Skip to main content
. 2022 Nov 21;50(3):727–741. doi: 10.1007/s00259-022-06036-9

Fig. 4.

Fig. 4

Diagnostic performance of PI-RADS, DL-CS-Res, and PIDL-CS for the detection of csPCa in the external validation cohorts. a The sensitivities and specificities of PI-RADS, DL-CS-Res and PIDL-CS in three external validation cohorts at chosen thresholds; b threshold points of DL-CS-Res and PI-RADS assessment for the detection of csPCa in three external validation cohorts; c threshold points of PIDL-CS and PI-RADS assessment for the detection of csPCa in three external validation cohorts. Receiver operating characteristics curves of DL-CS-Res and PIDL-CS are red lines, and those of PI-RADS assessment are blue lines. Asterisk symbol means significant (P < 0.05). Dagger symbol: compared with the PI-RADS assessment with the threshold of ≥ 3. Abbreviation: DL-CS-Res, the deep learning model based on ResNet3D network for the classification between clinically significant and non-clinically significant prostate cancer; PI-RADS, Prostate Imaging and Reporting and Data System; PIDL-CS, integrated model combining DL-CS-Res and PI-RADS assessment; SKH, Suzhou Kowloon Hospital; TZH, the People’s Hospital of Taizhou; SQH, the People’s Hospital of Suqian; CSH, Changshu No.1 People’s Hospital; mpMRI, multiparametric MRI