Leveraging 3D Heart Visualisation and Data Balancing Techniques for ECG Classification

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Cardiovascular diseases are a major global health concern.
• Electrocardiogram (ECG) analysis is crucial but can be time-consuming and error-prone.
• Automated methods are needed to improve accuracy and efficiency in cardiac abnormality diagnosis.

Purpose of the Study:

• To develop a deep learning pipeline for automated arrhythmia classification using ECG data.
• To address challenges posed by imbalanced datasets in cardiac analysis.
• To introduce a 3D visualization framework for enhanced diagnostic insight and anatomical localization.

Main Methods:

• A comprehensive deep learning pipeline was designed for automated arrhythmia classification.
• Specific data balancing strategies were implemented to handle imbalanced datasets.
• A novel 3D visualization framework was developed for interactive anatomical rendering.

Main Results:

• The proposed data balancing techniques significantly improved classification performance.
• The automated system achieved competitive or superior results compared to existing methods.
• 3D visualizations provided precise anatomical localization of arrhythmia substrates.

Conclusions:

• The deep learning pipeline offers a promising approach for accurate automated arrhythmia classification.
• The 3D visualization tool enhances clinical diagnosis and medical education.
• Further inter-patient cross-validation is recommended to establish generalizability.