Validation of Effectiveness of Multi-Parametric Endorectal MR Image Features for Prostate Cancer Detection and Correlation with Gleason Score

Yahui Peng¹, Yulei Jiang¹, Tatjana Antic², Maryellen L. Giger¹, Scott Eggener³, Aytekin Oto¹

We previously identified from Phillips MR images three quantitative multi-parametric endorectal MR image features effective for differentiation between prostate cancer and normal peripheral zone tissue regions of interest (ROIs), and moderately correlate with tumor ROI-specific Gleason score. We now validate these image features on GE MR images of 71 prostate cancer patients. Results show for both Philips and GE scanners that the 10th percentile, and average, ADC values, and T2-weighted signal-intensity histogram skewness are effective in differentiating prostate cancer from normal peripheral zone tissue ROIs, and that the ADC features correlate moderately and negatively with tumor ROI-specific Gleason scores.