Congenital Heart Disease recognition using Deep Learning/Transformer models

TL;DR

This paper explores deep learning and transformer models for non-invasive congenital heart disease detection using sound and image data, achieving over 73% accuracy on two datasets.

Contribution

It introduces dual-modality deep learning approaches for CHD diagnosis, combining sound and image data for improved accuracy.

Findings

73.9% accuracy on ZCHSound dataset

80.72% accuracy on DICOM Chest X-ray dataset

Demonstrates effectiveness of multimodal deep learning for CHD detection

Abstract

Congenital Heart Disease (CHD) remains a leading cause of infant morbidity and mortality, yet non-invasive screening methods often yield false negatives. Deep learning models, with their ability to automatically extract features, can assist doctors in detecting CHD more effectively. In this work, we investigate the use of dual-modality (sound and image) deep learning methods for CHD diagnosis. We achieve 73.9% accuracy on the ZCHSound dataset and 80.72% accuracy on the DICOM Chest X-ray dataset.