Multi-arrhythmias detection with an XML rule-based system from 12-Lead Electrocardiogram

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Cardiology

Background:

• Computer-aided detection of cardiac arrhythmias is vital in medical technology.
• Rule-based systems (RBS) provide transparent and interpretable results for arrhythmia detection.

Purpose of the Study:

• To aid cardiologists in diagnosis and reduce diagnostic uncertainty.
• To develop and evaluate an automated system for cardiac arrhythmia classification and recognition.

Main Methods:

• Utilized XML rules encoding cardiologist knowledge for arrhythmia classification.
• Conducted 13 experiments with varying knowledge bases to optimize detection of 13 cardiac arrhythmias.
• Employed an international dataset with 279 features from 12-lead ECG and patient data, using the "XMLRULE" software.

Main Results:

• Mean correct detection rate of 82.80% across 12 experiments for single arrhythmia detection.
• The final experiment, detecting 12 arrhythmias simultaneously, yielded 38.33% correct detection, with 90.55% sensitivity and 46.24% specificity.
• Performance for single arrhythmia detection surpassed other computational methods, but multi-arrhythmia detection was less successful.

Conclusions:

• Rule-based systems (RBS) are highly transparent for cardiac arrhythmia detection.
• While RBS excel in recognizing individual arrhythmias, rule conflicts and measure uncertainty limit multi-arrhythmia classification accuracy compared to other methods.