Myocardial infarction detection using ITD, DWT and deterministic learning based on ECG signals.

Area of Science:

• Cardiology
• Biomedical Signal Processing
• Artificial Intelligence

Background:

• Cardiovascular diseases (CVD), particularly myocardial infarction (MI), are leading causes of mortality worldwide.
• Current MI diagnosis relies on subjective visual inspection of electrocardiogram (ECG) and vectorcardiogram (VCG) signals, which is time-consuming and prone to error.
• Automated detection methods are needed to overcome the limitations of traditional MI diagnosis.

Purpose of the Study:

• To develop a novel, automated technique for detecting myocardial infarction (MI).
• To synthesize 12-lead ECG and Frank XYZ leads to create a hybrid cardiac vector for enhanced analysis.
• To leverage advanced signal processing and artificial intelligence for accurate MI classification.

Main Methods:

• Synthesized a 4-dimensional cardiac vector from 12-lead ECG and Frank XYZ leads.
• Applied intrinsic time-scale decomposition (ITD) to extract proper rotation components (PRCs) reflecting cardiac dynamics.
• Utilized discrete wavelet transform (DWT) and 3D phase space reconstruction for feature extraction from predominant PRCs.
• Employed neural networks for classification of healthy and MI cardiac vector signals.

Main Results:

• The proposed method achieved an average classification accuracy of 98.20% on the PhysioNet PTB database.
• Experiments included data from 148 MI patients and 52 healthy controls, using tenfold cross-validation.
• The technique effectively captured disparities in cardiac system dynamics between healthy and MI subjects.

Conclusions:

• The developed method demonstrates high effectiveness for automatic MI detection.
• The novel approach of synthesizing cardiac vectors and analyzing their dynamics shows promise for clinical application.
• This technique offers a potential advancement in the objective and efficient diagnosis of myocardial infarction.