Integrating electrocardiogram and fundus images for early detection of cardiovascular diseases

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Cardiology

Background:

• Cardiovascular diseases (CVD) pose a significant global health challenge.
• Advanced diagnostic tools are crucial for early detection and management of CVDs.
• The retina's vascular network mirrors cardiovascular health, and ECG provides cardiac insights.

Purpose of the Study:

• To develop an avant-garde methodology for early cardiovascular disease (CVD) tagging and triaging.
• To integrate electrocardiogram (ECG) readings and retinal fundus images for a holistic diagnostic perspective.
• To enhance CVD classification accuracy using a synergistic approach.

Main Methods:

• Applied Fast Fourier Transform (FFT) to ECG and retinal fundus images to convert data into the frequency domain.
• Computed Earth Mover's Distance (EMD) on frequency-domain features from both modalities.
• Concatenated EMD values to create a comprehensive feature set for a Neural Network classifier.

Main Results:

• Achieved a preliminary accuracy of 84% in classifying cardiovascular diseases.
• Demonstrated the potential of integrating spectral insights from FFT and nuanced data differences captured by EMD.
• Validated the robustness of the combined diagnostic strategy for CVD classification.

Conclusions:

• The synergistic integration of ECG and retinal fundus images shows significant promise for early CVD detection and triaging.
• The developed methodology offers a robust representation for CVD classification, leveraging FFT and EMD.
• Further refinement and validation are anticipated to enhance clinical applicability, particularly in resource-limited settings.