Enhancing heart disease diagnosis through ECG image vectorization-based classification

Area of Science:

• Biomedical Engineering
• Cardiology
• Artificial Intelligence in Healthcare

Background:

• Heart disease poses a significant health challenge, necessitating early detection and prevention strategies.
• Electrocardiograms (ECG) are crucial for diagnosing cardiac conditions.
• Automating ECG analysis and feature extraction can improve diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop an efficient image-vectorization technique for ECG analysis.
• To enhance the accuracy of ECG-based heart disease classification models.
• To reduce computational resources required for ECG analysis.

Main Methods:

• An image-vectorization technique involving image cropping, grid line removal, and pixel assignment was developed.
• Feature vectors were extracted using this technique and compared to those from VGG16.
• Artificial neural networks (ANNs) were trained using the image-vectorization-derived feature dataset.

Main Results:

• The image-vectorization technique produced feature vectors 589 times shorter than VGG16.
• This reduction significantly decreased memory requirements and computational power needs.
• Machine learning algorithms trained with image-vectorization showed improved performance over conventional methods like CNNs and VGG16.

Conclusions:

• The proposed image-vectorization technique offers a highly efficient method for ECG feature extraction.
• This approach enhances the performance of machine learning models for heart disease detection.
• The technique holds promise for more effective and accessible online detection and prevention of heart conditions.