This study evaluated a fully-automated cartilage lesion detection system utilizing a deep convolutional neural network (CNN) to segment bone and cartilage followed by a second CNN classification network to detect structural abnormalities within the segmented tissues. The CNN network was trained to detect cartilage lesions within the knee joint using sagittal fat-suppressed T2-weighted fast spin-echo images in 125 subjects. The proposed CNN model achieved high diagnostic accuracy for detecting cartilage lesions with a 0.914 area under curve on receiver operation characteristics analysis. The optimal threshold for sensitivity and specificity of the CNN model was 84.3% and 84.6% respectively.
This abstract and the presentation materials are available to members only; a login is required.