Diffusion Models and Masked Training Helps Digitization and Classification of Multiple Layout ECG Images

Recent advancements have led to the development of algorithms for interpreting ECG time series.However, the continued prevalence of ECG images highlights the urgent need for digitization and affordable data analysis.This is essential to ensure comprehensive cardiac care and to capture the diverse manifestations of cardiovascular diseases worldwide.As part of the George B. Moody Phy-sioNet Challenge 2024, our goal is to propose a series of pragmatic components for the digitization and classification of ECG images.First, we generate training samples with the ECG-Image-Kit and refine them using diffusion models for data augmentation.Then, we employ a U-net architecture to perform ECG digitization, utilizing this large scale ECG images paired with their corresponding ground-truth time series.Next, we pre-train a RegNet model for ECG classification using a large scale ECG time series data from open-source datasets.This pre-trained classifier is then further fine-tuned with the digitized ECG time series derived from ECG images.Additionally, we devise an adaptable meta-model and a masked training strategy to address issues related to varying lengths and asynchronization when digitizing diverse ECG image layouts.Our team, PKU NIHDS, reported a SNR of -1.103 on the reconstruction task and an F-measure of 0.421 on the classification task for the hidden test set.