Authors: Kshama Kodthalu Shivashankara, Deepanshi, Afagh Mehri Shervedani, Gari D. Clifford, Matthew A. Reyna, Reza Sameni

Published on: July 04, 2023

Impact Score: 8.3

Arxiv code: Arxiv:2307.01946

Summary

• What is new: Introduction of ECG-Image-Kit, a tool for generating synthetic ECG images with realistic artifacts from time-series data, to aid in digitizing paper ECGs.
• Why this is important: Paper-based ECGs are incompatible with advanced diagnosis software due to lack of digitization, hindering the use of machine learning for ECG diagnosis.
• What the research proposes: Using ECG-Image-Kit for creating synthetic ECG images and a deep learning pipeline to convert these images back into time-series data for analysis.
• Results: The pipeline accurately digitizes ECGs, maintaining clinical parameters essential for diagnosis, showcasing the effectiveness of generative digitization approaches.

Technical Details

Technological frameworks used: ECG-Image-Kit, combination of traditional computer vision and deep learning methods

Models used: Deep neural networks for converting synthetic ECG images into time-series data

Data used: 21,801 ECG images from the PhysioNet QT database

Potential Impact

Healthcare, specifically cardiology diagnostics services and companies developing ECG analysis software, could be majorly impacted, improving diagnosis and access to historical ECG data.

Want to implement this idea in a business?

We have generated a startup concept here: HeartDigit.