Objectives: To differentiate prostate cancer lesions with high and with low Gleason score by diffusion-weighted-MRI (DW-MRI).
Methods: This prospective study was approved by the responsible ethics committee. DW-MRI of 84 consenting prostate and/or bladder cancer patients scheduled for radical prostatectomy were acquired and used to compute apparent diffusion coefficient (ADC), intravoxel incoherent motion (IVIM: the pure diffusion coefficient D t, the pseudo-diffusion fraction F p and the pseudo-diffusion coefficient D p), and high b value (as acquired and Hessian filtered) parameters within the index lesion. These parameters (separately and combined in a logistic regression model) were used to differentiate lesions depending on whether whole-prostate histopathological analysis after prostatectomy determined a high (≥7) or low (6) Gleason score.
Results: Mean ADC and D t differed significantly (p of independent two-sample t test < 0.01) between high- and low-grade lesions. The highest classification accuracy was achieved by the mean ADC (AUC 0.74) and D t (AUC 0.70). A logistic regression model based on mean ADC, mean F p and mean high b value image led to an AUC of 0.74 following leave-one-out cross-validation.
Conclusions: Classification by IVIM parameters was not superior to classification by ADC. DW-MRI parameters correlated with Gleason score but did not provide sufficient information to classify individual patients.
Key points: • Mean ADC and diffusion coefficient differ between high- and low-grade prostatic lesions. • Accuracy of trivariate logistic regression is not superior to using ADC alone. • DW-MRI is not a valid substitute for biopsies in clinical routine yet.
Keywords: ADC; Diffusion-weighted MRI; IVIM; Logistic regression; Prostate cancer.