Graduation Term
Spring 2025
Degree Name
Master of Science (MS)
Department
School of Information Technology: Information Systems
Committee Chair
Qi Zhang
Committee Member
Yongning Tang
Committee Member
Chung-Chil Li
Abstract
Electrocardiogram (ECG) analysis is a fundamental diagnostic tool in cardiology, providing critical insights into cardiac function that directly impact patient care decisions and treatment outcomes. As healthcare systems face increasing demands, automated ECG interpretation using artificial intelligence offers promising solutions to improve diagnostic accuracy, reduce physicians workload, and enhance early detection of life threatening conditions.
This thesis compares two advanced deep learning architectures, CNN-GRU and Wide and Deep Transformer, for multi-label classification of 12-lead ECG data. Using data from the PhysioNet/Computing in Cardiology Challenge 2020, I evaluated both architectures across 27 different cardiac abnormalities. Results demonstrated that CNN-GRU architecture consistently outperformed Transformer models, achieving a validation AUC of 0.9169 compared to 0.8918 for the Transformer. The performance gap was most obvious for rare cardiac conditions, where CNN-GRU models successfully detected important abnormalities that Transformer models completely missed. Error analysis revealed both architectures struggled with signal quality variations and complex multi-label interactions, though CNN-GRU showed greater robustness overall.
This research provides insights into the strengths and limitations of different deep learning approaches for ECG analysis, establishing that combined convolutional-recurrent architectures are currently more effective than Transformer-based models for comprehensive cardiac abnormality detection. Future work should focus on developing hybrid architectures that addresses class imbalance issues and conduct more extensive validation across diverse patient populations. The findings have important implications for developing reliable clinical decision support systems for automated ECG interpretation that can improve cardiac care delivery.
Access Type
Thesis-Open Access
Recommended Citation
Kalal, Akhil Raghava, "Neural Network Algorithm and Analysis for Multi-Label ECG Data Classification" (2025). Theses and Dissertations. 2086.
https://ir.library.illinoisstate.edu/etd/2086
