Scalar invariant transform based deep learning framework for detecting heart failures using ECG signals

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Heart disease is a leading global cause of mortality.
• Early detection and treatment are crucial for improving patient outcomes.
• Electrocardiogram (ECG) analysis offers insights into cardiac health by monitoring heartbeat fluctuations.

Purpose of the Study:

• To develop and evaluate a novel automated Convolutional Neural Network (CNN) system for accurate cardiac disease detection.
• To leverage Scale-Invariant Feature Transform (SIFT) for enhanced feature extraction from ECG signals.
• To classify ECG signals into Arrhythmia (ARR), Congestive Heart Failure (CHF), and Normal Sinus Rhythm (NSR).

Main Methods:

• Utilized a custom Convolutional Neural Network (CNN) architecture.
• Employed Scale-Invariant Feature Transform (SIFT) for extracting unique features from ECG signal images.
• Compared SIFT with other feature extraction techniques like HOG and SURF.

Main Results:

• The SIFT-CNN model achieved an accuracy of 99.78% and an F1 score of 99.78% on a dataset of 162 ECG images.
• Achieved superior performance compared to models using HOG (99.45% accuracy) and SURF (78% accuracy).
• Demonstrated high classification accuracy for detecting Arrhythmia, Congestive Heart Failure, and Normal Sinus Rhythm.

Conclusions:

• The proposed SIFT-CNN model represents a significant advancement in automated cardiac disease detection.
• Combining SIFT feature extraction with a custom CNN model offers a novel and highly effective approach.
• This method shows exceptional performance, outperforming existing models for classifying heart conditions from ECG data.