The performance of three visual coding procedures and three computer programs in classification of electrocardiograms...

Area of Science:

• Cardiology
• Medical Informatics

Background:

• Electrocardiograms (ECGs) are crucial for diagnosing myocardial infarction (MI).
• The Minnesota Code standardizes ECG interpretation.
• Automated ECG analysis aims to improve efficiency and consistency.

Purpose of the Study:

• To evaluate the performance of visual and computer-based ECG coding procedures using the Minnesota Code.
• To compare the accuracy of different coding methods in identifying myocardial infarction.

Main Methods:

• A library of 228 myocardial infarction patient ECGs and 294 control subject ECGs was used.
• Three visual coding procedures and three computer programs were assessed.
• Performance was evaluated based on consistency and accuracy in classifying ECGs.

Main Results:

• Experienced visual coders showed higher consistency than less experienced teams.
• Computer program performance was comparable to visual coding.
• All methods demonstrated limited accuracy in discriminating infarcts from non-infarcts.
• 'Soft' criteria yielded higher sensitivity but lower specificity; 'hard' criteria improved specificity at the cost of sensitivity.

Conclusions:

• Computerized ECG coding according to the Minnesota Code is feasible.
• Combining automated coding with visual verification may enhance precision.
• Current coding criteria have limitations in accurately diagnosing myocardial infarction.