WPI - Computer Science, MS Thesis Presentation, Jiaqi Ji " A Deep Semantic Segmentation Framework for Digitizing Paper Electrocardiograms with Data Augmentation using a Diffusion Model"
WPI – Computer Science Department
Monday, April 24, 2023
Time: 10:00 a.m. – 11:00 a.m.
Location: Fuller Labs 311
Advisor: Prof. Emmanuel Agu
Reader: Prof. Dmitry Korkin
Electrocardiograms (ECGs) play a key role in diagnosis of cardiac diseases. ECGs have traditionally been analyzed manually by human experts, which is time consuming and yields inconsistent results. In electronic form, ECGs can be analyzed using modern methods such as machine learning. However, a large number of ECGs are currently stored in paper forms, especially in developing countries.
Moreover, paper ECGs are easily damaged, degenerate easily, and are difficult to preserve. In this paper, the problem of digitizing ECGs to facilitate Machine Learning (ML) analyses is explored. Due to the wide variety of colors and signal variations in real ECGs, digitization using traditional image processing techniques for ECG digitization is challenging. We propose a neural networks approach that uses the Unet segmentation to identify regions of medical paper containing the ECG signal.
To facilitate deep learning approaches, first we augmented the small ECG dataset by using a diffusion model to generate synthetic ECG signals. In rigorous evaluation, our best-performing model using the VGG16 model, achieved F1 score of 92.52%, and mIoU of 86.88%, outperforming state of the art baseline models.