Abstract: To evaluate the diagnostic value of radiomics based on high-resolution T2WI combined with apparent diffusion coefficient imaging (ADC) for prostate cancer. Data from 120 patients who underwent MRI and Confirmed by pathology were reviewed by a retrospective study, including 71 prostate cancer and 49 no-prostate cancer. The lesions of these patients were divided into peripheral zone (PZ) and central zone (CZ) groups, respectively. Radcloud platform was used for data analysis of MRI images, and the area under the curve (AUC) was calculated by receiver operating characteristic curve (ROC). T2WI, ADC, T2WI+ADC computer aided diagnostic model and Prostate Imaging Report and Data System version 2 (PI-RADS v2) were compared in 247distinguishing benign and malignant prostate lesion. The results show in the test set, the performance of Model T2WI+ADC was significantly higher than Model T2WI and PI-RADS in the peripheral zone (P<0.05), Model ADC was significantly higher than Model T2WI and PI-RADS in the central zone (P<0.05). For PI-RADS 3 lesion, the area under curve of the Model T2WI+ADC were 0.89 and 0.94 in the peripheral zone and central zone, respectively. Radiomics based on high-resolution T2WI combined with apparent diffusion coefficient imaging achieved a higher diagnostic efficiency than PI-RADS v2, including PI-RADS 3 lesion.

Key words: radiomics, prostate cancer, magnetic resonance imaging, PI-RADS v2