Search papers, labs, and topics across Lattice.
Echo2ECG, a novel multimodal self-supervised learning framework, is introduced to enhance ECG representations by incorporating cardiac morphological information from multi-view echocardiograms. It addresses the limitations of existing methods that rely on single-view Echo alignment, which leads to a representational mismatch. By pretraining on paired ECG and multi-view Echo data, Echo2ECG learns ECG representations that outperform state-of-the-art unimodal and multimodal baselines in structural cardiac phenotype classification and Echo study retrieval tasks.
Multi-view Echo data can be used to train ECG encoders that are 18x smaller yet outperform larger models at predicting cardiac morphology.
Electrocardiography (ECG) is a low-cost, widely used modality for diagnosing electrical abnormalities like atrial fibrillation by capturing the heart's electrical activity. However, it cannot directly measure cardiac morphological phenotypes, such as left ventricular ejection fraction (LVEF), which typically require echocardiography (Echo). Predicting these phenotypes from ECG would enable early, accessible health screening. Existing self-supervised methods suffer from a representational mismatch by aligning ECGs to single-view Echos, which only capture local, spatially restricted anatomical snapshots. To address this, we propose Echo2ECG, a multimodal self-supervised learning framework that enriches ECG representations with the heart's morphological structure captured in multi-view Echos. We evaluate Echo2ECG as an ECG feature extractor on two clinically relevant tasks that fundamentally require morphological information: (1) classification of structural cardiac phenotypes across three datasets, and (2) retrieval of Echo studies with similar morphological characteristics using ECG queries. Our extracted ECG representations consistently outperform those of state-of-the-art unimodal and multimodal baselines across both tasks, despite being 18x smaller than the largest baseline. These results demonstrate that Echo2ECG is a robust, powerful ECG feature extractor. Our code is accessible at https://github.com/michelleespranita/Echo2ECG.
