Photo Credit: iStock.com/Mohammed Haneefa Nizamudeen

Diffusion basis spectrum imaging combined with AI enhances diagnostic accuracy for prostate cancer

A study published in the June 2025 issue of Journal of Urology evaluated artificial intelligence models applied to diffusion basis spectrum imaging metrics for predicting clinically significant prostate cancer before biopsy.  

They assessed 241 patients who underwent prostate magnetic resonance imaging with conventional and diffusion basis spectrum imaging sequences before biopsy between February 2020 and March 2024. Artificial intelligence models used DBSI metrics as input classifiers, with biopsy pathology serving as the reference standard. The DBSI-based model was compared to prostate-specific antigen, PSA density [PSAD], and Prostate Imaging Reporting and Data System [PI-RADS] for discriminating clinically significant prostate cancer defined by Gleason score greater than 7.  

The results showed that the diffusion basis spectrum imaging–based model independently predicted clinically significant prostate cancer with an odds ratio of 2.04 (95% CI, 1.52–2.73; P < .01), alongside PSA density (OR 2.02; 95% CI, 1.21–3.35; P = .01) and Prostate Imaging Reporting and Data System classification (OR 4.00 for PI-RADS 3; 95% CI, 1.37–11.6; P = .01; OR 9.67 for PI-RADS 4–5; 95% CI, 2.89–32.7; P < .01), after adjustment for age, family history, and race. The DBSI-based model alone showed similar performance to PSA density plus PI-RADS (AUC 0.863 vs 0.859; P = .89). The combination of the DBSI-based model and PI-RADS achieved the highest discrimination (AUC 0.894; P < .01). Using the DBSI-based model in patients with PI-RADS 1–3 could have reduced biopsies by 27%, missing 2% of clinically significant prostate cancer compared to biopsying all.  

Investigators concluded that the diffusion basis spectrum imaging–based artificial intelligence model accurately predicted clinically significant prostate cancer and, when combined with PI-RADS, could potentially reduce unnecessary prostate biopsies. 

Source: auajournals.org/doi/10.1097/JU.0000000000004456