The Synthesis Company of San Francisco Mountain Logo

Abstract

BACKGROUND: A critical responsibility in healthcare is the classification of cardiac diseases to facilitate early diagnosis and prevent cardiovascular consequences. Optimizing model performance and managing noisy and high-dimensional datasets are common issues for traditional approaches. OBJECTIVE: Using the MIT-BIH dataset, this paper describes a novel method for analyzing ECG signals in order to detect and classify various types of cardiovascular diseases (CVD). The proposed approach includes a comprehensive workflow that starts with dataset acquisition and progresses to signal buffering to prevent data loss and artifact removal. METHODS: To achieve optimal fusion, signal optimization is performed using Cat Swarm Optimization, and similar signals are clustered using Fuzzy C-Means (FCM). Signal decomposition and feature extraction are achieved using the Stockwell Transform and Empirical Mode Decomposition (EMD), respectively, while texture analysis is accomplished using Local Phase Quantization (LPQ). The classification is carried out using the ALEXIS Model, which combines AlexNet for feature learning, LSTM for temporal analysis, and SVM for final classification. RESULTS: The proposed method was compared to several other approaches, including CNN-LSTM-SE, XGBoost, and traditional machine learning methods. The proposed methodology outperformed all other methods, with an accuracy of 99.432%, sensitivity of 99.271%, specificity of 99.78%, and an F1 score of 99.311%. CONCLUSIONS: These findings demonstrate the integrated approach’s robustness and effectiveness in accurately classifying ECG signals into five types of CVD: congestive heart failure (CHF), valve stenosis, myocardial infarction, vascular disease, and arrhythmia. The high-performance metrics indicate that this methodology has the potential to significantly improve the accuracy and reliability of CVD diagnosis based on ECG signals.

High-Precision ECG Signal Classification for Cardiovascular Disease Detection with the ALEXIS Model

Abstract

Related Papers