Precise zonal diagnosis: multi-b-value DWI model reveals differential predictors of clinically significant prostate cancer in peripheral and transition zones.

objectivesTo evaluate the diagnostic performance of multi-b-value DWI models for clinically significant prostate cancer (csPCa), identify zone-specific predictors in the peripheral (PZ) and transition zones (TZ), and validate the model's robustness across different MRI vendors. MATERIALS AND

methodsThis retrospective study enrolled 238 patients, comprising a primary cohort (n = 162) and an independent cross-vendor validation cohort (n = 76). Seven diffusion models (mono-exponential model (MEM), intravoxel incoherent motion (IVIM), diffusion kurtosis imaging (DKI), stretched-exponential model (SEM), fractional order calculus (FROC), continuous-time random walk (CTRW), and IVIM-DKI model) were fitted to generate 18 parameters. LASSO regression and generalized estimating equations (GEE) identified independent predictors. Model performance was assessed using ROC curves and decision curve analysis (DCA). Subgroup analyses were performed in PZ/TZ.

resultsMEM_ADC and CTRW_alpha were identified as robust independent predictors of csPCa. In the test set of the primary cohort, the Clinical+Multib_DWI model achieved an AUC of 0.85. Although the improvement over the Clinical+ADC model (AUC = 0.80) was not statistically significant (p > 0.05), the multi-b-value model demonstrated superior clinical net benefit. Crucially, in the cross-vendor validation cohort, the model maintained robust diagnostic accuracy (AUC = 0.88). Subgroup analysis revealed that CTRW_alpha exhibited strong diagnostic value for TZ lesions (AUC = 0.86) and TZ PI-RADS 3 lesions (AUC = 0.82).

conclusionMEM_ADC and CTRW_alpha are zone-specific predictors of csPCa. While the multi-b-value model did not significantly outperform the ADC model in AUC, it offered superior clinical utility through higher net benefit and demonstrated cross-vendor robustness, supporting the translational potential of advanced diffusion models. CRITICAL RELEVANCE STATEMENT: This study identifies zone-specific diffusion predictors for prostate cancer. By demonstrating robustness across different MRI vendors, the findings demonstrate that advanced diffusion models can be successfully translated from specialized protocols to clinical settings, providing superior decision-making utility regarding biopsy necessity. KEY POINTS: Advanced diffusion models lack cross-vendor validation for prostate cancer diagnosis. Selected diffusion parameters demonstrated robust cancer prediction across independent scanner vendors. Zone-specific evaluation offers superior clinical benefit for personalized biopsy decisions.