An automated data extraction system from 12 lead ECG images

Area of Science:

• Biomedical Engineering
• Digital Signal Processing

Background:

• Traditional electrocardiogram (ECG) data is often recorded on paper, posing challenges for digital analysis and long-term storage.
• Converting analog ECG records to a digital format is crucial for modern healthcare and research.

Purpose of the Study:

• To develop a software-based system for normalizing and acquiring ECG data from both normal and abnormal records.
• To enable the transfer of wave data from paper to a digital time database.

Main Methods:

• Utilizing a flatbed scanner to create an image database of 12-lead ECG signals.
• Converting TIF gray-tone images to binary images using histogram analysis.
• Employing smearing runlength and thinning algorithms to process images and extract skeletal data.
• Regenerating ECG signals from pixel coordinate data and analyzing database consistency via standard deviation.

Main Results:

• Successfully converted scanned ECG images into a digital database.
• Developed a method to extract essential ECG signal data, minimizing redundant points.
• Demonstrated the consistency of the digitized ECG database through graphical analysis.

Conclusions:

• The developed software system provides an effective method for digitizing and normalizing ECG data from paper records.
• This system facilitates the creation of a reliable digital ECG database for further analysis and research.