Dr. Larisa Tereschenko is a researcher at OHSU who relies on Electrocardiogram (ECG) data for her work. A digitizer that is convenient to use in the lab will enable her to leverage large stores of paper ECG data and advance her research on heart disease. Furthermore, the free availability of this tool will enable the entire field of cardiovascular researchers to digitally analyze these paper-only data sets.
The heart generates electricity with every heartbeat, and by using electrodes we can measure the differences in electrical potentials. The shapes of the potential graphs can be analyzed by visual examination to detect abnormalities. This is still common practice for the diagnosis of heart diseases. This process of measuring the potentials of the heart requires sensitive equipment, because the electricity produced by the heart is minute. ECG devices graph the electrical potential of the heart and were first developed in the 1800s. In the early 1900s, ECG machines were accurate and cheap enough to begin use as clinical tools. The standard number of leads is 12, which requires 6 leads to be placed on the torso, and another 4 - on each arm and leg.
Now, we have cheap electronic devices (e.g., wearables) that make it remarkably easy to collect this data and provide large amounts of data. However, we still don’t have good applications of these tools. Research is necessary to discover what is useful and how to use it.
Unfortunately, large amounts of data are still trapped in paper print form. The original ECG machines were analogue, and they directly drew the potentials on paper. In the 70s and 80s these machines transitioned to digital, but the digital data was often discarded—since clinicians are primarily interested in the graphical format. While this paper format is useful for diagnosis, it is not ideal for algorithms that analyze the signals, which expect a digital signal input.
Previous tools for ECG digitizing and their limitations
There exists a closed-source, proprietary tool called ECGScan that was developed by Badilini, Erdem, Zareba, and Moss. The system removes the grid by ignoring repeating visual features, and traces the ECG waveforms via an active contour algorithm. ECGScan performs well in the digitization of multi and single lead ECGs, but is not fully automated. There is a lack of open-source ECG digitization tools, despite the numerous algorithms that have been developed to solve the technical challenges involved .