No one wants to look like a fool at a social gathering by showing up with the wrong outfit. You wouldn’t wear a tux to a backyard BBQ, or Bermuda shorts and a pineapple shirt to a black-tie benefit. As you stand in front of your closet, foreknowledge about the nature of the event will determine which hangers you reach for.
Social embarrassment is hardly pleasant, but it’s not as crucial as saving your life by choosing the right prostate cancer (PCa) treatment. To do so, you and your doctor need as much accurate foreknowledge as possible about the location, size and extent of a prostate cancer tumor in your gland. The importance of human expertise has long been recognized by all PCa patient support/education programs. Such organizations have consistently encouraged patients to have an expert radiologist perform and interpret their prostate MRI scan; and if that’s not possible, send their MRI CD to an expert for a second opinion. Accurate image-based diagnosis before a needle biopsy can save a man’s life, especially if the disease appears aggressive and/or extensive, requiring immediate action. It can also save his quality of life by avoiding overtreatment if the scan reveals an index lesion suitable for a focal treatment or for Active Surveillance.
Can a computer perform as well as a human for diagnosing the presence and characteristics of PCa based on imaging? Increasingly, it appears that computer-aided diagnosis (CAD) can, thanks to various methods of Artificial Intelligence that are being designed and tested.
For example, a 2022 study reported on a new Deep Learning-based program that combines a PCa localization network with an integrated multi-modal classification network.[i] For classifying prostate cancer and non-cancerous tissues based on multiparametric MRI (mpMRI) it demonstrated excellent sensitivity of 95%.
Perhaps an even more important 2023 study of CAD compared the diagnosis of an AI program called the Watson Elementary® (WE®) CAD system with actual radical prostatectomy (RP) specimens obtained surgically.[ii] Initially, all study patients who were scheduled for RP between 2020-21 underwent mpMRI, and their PI-RADS scores were recorded. Their image results were then analyzed by the WE CAD program. Compared to the post-surgical specimens, the research team calculate the program’s ability to “…predict the presence of PCa, to correctly locate the dominant lesion, to delimit the largest diameter of the dominant lesion, and to predict the extraprostatic extension (EPE).”
The WE program’s identification of highly suspicious areas had 92% accuracy for predicting PCa, and its accuracy was confirmed in 86% of PI-RADS ≥ 4 lesions. Even more encouragingly, “In 98% of cases, visible tumor at WE® showed that the highly suspicious areas were within the same prostate sector of the dominant tumor nodule at pathology.” In fact, the WE program was more precise than mpMRI alone at identifying the diameter of tumors. Thus, the authors concluded that, “The WE® system resulted accurate in the PCa dominant lesion detection, localization and delimitation providing additional information concerning EPE prediction.”
As we’ve pointed out in an earlier blog, no one is suggesting that AI is about to replace expert radiologists. In its current form, it helps expedite both efficiency and accuracy of the work of human readers who interpret the MRI scans. In the diagnostic pathway, this is beneficial to patients and their doctors, since the imaging results are returned faster. Speed makes a difference, especially if the patient’s disease level is high-risk, so time is of the essence. It is also advantageous for those with low risk disease, who now have correct information that they can take time to explore less aggressive treatment options, choosing one that best fits their lifestyle and psychological comfort level.
Stay tuned for more good news about AI and how it can help both physicians and those they treat.
NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain, or have any other health concerns or questions of a personal medical nature.
[i] Yi Z, Ou Z, Hu J, Qiu D et al. Computer-aided diagnosis of prostate cancer based on deep neural networks from multi-parametric magnetic resonance imaging. Front Physiol. 2022 Aug;Vol. 13. https://doi.org/10.3389/fphys.2022.918381
[ii] Vittori G, Bacchiani M, Grosso AA, Raspollini MR et al. Computer-aided diagnosis in prostate cancer: a retrospective evaluation of the Watson Elementary® system for preoperative tumor characterization in patients treated with robot-assisted radical prostatectomy. World J Urol. 2023 Jan 3. doi: 10.1007/s00345-022-04275-x.