REM-Net: MEMS-Based Synchronized PCG-ECG Analysis Framework for High-Precision CAD Diagnosis

Early detection of coronary artery disease (CAD) demands both high-fidelity signal acquisition and sophisticated analysis methods to capture subtle cardiac abnormalities. Traditional approaches face challenges in achieving precise synchronization and maintaining signal quality across different modalities. We address these limitations by introducing a high-sensitivity micro-electromechanical systems (MEMSs)-based synchronous auscultation system that enables unprecedented precision in simultaneous phonocardiogram–electrocardiogram (PCG-ECG). The MEMS sensors, featuring superior sensitivity, low noise floor, and excellent frequency response, capture detailed cardiac acoustic and electrical signals with minimal distortion and precise temporal alignment. Building upon this high-quality signal acquisition foundation, we propose ResNet18-Encoder-MCA (REM-Net), an innovative deep-learning framework that integrates ResNet18-based feature extraction with a mutual-cross-attention (MCA) mechanism to fully exploit the rich, synchronized cardiac information. Extensive experiments demonstrate that our MEMS-enabled approach achieves exceptional performance, with 98.81% accuracy (Acc) and 98.83% F1 Score on our collected data. The framework’s robustness is further validated on the 2016 PhysioNet/CinC Challenge dataset, achieving 99.05% Acc and 98.79% F1 Score. These results demonstrate that the combination of high-precision MEMS-based signal acquisition and advanced deep-learning analytics not only advances the state-of-the-art (SOTA) in CAD diagnosis but also provides a reliable, noninvasive, and cost-effective solution for clinical applications. Our approach represents a significant step forward in early CAD screening, where the superior signal quality enabled by MEMS technology proves crucial for accurate diagnosis.