Abstract
Cardiovascular disease (CVD) is the leading cause of death worldwide. Hypertrophic Cardiomyopathy (HCM) is the most common genetic disease in which the heart’s Left Ventricular (LV) wall becomes thicker and stiffer, making it difficult to pump blood. HCM affects 1:200 to 1:500 people and can result in Sudden Cardiac Death (SCD), heart failure, and abnormal heart rhythms leading to stroke. Early diagnosis and treatment of HCM can improve outcomes. An echocardiogram, a heart ultrasound, is routinely performed on patients and is currently the gold standard for HCM diagnosis. However, expert analyses of echocardiograms can be inconsistent, resulting in missed diagnoses. Deep Video Action Recognition (VAR) models have achieved state-of-the-art performance for the task of recognizing human actions, such as running and walking, in a video. In this paper, we innovatively propose HCM-Dynamic-Echo, an end-to-end deep learning framework that uses the SlowFast VAR architecture, for the binary classification of echocardiogram videos as having HCM vs. normal. SlowFast has two arms: arm 1 (slow pathway) analyzes spatial features, while arm 2 (fast pathway) captures temporal structural information to increase video recognition accuracy. Furthermore, we employed transfer learning, pre-training HCM-Dynamic-Echo on the large Stanford EchoNet-Dynamic echocardiogram dataset, enabling HCM detection in a smaller echocardiogram video dataset. In rigorous evaluation, HCM-Dynamic-Echo outperformed state-of-the-art baselines, achieving an accuracy of 93.13%, a F1-score of 92.98%, Positive Predictive Value (PPV) of 94.64%, specificity of 94.87%, and an Area Under the Curve (AUC) of 93.13%. To the best of our knowledge, our work is the first that innovatively utilized the SlowFast VAR architecture for predicting HCM in racially and ethnically diverse echocardiogram videos.