Detecting clinically significant prostate cancer with a distributed parameter model based on quantitative dynamic...

Prostate cancer (PCa) is a common malignancy with heterogeneous biological behavior, and accurate differentiation between clinically significant (csPCa) and clinically insignificant PCa (ciPCa) is critical for guiding treatment decisions. Multiparametric magnetic resonance imaging (mp-MRI) has become an essential tool for PCa detection; however, the role of dynamic contrast-enhanced MRI (DCE-MRI) remains limited in the current Prostate Imaging Reporting and Data System (PI-RADS) guidelines due to its largely qualitative nature. Quantitative analysis of DCE-MRI through use of advanced pharmacokinetic models may provide additional diagnostic value. This study aimed to evaluate the feasibility of a distributed parameter (DP) model based on quantitative DCE-MRI to differentiate csPCa from ciPCa.

Patients with suspected PCa were prospectively enrolled and underwent 3.0-T DCE-MRI between June 2022 and May 2025. Voxel-wise kinetic parameters were estimated with the DP model, the extended Tofts model (ETM), and the adiabatic tissue homogeneity (ATH) model. Regions of interest (ROIs) were manually delineated by radiologists on parameter maps, including the permeability surface area product (PS) from the DP and ATH models and the transfer constant (Ktrans) maps from the ETM. Group comparisons were performed to evaluate the ability of individual DCE-derived parameters to differentiate ciPCa from csPCa, with the Gleason score serving as the reference standard. The diagnostic performance of these parameters was further assessed via receiver operating characteristic (ROC) analysis. In addition, the diagnostic performance of biparametric MRI and mp-MRI was compared to evaluate the incremental value of incorporating quantitative DCE parameters into the PI-RADS.

A total of 70 patients comprising 88 biopsy-proven lesions (42 ciPCas and 46 csPCas) were included. Several DP model-derived parameters (PS, mean transit time, plasma volume, and blood flow) were significantly higher in patients with csPCa than in those with ciPCa (all P<0.05), with PS showing the strongest discriminative ability (P<0.001). ROC analysis revealed that DP-derived PS achieved the highest performance [area under the curve (AUC) =0.87, sensitivity =0.74, specificity =0.83, and accuracy =0.78], outperforming parameters derived from the ETM (AUC ≤0.71) and ATH model (AUC ≤0.82). Subgroup analyses stratified by lesion zone and size consistently confirmed the superiority of the DP model. Incorporation of DCE parameters into mp-MRI improved diagnostic performance as compared with biparametric MRI (AUC: 0.92 vs. 0.88).

Quantitative DCE-MRI based on a DP model outperformed the ETM and ATH model in distinguishing csPCa from ciPCa. Among all evaluated parameters, PS derived from the DP model demonstrated the strongest diagnostic ability, supporting its potential as a robust quantitative imaging biomarker for PCa risk stratification.