Interpretable Anomaly Detection in Echocardiograms with Dynamic Variational Trajectory Models

Alain Ryser, Laura Manduchi, Fabian Laumer, Holger Michel, Sven Wellmann, Julia E Vogt

Abstract: We propose a novel anomaly detection method for echocardiogram videos. The introduced method takes advantage of the periodic nature of the heart cycle to learn different variants of a variational latent trajectory model (TVAE). The models are trained on the healthy samples of an in-house dataset of infant echocardiogram videos consist- ing of multiple chamber views to learn a nor- mative prior of the healthy population. During inference, maximum a posteriori (MAP) based anomaly detection is performed to detect out-of- distribution samples in our dataset. The proposed method reliably identifies severe congenital heart defects, such as Ebstein’s Anomaly or Shone- complex. Moreover, it achieves superior perfor- mance over MAP-based anomaly detection with standard variational autoencoders on the task of detecting pulmonary hypertension and right ven- tricular dilation. Finally, we demonstrate that the proposed method provides interpretable ex- planations of its output through heatmaps which highlight the regions corresponding to anomalous heart structures.