Tunable Q-factor wavelet transform based identification of diabetic patients using ECG signals

Area of Science:

• Biomedical Engineering
• Computational Medicine
• Cardiology

Background:

• Diabetes mellitus is a chronic condition causing elevated blood glucose, leading to widespread organ damage.
• Diabetic cardiovascular autonomic dysfunction alters electrocardiogram (ECG) signal morphology.
• Automated detection of diabetes through ECG analysis remains a significant challenge.

Purpose of the Study:

• To develop and validate a machine learning-based system for automated diabetes detection using ECG signals.
• To investigate the efficacy of multi-resolution analysis and entropy-based features for diabetes identification.
• To compare the performance of the proposed method against existing techniques.

Main Methods:

• ECG signals were decomposed into sub-bands using the tunable Q-factor wavelet transform (TQWT).
• Four entropy-based features were extracted from each sub-band and selected using the K-W test.
• A Support Vector Machine (SVM) classifier with 10-fold cross-validation was employed for automated detection.

Main Results:

• The $$ sub-band of TQWT combined with an SVM classifier using a Coarse Gaussian kernel achieved 91.5% classification accuracy.
• The proposed method demonstrated superior performance compared to other existing methods on the same dataset.
• Feature selection using the K-W test effectively identified relevant characteristics for diabetes detection.

Conclusions:

• The developed machine learning system offers a promising, non-invasive approach for automated diabetes detection.
• Multi-resolution analysis of ECG signals, particularly specific TQWT sub-bands, holds significant potential for identifying diabetic complications.
• The findings suggest that ECG-based analysis can serve as a valuable tool for early diabetes screening and management.