Digitizing paper ECGs at scale: an open-source algorithm for clinical research.

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Computer Vision

Background:

• Billions of clinical electrocardiograms (ECGs) are preserved as paper scans, hindering their use in modern automated diagnostic systems.
• Lack of digital formats limits the accessibility and analytical potential of vast historical ECG archives.

Purpose of the Study:

• To develop and validate a fully automated, modular framework for converting scanned or photographed ECGs into usable digital signals.
• To improve the state-of-the-art in ECG image digitization for clinical and research applications.
• To promote reproducibility and further development through open-source software release.

Main Methods:

• Development of a modular, automated framework for ECG image processing.
• Validation on a large dataset of 37,191 ECG images, including data from Akershus University Hospital and the Emory Paper Digitization ECG Dataset.
• Evaluation of algorithm performance on images with various artifacts like perspective distortion, wrinkles, and stains.

Main Results:

• The framework achieved a mean signal-to-noise ratio of 19.65 dB on scanned ECGs with common artifacts.
• The model demonstrated superior performance across all subcategories compared to existing state-of-the-art methods.
• Successful digitization of ECGs from diverse sources, including those with significant image degradation.

Conclusions:

• The developed framework effectively converts scanned ECGs into digital signals, unlocking the potential of retrospective ECG archives.
• The open-source release of the software democratizes access to AI-driven diagnostics and encourages further innovation.
• This work significantly enhances the utility of historical ECG data for both clinical practice and scientific research.