Comparing Deep Neural Network for Multi-Label ECG Diagnosis From Scanned ECG

TL;DR

This paper tackles the practical problem of multi-label ECG diagnosis from scanned paper records and compares AlexNet, VGG, ResNet, and Vision Transformer on scanned ECGs. It investigates an indirect digitization pathway (VinDigitizer) to reconstruct digital 12-lead signals from paper images, and reports ongoing work on direct image-based classification. The digitization results show a mean $SNR$ of $0.01$ $dB$, highlighting challenges from printing and scanning artifacts, while model robustness to these artifacts varies. The work offers insights into the feasibility, limitations, and necessary preprocessing or domain adaptation steps for integrating image-based ECG diagnosis into clinical workflows and telemedicine.

Abstract

Automated ECG diagnosis has seen significant advancements with deep learning techniques, but real-world applications still face challenges when dealing with scanned paper ECGs. In this study, we explore multi-label classification of ECGs extracted from scanned images, moving beyond traditional binary classification (normal/abnormal). We evaluate the performance of multiple deep neural network architectures, including AlexNet, VGG, ResNet, and Vision Transformer, on scanned ECG datasets. Our comparative analysis examines model accuracy, robustness to image artifacts, and generalizability across different ECG conditions. Additionally, we investigate whether ECG signals extracted from scanned images retain sufficient diagnostic information for reliable automated classification. The findings highlight the strengths and limitations of each architecture, providing insights into the feasibility of image-based ECG diagnosis and its potential integration into clinical workflows.

Figures (3)

• Figure 1: Sample scanned paper ECG
• Figure 2: Signal extraction procedure.
• Figure 3: Digitization result of Lead II, patient 001, ID s0010_re.