A Fully-Automated Paper ECG Digitisation Algorithm for e-Health Applications

The electrocardiogram (ECG) is an important non-invasive tool for diagnosing heart and cardiovascular diseases. However, traditional ECG systems often provide images, which makes it challenging to directly utilize the image-based data in various applications that require 1-D signals. Therefore, the digitalization of ECG reports is essential for the establishment of electronic medical archives, advanced diagnostic tools, and intelligent healthcare systems. Previous efforts to digitize paper-based ECGs have been limited by the need to manually select regions of interest, limited accuracy in handling overlapping ECG leads, and lack of batch processing capabilities. We present a fully automated method for digitizing ECG reports that requires only a predefined lead distribution, eliminating the need for manual lead position selection. This method effectively addresses the challenge of overlapping leads and has the capability for automated batch processing. Validation against a subset of the PTB-XL dataset demonstrated high accuracy and reliability (Root Mean Square Error: 0.036 (mV), Pearson Correlation Coefficient: 0.977). To further verify the generalizability of the method, we applied the algorithm to hospital ECG reports. The results, evaluated by expert cardiologists, confirmed that our method produced consistently high-quality digitalization and demonstrated its applicability to different types of ECG reports.