Extended Data Table 5 DNN performance by demographic strata in the UCSF test dataset
From: Multiview deep learning improves detection of major cardiac conditions from echocardiography
Left/Right Ventricular Abnormality (value (95%CI)) | AUC | Sensitivity | Specificity | F1 |
|---|---|---|---|---|
Males (N=3002) | 0.904 (0.894–0.914) | 0.826 (0.805–0.847) | 0.81 (0.796–0.824) | 0.725 (0.707–0.744) |
Females (N=3241) | 0.907 (0.894–0.918) | 0.79 (0.761–0.817) | 0.864 (0.853–0.875) | 0.655 (0.632–0.678) |
Age 18–60 (N=2550) | 0.912 (0.900–0.922) | 0.834 (0.811–0.861) | 0.84 (0.826–0.854) | 0.723 (0.701–0.745) |
Age 61- (N=3695) | 0.905 (0.895–0.915) | 0.795 (0.773–0.818) | 0.84 (0.828–0.851) | 0.676 (0.655–0.695) |
Diastolic Dysfunction (value (95%CI)) | AUC | Sensitivity | Specificity | F1 |
Males (N=1069) | 0.819 (0.795–0.843) | 0.718 (0.692–0.745) | 0.758 (0.719–0.799) | 0.790 (0.769–0.809) |
Females (N=1145) | 0.851 (0.832–0.872) | 0.755 (0.729–0.780) | 0.787 (0.750–0.820) | 0.812 (0.793–0.830) |
Age 18–60 (N=721) | 0.830 (0.805–0.854) | 0.543 (0.490–0.592) | 0.876 (0.851–0.901) | 0.624 (0.579–0.665) |
Age 61- (N=1493) | 0.752 (0.726–0.780) | 0.779 (0.760–0.797) | 0.589 (0.540–0.644) | 0.838 (0.824–0.851) |
Valve Regurgitation (value (95%CI)) | AUC | Sensitivity | Specificity | F1 |
Males (N=2088) | 0.903 (0.887–0.919) | 0.849 (0.815–0.885) | 0.821 (0.806–0.835) | 0.564 (0.532–0.598) |
Females (N=2060) | 0.903 (0.886–0.918) | 0.795 (0.748–0.839) | 0.837 (0.822–0.850) | 0.498 (0.454–0.535) |
Age 18–60 (N=1715) | 0.926 (0.911–0.941) | 0.808 (0.756–0.857) | 0.879 (0.865–0.893) | 0.552 (0.506–0.598) |
Age 61- (N=2433) | 0.888 (0.872–0.903) | 0.835 (0.799–0.869) | 0.792 (0.777–0.807) | 0.526 (0.493–0.557) |
