Feasibility and repeatability of DWI-derived diffusion heterogeneity for discriminating clinically significant prostate cancer.

PURPOSE: To evaluate feasibility and repeatability of entropy derived directly from diffusion weighted imaging (DWI) of prostate cancer (PCa), and compare its performance to entropy derived from the conventional monoexponential model calculated apparent diffusion coefficient (ADC). METHODS: 112 men with PCa underwent two repeated 3 T MRIs before prostatectomy. DWI was obtained using 12 b-values (0-2000 s/mm2). DWI entropy was derived directly from source DWI image using b-values of 100 s/mm2, 500 s/mm2, and 900 s/mm2. ADC based radiomic features were calculated for comparison. Repeatability was assessed using intraclass correlation coefficient, ICC (3,1). Discrimination of clinically significant PCa (csPCa) and correlations with Gleason Grade Group (GGG) were assessed using area under receiver operating curve (AUC) and Spearman's ρ, respectively. RESULTS: DWI entropy demonstrated better repeatability (ICC = 0.81; 95% CI: 0.74-0.87) than ADC entropy (ICC = 0.57; 95% CI: 0.43-0.68). DWI entropy (AUC = 0.85; 95% CI: 0.74-0.96) showed a numerically higher AUC than ADC entropy (AUC = 0.75; 95% CI: 0.58-0.92) in csPCa discrimination, although the difference was not statistically significant (P = 0.176). DWI entropy showed a moderate correlation with GGG (ρ = 0.5; 95% CI: 0.3-0.6), while ADC entropy demonstrates weak correlation (ρ = 0.4; 95% CI: 0.2-0.5), although the difference was not statistically significant (P = 0.056). CONCLUSION: Entropy derived directly from DWI demonstrates potential for discriminating csPCa. Prospective validation in broader diagnostic population is warranted.