Detection of Cardiovascular Disease with Minimal Leads Using efficient Machine Learning Techniques

• Published in: 2024 IEEE International Conference for Women in Innovation, Technology & Entrepreneurship (ICWITE) pp. 368 - 374
• Main Authors: Shaji, Shereena, Pathinarupothi, Rahul Krishnan, Unnikrishna Menon, K A
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.02.2024
• Subjects: Amrita Spandanam (ASMD); Artificial neural networks; Cardiovascular diseases; Classification algorithms; Data models; EDA (Exploratory Data Analysis); Electrocardiography; Filtering; Heart Rate Variability; Machine learning algorithms; Motion Artifacts; PCA; Training data
• DOI: 10.1109/ICWITE59797.2024.10503171

Medical experts in India extensively use a 12-lead ECG to detect cardiac issues and other illnesses. The majority of cardiac abnormalities are signs of chronic heart failure. As a result, they need to be diagnosed right away. An ECG is automatically classified, allowing for earlier diagnosis than a clinician could. Even though the 12 lead ECG is the standard diagnosing system, there is a need for a minimal lead ECG screening system because it has several advantages, including fewer leads, which reduces wiring complexity, ease of use, and collaboration with wearable devices. In this paper, we proposed various machine learning models to predict rhythm/morphology problems in 12-lead ECGs that last a few seconds to a few minutes. We have used four machine learning approaches, Logistic Regression, Support vector machine, Decision Tree (DT), and Artificial Neural Networks (ANN) in the full sets of features. Results were evaluated based on accuracy, and precision. The best results were obtained with ANN and Decision Tree algorithm an accuracy of 50%. CNN for classifying cardiovascular diseases like 2- Atrial fibrillation (AF) and 5- Right bundle branch block (RBBB) with an accuracy of 0.978 for training data and 0.663 for the testing data.