Zone-Specific PSA Density from AI-Assisted mp-MRI as a Predictor of Prostate Cancer in the PSA 4-20 ng/mL Range

Background/Objectives: Serum prostate-specific antigen (PSA) lacks sufficient specificity for prostate cancer (PCa) detection. We evaluated the diagnostic value of AI-assisted MRI-derived zone-specific PSA density parameters, including transition zone PSA density (TZ-PSAD) and peripheral zone PSA density (PZ-PSAD), in men undergoing prostate biopsy. Methods: This retrospective cohort study included 354 patients who underwent prostate biopsy due to elevated PSA (2024-2026), stratified into PSA 4-10 ng/mL (n = 272) and PSA 10-20 ng/mL (n = 82) groups. Zonal volumes were measured using AI-assisted mp-MRI segmentation per PI-RADS v2.1, and PSAD, TZ-PSAD, and PZ-PSAD were calculated accordingly. MRI readings were performed blinded to pathology. Results: TZ-PSAD was significantly higher in PCa patients in both PSA groups (p < 0.001), demonstrating superior AUC compared to PSAD in both the PSA 4-10 ng/mL (0.711 vs. 0.660) and PSA 10-20 ng/mL (0.841 vs. 0.676) groups. At an optimal cut-off of ≥0.34 ng/mL², TZ-PSAD yielded 44.8% sensitivity and 90.8% specificity in the PSA 4-10 group, improving to 73.8% sensitivity and 94.7% specificity at ≥0.48 ng/mL² in the PSA 10-20 group. On multivariate analysis, TZ-PSAD and PI-RADS ≥ 3 emerged as independent predictors of PCa in both groups. Serum PSA, PZ-PSAD, and PZ volume failed to differentiate BPH from PCa in either group. Conclusions: TZ-PSAD and conventional PSAD were both significantly associated with prostate cancer, whereas PZ-PSAD was not, and TZ-PSAD, combined with PI-RAD ≥3, emerged as an independent predictor in the PSA gray zone.