Detection and localization of myocardial infarction using K-nearest neighbor classifier

Area of Science:

• Biomedical Engineering
• Cardiology
• Machine Learning in Healthcare

Background:

• Myocardial infarction (MI) diagnosis relies on ECG analysis.
• Accurate detection and localization of MI are crucial for timely treatment.
• Existing methods may have limitations in accuracy or computational efficiency.

Purpose of the Study:

• To develop and evaluate an automated system for detecting and localizing myocardial infarction (MI) using ECG data.
• To assess the performance of the K-nearest neighbor (KNN) classifier with time-domain ECG features for MI.
• To investigate the impact of data pruning on diagnostic accuracy and computational cost.

Main Methods:

• Extraction of time-domain ECG features (T wave amplitude, Q wave, ST level deviation) from 12-lead ECG signals.
• Application of the K-nearest neighbor (KNN) classifier for MI detection and localization.
• Utilizing the PTB database comprising 20,160 ECG beats for analysis.
• Implementation of the Arif-Fayyaz pruning algorithm to optimize data storage and computation.

Main Results:

• KNN achieved 99.9% sensitivity and specificity for MI detection without pruning.
• Data pruning reduced training by 93% with a slight decrease in sensitivity (97%) and specificity (99.6%).
• Classification accuracy above 90% for all eleven classes (10 MI locations + normal), with an overall accuracy of 98.8% before pruning and 98.3% after pruning.

Conclusions:

• The proposed automated KNN-based method demonstrates high accuracy and simplicity for MI detection and localization.
• Data pruning effectively reduces computational load and storage requirements with minimal impact on diagnostic performance.
• This approach holds significant potential for improving MI diagnosis in practical clinical settings.