Integrating Multimodal Deep Learning Architectures and Wearable Biosensor Telemetry for Real-Time Prediction of Adverse Cardiovascular Outcomes in Post-Discharge Ischemic Heart Disease Patients

• Abilly Elly

  Texas University

  Author

Ischemic heart disease (IHD) remains a leading cause of global morbidity and mortality, with post-discharge patients facing a persistently high risk of adverse cardiovascular outcomes. Current risk stratification tools, predominantly reliant on episodic clinical assessments, fail to capture the dynamic physiological changes that precede acute events. This prospective cohort study addresses this critical gap by integrating multimodal deep learning architectures with continuous wearable biosensor telemetry to enable real-time prediction of major adverse cardiovascular events (MACE) in post-discharge IHD patients. A hybrid deep learning framework combining Convolutional Neural Networks (CNNs) for spatial feature extraction from electrocardiogram (ECG) and photoplethysmography (PPG) waveforms with a Temporal Attention-Gated Recurrent Unit (TA-GRU) for capturing temporal dependencies was developed and validated on a prospective cohort of 3,247 patients monitored over eight months post-discharge . The proposed framework achieved exceptional predictive performance with an area under the curve (AUC) of 0.89, surpassing traditional machine learning models (AUC = 0.86) and conventional risk scores (AUC = 0.81) . The model demonstrated 89.4% sensitivity at a specificity of 85.2%, with a mean early warning lead time of 4.7 hours prior to clinical manifestation. This research provides a validated framework for transforming continuous physiological data into actionable clinical intelligence, supporting the paradigm shift from reactive to predictive cardiovascular care.

• PDF

Copyright (c) 2026 Abilly Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.