Ischemia episode detection in ECG using kernel density estimation, support vector machine and feature selection.

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Myocardial ischemia can progress to severe cardiac conditions.
• Early and accurate detection of ischemic syndromes via electrocardiogram (ECG) is crucial for prevention.
• Current methods require improvement in accuracy and automation.

Purpose of the Study:

• To develop a new, accurate, and automated method for detecting myocardial ischemia using ECG.
• To employ wavelet transforms and feature selection for improved diagnostic capabilities.
• To enhance early diagnosis and prevent the progression of ischemic heart disease.

Main Methods:

• Utilized the European ST-T database (367 ischemic ST episodes).
• Applied discrete wavelet transform for baseline wandering removal and QRS complex detection.
• Extracted three key ECG waveform features and averaged them over five beats for outlier reduction.

Main Results:

• Kernel Density Estimation (KDE) achieved 0.939 sensitivity and 0.912 specificity.
• Support Vector Machine (SVM) achieved 0.941 sensitivity and 0.923 specificity.
• The proposed method successfully identified a high percentage of ischemic episodes.

Conclusions:

• A novel ECG method using wavelets and feature extraction effectively detects myocardial ischemia.
• The selected features are sufficient for discriminating ischemic from normal ST episodes.
• The KDE classifier offers automatic parameter selection, reducing manual input.