(UroToday.com) The 2023 EAU annual meeting included a game changing session, featuring a discussant presentation by Dr. Jochen Walz putting into context preliminary results from the PI-CAI challenge assessing artificial intelligence and radiologists at prostate cancer detection on MRI. Dr. Walz started his presentation by noting that when he asked ChatGPT to “discuss PI-CAI Challenge” the response was “the PI-CAI Challenge, also known as the “Privacy-Preserving AI Challenge” is an annual competition that brings together researchers, academics, and industry professionals to develop innovative solutions for privacy-preserving artificial intelligence.” So, he notes, perhaps ChatGPT is not “perfect”.
Based on the EAU guidelines that men with suspicion of prostate cancer should have an MRI prior to prostate biopsy, Dr. Walz emphasized that the tsunami of prostate MRI is coming and thus we certainly have a need for artificial intelligence in prostate cancer diagnosis. Furthermore, we need to reduce scanning time, with biparametric MRI (eliminating the DCE phase) being proposed as a way to do this:
Dr. Walz noted that one of the main limitations of the PI-CAI challenge was that the standard of truth was follow-up, targeted biopsy, and systematic biopsy. Systematic biopsy is associated with a high false negative rate of 25-35% if compared to template biopsy, and targeted biopsy performance is also not 100% reliable. The ideal standard of truth would be whole mount sections from radical prostatectomies (and even using prostate specimens from cystoprostatectomies as negative controls). He notes that if artificial intelligence is to out-perform “humans”, the training set needs to provide a standard of truth that is “human independent” and unbiased by a human’s performance.
Additionally, Dr. Walz highlighted that the sensitivity of multiparametric MRI is already quite good (87-93%), thus artificial intelligence research should be focused on improving specificity (41-47%) and negative predictive value (72-89%). If negative predictive value were to improve, ideally, we would be able to limit the number of prostate biopsies we would have to perform. Dr. Walz notes that another of the aims that the PI-CAI challenge should endorse is improving imaging quality. In 2020, Giganti et al. [1] proposed the prostate imaging quality (PI-QUAL) scoring system, a new quality control scoring system for multiparametric MRI of the prostate from the PRECISION trial. This system scores MRI quality from 1-5 and provides clinical implications:
The clinical meaning of PI-QUAL is as follows:
Higher PI-QUAL scores for image quality are associated with decreased uncertainty in MRI decision-making and improved efficiency of diagnostic pathway delivery.
Dr. Walz also emphasized that with the PI-CAI challenge, we have the opportunity to level up reader performance, particularly given that there can be tremendous inter-reader variability among radiologists reading prostate MRIs.
Dr. Walz concluded his presentation discussing preliminary results from the PI-CAI challenge by asking ChatGPT “AI in prostate cancer diagnosis”, which resulted in a more appropriate answer as follows:
- Image analysis: artificial intelligence can analyze medical images such as MRIs, CT scans, and ultrasound images to identify suspicious areas in the prostate that may be cancerous. This can help radiologists and other medical professionals to make more accurate diagnoses
- Overall, artificial intelligence has the potential to significantly improve the accuracy and efficiency of prostate cancer diagnosis and treatment, leading to better outcomes for patients
Presented by: Jochen Walz, MD, Institut Paoli Calmettes Cancer Center, Marseille, France
Written by: Zachary Klaassen, MD, MSc – Urologic Oncologist, Assistant Professor of Urology, Georgia Cancer Center, Augusta University/Medical College of Georgia, @zklaassen_md on Twitter during 2023 European Association of Urology (EAU) Annual Meeting, Milan, IT, Fri, Mar 10 – Mon, Mar 13, 2023.
References:
- Giganti F, Allen C, Emberton M, et al. Prostate imaging quality (PI-QUAL): A new quality control scoring system for multiparametric MRI of the prostate from the PRECISION trial. Eur Urol Oncol. 2020 Oct;3(5):615-619.