High Precision Digitization of Paper-Based ECG Records: A Step Toward Machine Learning

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Cardiology

Background:

• Electrocardiograms (ECGs) are crucial for diagnosing heart conditions, but traditional methods rely on outdated datasets and algorithms with limited accuracy.
• Improving ECG diagnostic accuracy necessitates leveraging historical ECG data, requiring conversion of scanned or printed ECGs into digital formats.
• Machine learning algorithms offer potential for enhanced ECG analysis, but require digitized raw ECG signal data (time in milliseconds, voltage in millivolts).

Purpose of the Study:

• To present a MATLAB-based tool and algorithm for converting printed or scanned electrocardiogram (ECG) formats into digitized ECG signals.
• To facilitate the use of historical ECG data for advanced analysis and diagnostic purposes.

Main Methods:

• Utilized 30 scanned ECG curves for the study.
• Implemented image processing techniques to detect regions of interest and extract ECG signals.
• Employed serial digitization and validation steps to ensure accuracy of the converted ECG data.

Main Results:

• Achieved very high correlation values for standard ECG parameters during validation.
• Demonstrated high precision for PR interval (0.984 +/-0.021), QRS interval (1 +/- SD), QT interval (0.981 +/-0.023), and RR interval (1 +/-0.001).
• P-values for all validated parameters were less than 0.001, indicating statistical significance.

Conclusions:

• Digitized ECG signals from historical paper or scanned sources can be obtained with over 95% precision.
• This digitization enables the application of machine learning algorithms to historical ECG data for identifying heart disease patterns.
• Facilitates improved diagnostic and prognostic evaluation of patients with cardiovascular disease using historical ECG information.