Zone-specific diagnostic performance of synthetic MRI quantitative parameters, ADC, and PSA density for clinically significant prostate cancer: comparison with PI-RADS v2.1.

BACKGROUND

The diagnostic performance of synthetic magnetic resonance imaging-derived quantitative tissue parameters relative to Prostate Imaging Reporting and Data System version 2.1 (PI-RADS v2.1) remains incompletely understood. This study evaluated the zone-specific diagnostic performance of magnetic resonance image compilation (MAGiC)-derived T1, T2, and proton density (PD), apparent diffusion coefficient (ADC), and prostate-specific antigen density (PSAD) for clinically significant prostate cancer (csPCa).

METHODS

This retrospective diagnostic accuracy study included 141 treatment-naïve patients with 141 MRI-visible index lesions and histopathologic confirmation. Sixty-three lesions were located in the peripheral zone (PZ) and 78 in the transition zone (TZ). Two radiologists independently assigned PI-RADS v2.1 categories. T1, T2, PD, and ADC were measured using a three-slice region-of-interest approach. Zone-specific logistic regression models combining T1, T2, ADC, and PSAD were evaluated by receiver operating characteristic analysis. Ridge-penalized regression was used to generate predicted probabilities for the TZ models.

RESULTS

The cohort included 52 csPCa and 89 clinically insignificant lesions. T1, T2, and ADC were lower in csPCa in both zones (all P < 0.001), whereas PD did not differ. ADC showed the highest individual AUC in the PZ (0.881) and TZ (0.878). The T1 + T2 + ADC model showed higher apparent AUCs than PI-RADS v2.1 in the PZ (0.947 vs 0.799; P = 0.002) and TZ (0.961 vs 0.803; P = 0.013). Adding PSAD increased the AUCs to 0.961 and 0.963, but the increments were not significant (P = 0.329 and P = 0.700). PI-RADS agreement was substantial, and quantitative measurements showed good-to-excellent reproducibility.

CONCLUSIONS

MAGiC-derived T1 and T2 measurements combined with ADC showed promising zone-specific discrimination of csPCa and higher apparent performance than PI-RADS v2.1. External validation and assessment of calibration and clinical utility are required before implementation.