Evaluation of Dispersion MRI for Improved Prostate Cancer Diagnosis in a Multicenter Study.

Abstract

The purpose of this study is to compare dispersion MRI and Tofts model (TM) for analysis of quantitative dynamic contrast-enhanced (DCE) MRI (DCE-MRI) for localization of prostate cancer and to assess the correlation between quantitative DCE-MRI parameters and tumor grade. This retrospective multicenter study included 80 patients with biopsy-proven prostate cancer who underwent DCE-MRI followed by radical prostatectomy. DCE-MRI parameters were extracted from dispersion MRI analysis (the dispersion parameter [k_d], the flux rate [k_ep], and the intravascular mean transit time) and TM analysis (the forward volume transfer constant [K^trans], k_ep, and the extravascular extracellular volume fraction [v_e]). ROIs representing benign and malignant tissue were drawn on each DCE-MRI slice according to the histopathologic findings, and the diagnostic performance of the estimated parameters for the diagnosis of prostate cancer was evaluated using fivefold cross-validation and ROC curve analysis. Further analysis was conducted for the two most relevant parameters (i.e., k_d [for dispersion MRI] and k_ep [for TM]), to investigate the correlation between DCE-MRI parameters and tumor grade. DCE-MRI parameters were significantly different between benign and malignant prostate tissue (p < 0.0001). The dispersion MRI parameter k_d outperformed all other DCE-MRI parameters for prostate cancer diagnosis, showing the highest area under the ROC curve value (p < 0.0001). Only a weak linear correlation (Pearson r = 0.18; p < 0.05) was found between the dispersion parameter and the Gleason grade group. Dispersion MRI outperformed TM analysis, improving the diagnostic performance of quantitative DCE-MRI for prostate cancer localization. Of the DCE-MRI parameters, k_d (for dispersion MRI) and k_ep (for TM) provided only poor characterization of tumor grade.