Date of Award

5-2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

First Advisor

Neji Mensi

Second Advisor

Vivek Sharma

Third Advisor

Hemant Purohit

Fourth Advisor

Brian A. Lail

Abstract

Cardiovascular disease remains the leading cause of preventable mortality worldwide, yet cardiac monitoring systems still rely on static two-dimensional waveform displays that require continuous expert interpretation and provide limited real-time insight. This work presents a real-time Electrocardiogram (ECG) Digital Twin framework that integrates sensing, embedded machine learning, wireless communication, and immersive visualization into a unified end-to-end system for continuous cardiac monitoring.

The proposed system acquires single-lead ECG signals at 125 Hz using a MAX30003 analog front-end interfaced with an ESP32-S3 microcontroller. A compact Convolutional Neural Network (CNN), quantized to a 38.7 kB TensorFlow Lite Micro model, performs five-class arrhythmia classification (Normal, Supraventricular Ectopic Beat, Ventricular Ectopic Beat, Fusion, and Unknown) following the AAMI EC57 standard. The model executes entirely on-device with a mean inference latency of 3.8 ms and achieves 98.28% accuracy on the inter-patient MIT-BIH DS2 benchmark.

Classified results are transmitted via MQTT over a secure cloud broker to a standalone Virtual Reality (VR) application developed in Unity. The VR system, CardioTwin, renders a real-time 3D cardiac model synchronized with the patient’s R–R interval and overlays live ECG signals, arrhythmia labels, and key cardiac metrics such as heart rate. A platform-aware transport abstraction enables seamless switching between communication interfaces without affecting the visualization pipeline.

The complete system achieves an end-to-end latency of 52.4 ± 11.3 ms while maintaining 88.7 frames per second on a standalone XR headset. This integrated approach demonstrates the feasibility of combining edge AI and immersive visualization to enhance real-time cardiac monitoring and support faster, more informed clinical decision-making.

Comments

This thesis presents a real-time ECG Digital Twin system integrating edge TinyML and immersive VR for cardiac monitoring.

Download

Available for download on Tuesday, May 09, 2028

Share

COinS