Figure 1 | Scientific Reports

Figure 1

From: A deep learning framework assisted echocardiography with diagnosis, lesion localization, phenogrouping heterogeneous disease, and anomaly detection

Figure 1

Overview of the study design of AIEchoDx. (a) The echocardiogram videos and corresponding electronic medical records were collected for five diagnostic classes, including atrial septal defect (ASD), dilated cardiomyopathy (DCM), hypertrophy cardiomyopathy (HCM), prior myocardial infarction (prior MI), and normal subjects (Normal). (b) Examples of apical 4-chamber view (A4c) images of ASD, DCM, HCM, prior MI, and Normal (LV, left ventricle; LA, left atrium; RV, right ventricle; RA, right atrium). The white arrows displayed the morphology changes and disease hallmarks for clinical diagnosis. (c–e) The distributions of patients (c), videos (d), and single A4c slice per class (e) in the echocardiography dataset to train the deep neural network. (f) The distributions of the training, validation dataset, and test dataset 1. (g) The architecture of the AIEchoDx framework for analyzing echocardiogram videos.

Back to article page