Hyperglycemia Identification Using ECG in Deep Learning Era

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Signal Processing

Background:

• Smart wearable biosensors are increasingly used in the medical Internet of Things (IoT).
• Electrocardiogram (ECG) signals are crucial for cardiovascular diagnostics.
• Non-invasive methods for hyperglycemia detection using ECG are actively researched.

Purpose of the Study:

• To propose a novel deep learning architecture for identifying hyperglycemia from ECG signals.
• To introduce an improved fiducial feature extraction technique to enhance classifier performance.
• To evaluate the efficacy of the proposed method in detecting hyperglycemia.

Main Methods:

• Development of a novel 10-layer deep neural network architecture.
• Implementation of a new fiducial feature extraction technique for ECG signals.
• Validation using ECG data from 1119 diverse subjects.

Main Results:

• The proposed algorithm achieved an Area Under the Curve (AUC) of 94.53%.
• High sensitivity (87.57%) and specificity (85.04%) were recorded for hyperglycemia detection.
• A relative performance improvement of 53% compared to existing literature models.

Conclusions:

• ECG signals contain intrinsic information indicative of blood glucose concentration.
• The developed deep learning model offers an effective and non-invasive method for hyperglycemia detection.
• The findings suggest a promising advancement in using wearable biosensors for diabetes management.