Coronary artery calcium (CAC) is biomarker of advanced subclinical coronary artery disease and predicts myocardialinfarction and death prior to age 60 years. The slice-wise manual delineation has been regarded as the gold standard ofcoronary calcium detection. However, manual efforts are time and resource consuming and even impracticable to beapplied on large-scale cohorts. In this paper, we propose the attention identical dual network (AID-Net) to perform CACdetection using scan-rescan longitudinal non-contrast CT scans with weakly supervised attention by only using per scanlevel labels. To leverage the performance, 3D attention mechanisms were integrated into the AID-Net to providecomplementary information for classification tasks. Moreover, the 3D Gradient-weighted Class Activation Mapping(Grad-CAM) was also proposed at the testing stage to interpret the behaviors of the deep neural network. 5075 non-contrastchest CT scans were used as training, validation and testing datasets. Baseline performance was assessed on the samecohort. From the results, the proposed AID-Net achieved the superior performance on classification accuracy (0.9272) andAUC (0.9627).
展开▼
