretation is of your own image. That is important. That is how VI-RADS and PI-RADS also started. Because we spoke the same language, we could see the differences between the groups and the radiologists. So with image quality, that's the same. So just start. Look at the lexicons, look at the images, discuss this with your colleagues and with experts, and then just start. So if you do that, you can train yourself during your clinical work. So I think that is the best advice. And if you have problems with image quality, go to an expert center and ask, "How are you dealing with that? Is the bladder too much distended? What about these SNR, the spatial resolution and the signal in the bladder wall for diffusion? Is your DCE correct? How can you suppress motion artifacts? How you can deal with air in the rectum, susceptibility artifacts? Et cetera, et cetera." I hope that gives a little bit of a clue how you should start.
You should start by doing it and also after that, you have the MDT and in the MDT, there is also some time to educate the urologist. So what we do in the MDT typically is clinical information and radiologist presents the data and then the pathologist, but the radiologist saying, "Well, this image is a good quality because of da, da, da, da," or "Is not good quality VI-QUAL 2," so that everybody is a little bit understanding what is a 1, what's the 2, what's a 3? And sometimes a 1 gives good results from a 1 inadequate image, sometimes you see with a very gross tumor that it is a gross tumor, but there's also some, well, you can miss small tumors and so from the practice you learn.
Ashish Kamat: Absolutely. No, very, very important critical points. And I think just as with pathology or just as with anything else, recognizing the importance of the quality of the information coming in is very important because we make so many decisions based on that. And I'm sure there are efforts ongoing where we can use machine learning to actually give feedback on the quality of the image, hopefully in real time so the patient doesn't have to come again to the scanner and get scanned.
Jelle Barentsz: What my dream is, and I think it can be realized, is that you have a scanner, the patient is in a scanner, and then a certain image sequence is being performed, let's say a T2, and then automatically AI is being performed and then with a red light, blip, blip. Okay, this is not a good sequence. It needs to be repeated, quality is bad, and with some clues, that would be very nice. It also gives more reliability in the performance of a technician. Now, MRI is very highly technician-dependent. If you have a good technician, the quality is good. If you have a bad technician, the quality is not that good. And I can see based on the image, there's a signature of the guy who's making the MRI and that we need to try to avoid with artificial intelligence.
Ashish Kamat: Absolutely. Professor Barentsz, thank you so much for taking the time. Very important topic. Thank you so much for joining us.
Jelle Barentsz: Okay. Thank you. You're welcome.