Extended Data Fig. 1: Performance of video-based echo 21-view classifier and color doppler detector.
From: Multiview deep learning improves detection of major cardiac conditions from echocardiography
(A) Receiver operating characteristic curves and (B) confusion matrix for the 21 class echo view classifier. The view classifier DNN achieved a mean AUC of 0.972 across 21 classes. (C) Receiver operating characteristic curve and (D) confusion matrix for the DNN color doppler classifier. The color doppler classifier achieved an AUC of 0.991. View class number view labels: 0: parasternal long axis, 1: parasternal long axis (zoom on left atrium), 2: parasternal long axis (standard zoom), 3: apical two chamber, 4: apical two chamber (left atrium occluded), 5: apical two chamber (left ventricle occluded), 6: apical three chamber, 7: apical three chamber (left atrium occluded), 8: apical three chamber (left ventricle occluded), 9: apical four chamber, 10: apical four chamber (left ventricle occluded), 11: apical four chamber (left atrium occluded), 12: apical five chamber,13: parasternal short axis apical level 14: parasternal short axis (mitral valve level), 15: parasternal short axis (zoom on left atrium), 16: parasternal short axis (aortic valve zoom), 17: subcostal, 18: suprasternal, 19: right ventricle inflow, 20: other.
