Diagnosis of cardiovascular abnormalities from compressed ECG: a data mining-based approach

Area of Science:

• Biomedical Engineering
• Health Informatics
• Data Science

Background:

• Compressed electrocardiogram (ECG) signals are essential for efficient telecardiology due to large data sizes.
• Conventional ECG diagnosis requires decompression, introducing delays critical for cardiovascular disease (CVD) patients.
• Real-time detection of cardiac abnormalities is crucial for timely intervention.

Purpose of the Study:

• To develop an innovative technique for real-time classification of CVD directly from compressed ECG data.
• To enable automatic notification systems for life-threatening cardiac abnormalities.
• To reduce diagnostic delays in telecardiology applications.

Main Methods:

• Utilizes data mining techniques including attribute selection and Expectation Maximization (EM)-based clustering on a hospital server.
• Generates a set of constraints representing cardiac abnormalities from compressed ECG data.
• Employs a rule-based system on a patient's mobile phone for real-time abnormal beat identification.

Main Results:

• The proposed system achieves an average accuracy of 97% in detecting various cardiac abnormalities.
• Successfully identifies conditions such as ventricular flutter/fibrillation, premature ventricular contraction, and atrial fibrillation.
• Demonstrates the feasibility of direct analysis of compressed ECG data.

Conclusions:

• The data mining technique on compressed ECG packets enables faster cardiac abnormality identification.
• The proposed system significantly enhances the efficiency of telecardiology diagnosis.
• This approach facilitates rapid detection and notification for critical cardiovascular events.