Using Artificial Intelligence to Map Cancer in the Prostate

By: Chris Schimpf, BS
Posted: Monday, August 7, 2023

Artificial intelligence (AI) may be useful in determining the extent of cancer within the prostate, potentially improving and standardizing the definition of treatment margins for focal therapy and reducing rates of cancer recurrence, according to a study published in European Urology Open Science. Geoffrey A. Sonn, MD, of Stanford University School of Medicine, and colleagues developed an AI deep-learning model that combines multimodal imaging and biopsy data to produce three-dimensional cancer estimation maps and assess positive margin risk in patients with prostate cancer. In their retrospective study, the researchers found the model to be more effective than conventional methods and noted considerable potential in advancing accurate, patient-specific diagnosis and treatment planning.

“Precision management of prostate cancer has the potential to optimize therapy while preserving quality of life, but targeted treatment first requires accurate tumor localization,” the investigators stated. “In our study, treatment of the original regions of interest would have resulted in positive margins for every patient. It is clear that current multiparametric MRI contouring protocols, which were developed for diagnosis, are not suitable for targeted treatment.”

Biopsy data from multiple institutions were used to train the AI model, and then testing was conducted using an independent data set of 50 consecutive patients who underwent radical prostatectomy for intermediate-risk cancer. Comparing the AI model with the data set’s conventional methods, the team observed that the mean sensitivity for cancer-bearing voxels was higher for AI margins (97%) than for conventional regions of interest (37%; P < .001), 10-mm region-of-interest margins (93%; P = .24), and hemigland margins (94%; P < .001). For index lesions, AI margins were more often negative (90%) than conventional regions of interest (0%; P < .001), 10-mm region-of-interest margins (82%; P = .24), and hemigland margins (66%; P = .004).

Disclosure: For full disclosures of the study authors, visit sciencedirect.com.

European Urology Open Science

---