Voxel-level Classification of Prostate Cancer Using a Four-Compartment Restriction Spectrum Imaging Model

Abstract

AbstractPurposeDiffusion MRI is integral to detection of prostate cancer (PCa), but conventional ADC cannot capture the complexity of prostate tissues. A four-compartment restriction spectrum imaging (RSI4) model was recently found to optimally characterize pelvic diffusion signals, and the model coefficient for the slowest diffusion compartment, RSI4-C1, yielded greatest tumor conspicuity. In this study, RSI4-C1 was evaluated as a quantitative voxel-level classifier of PCa.MethodsThis was a retrospective analysis of 46 men who underwent an expanded-acquisition pelvic MRI for suspected PCa. Twenty-three men had no detectable cancer on biopsy or clinical follow-up; the other 23 had biopsy-proven PCa corresponding to a lesion on MRI (PI-RADS category 3-5). High-confidence cancer voxels were delineated by expert consensus, using imaging data and biopsy results. The entire prostate was considered benign in patients with no detectable cancer. Diffusion images were used to calculate RSI4-C1 and conventional ADC. Voxel-level discrimination of PCa from benign prostate tissue was assessed via receiver operating characteristic (ROC) curves generated by bootstrapping with patient-level case resampling. Specifically, we compared RSI4-C1 and conventional ADC on mean (and 95% CI) for two metrics: area under the curve (AUC) and false-positive rate for a sensitivity of 90% (FPR90). Classifier images were also compared.ResultsRSI4-C1 outperformed conventional ADC, with greater AUC [0.977 (0.951-0.991) vs. 0.921 (0.873-0.949)] and lower FPR90 [0.033 (0.009-0.083) vs. 0.201 (0.131-0.300)].ConclusionRSI4-C1 yielded a quantitative, voxel-level classifier of PCa that was superior to conventional ADC. RSI classifier images with a low false-positive rate might improve PCa detection.