Fig. 7From: Fully automated pelvic bone segmentation in multiparameteric MRI using a 3D convolutional neural networkSCORE results on DWI and ADC images. a The range of the CNN-predicted segmentations of sacrococcyx (yellow label) and ilium (deep blue label) was slightly more extensive than the manual segmentations (white background), attaining DSC values of 0.92 and 0.88, respectively. According to the SCORE system, 10 on the DWI image and 8 on the ADC image were obtained. b The range of the CNN-predicted segmentation of lumbar vertebra (light blue label) was slightly smaller than manual segmentation (white background) with a DSC value of 0.82. According to the SCORE system, 8 on the DWI image and 10 on the ADC image were obtainedBack to article page