Artificial Intelligence in Medicine: Designing a Virtual Prostate Biopsy Using MRI Scans
Your PSA blood test came back abnormally high. Your doctor says you might have prostate cancer (PCa) and wants to do a needle biopsy, but guess what? You have a choice. Either:
a) He numbs you up and pokes your prostate gland with needles in order to extract tissue samples, or
b) He sends you for a “virtual biopsy” in which an MRI of your prostate will use two imaging sequences, which will then be analyzed by Artificial Intelligence (AI) software. If it shows you have PCa, it will also reveal if it’s low aggression or high aggression.
Well, the day has not yet come for that scenario, but if it had, what would you choose? The idea of a virtual biopsy is appealing. However, there’s a slight catch. If the AI software says you have aggressive PCa, you’ll still need a needle biopsy. Still, the virtual biopsy is a reasonable next step, because if it’s low aggression, you can skip the biopsy for now, and go on Active Surveillance. Or, if it identifies high aggression, you can go for an in-bore MRI guided targeted biopsy for a precise diagnosis.
New algorithm benefits patients
Increasingly, AI is being embraced in all areas of radiology, including MRI of the prostate. In fact, the Sperling Prostate Center is in the vanguard of integrating AI into PCa diagnosis. One type of AI is Machine Learning (ML). TechTarget tells us that ML involves “training” an algorithm, or prediction calculation, by inputting historical data to predict new output values.
Now, a group of Italian physicians and engineers have developed a Machine Learning program that can distinguish low vs. high aggressive PCa in prostate MRI scans. Nicoletti, et al. (2021)[i]trained, tested and validated a diagnostic algorithm based on two MRI parameters: Apparent Diffusion Coefficient (ADC) map and T2-weighted imaging.
This algorithm was trained to recognize prostate imaging features that predict the degree of aggression, which was classified as either low aggressive (Grade Group ≤ 2) or high aggressive (Grade Group ≥ 3). After refining and testing the program, the ML algorithm succeeded in distinguishing between low and high aggressive lesions with high sensitivity (77%) and very high specificity (93%).
This would benefit PCa patients with low aggressive indications who would like to avoid a biopsy, if possible. Although their “virtual biopsy” does not provide a definitive diagnosis (no needles are used to capture tissue for pathology analysis), it is safe and noninvasive. As the authors explain, it “could help clinician[s] to noninvasively distinguish between PCas that might need active treatment and those that
could potentially benefit from active surveillance, avoiding biopsy-related complications.” This is truly putting the power of AI and ML at the service of patients.
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] Nicoletti G, Barra D, Defeudis A, Mazzetti S, Gatti M, Faletti R, Russo F, Regge D, Giannini V. Virtual biopsy in prostate cancer: can machine learning distinguish low and high aggressive tumors on MRI? Annu Int Conf IEEE Eng Med Biol Soc. 2021 Nov;2021:3374-3377.