Abstract
PURPOSE: To retrospectively assess the utility of whole-lesion apparent diffusion coefficient (ADC) metrics in characterizing the Gleason 4 component of Gleason 7 prostate cancer (PCa) at radical prostatectomy. MATERIALS AND METHODS: Seventy patients underwent phased-array coil 3T-magnetic resonance imaging (MRI) before prostatectomy. A uropathologist mapped locations and Gleason 4 percentage (G4%) of Gleason 7 tumors. Two radiologists independently reviewed ADC maps, aware of tumor locations but not G4%, and placed a volume-of-interest (VOI) on all slices including each lesion on the ADC map to obtain whole-lesion mean ADC and ADC entropy. Entropy reflects textural variation and increases with greater macroscopic heterogeneity. Performance for characterizing Gleason 7 tumors was assessed with mixed-model analysis of variance (ANOVA) and logistic regression. RESULTS: Among 84 Gleason 7 tumors (G4% 5%-85%, median 30%; 59 Gleason 3+4, 25 Gleason 4+3), ADC entropy was significantly higher in Gleason 4+3 than Gleason 3+4 tumors (R1: 5.27 ± 0.61 vs. 4.62 ± 0.78, P = 0.001; R2: 5.91 ± 0.32 vs. 5.57 ± 0.56, P = 0.004); mean ADC was not significantly different between these groups (R1: 0.90 ± 0.15*10(-3) cm(2) /s vs. 0.98 ± 0.21*10(-3) cm(2) /s, P = 0.075; R2: 1.06 ± 0.19*10(-3) cm(2) /s vs. 1.14 ± 0.16*10(-3) cm(2) /s, P = 0.083). The area under the receiver operating characteristic (ROC) curve (AUC) for differentiating groups was significantly higher with ADC entropy than mean ADC for one observer (R1: 0.74 vs. 0.57, P = 0.027; R2: 0.69 vs. 0.61, P = 0.329). For R1, correlation with G4% was moderate for ADC entropy (r = 0.45) and weak for mean ADC (r = -0.25). For R2, correlation with G4% was moderate for ADC entropy (r = 0.41) and mean ADC (r = -0.32). For both readers, ADC entropy (P = 0.028-0.003), but not mean ADC (P = 0.384-0.854), was a significant independent predictor of G4%. CONCLUSION: Whole-lesion ADC entropy outperformed mean ADC in characterizing Gleason 7 tumors and may help refine prognosis for this heterogeneous PCa subset.