Text-to-ECG: 12-Lead Electrocardiogram Synthesis conditioned on Clinical Text Reports

Hyunseung ChungJiho KimJoon-myoung KwonKi-Hyun JeonMin Sung LeeEdward Choi

Mar 2023

摘要

Electrocardiogram (ECG) synthesis is the area of research focused ongenerating realistic synthetic ECG signals for medical use without concernsover annotation costs or clinical data privacy restrictions. Traditional ECGgeneration models consider a single ECG lead and utilize GAN-based generativemodels. These models can only generate single lead samples and require separatetraining for each diagnosis class. The diagnosis classes of ECGs areinsufficient to capture the intricate differences between ECGs depending onvarious features (e.g. patient demographic details, co-existing diagnosisclasses, etc.). To alleviate these challenges, we present a text-to-ECG task,in which textual inputs are used to produce ECG outputs. Then we proposeAuto-TTE, an autoregressive generative model conditioned on clinical textreports to synthesize 12-lead ECGs, for the first time to our knowledge. Wecompare the performance of our model with other representative models intext-to-speech and text-to-image. Experimental results show the superiority ofour model in various quantitative evaluations and qualitative analysis.Finally, we conduct a user study with three board-certified cardiologists toconfirm the fidelity and semantic alignment of generated samples. our code willbe available at https://github.com/TClife/text_to_ecg

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答

被引用

笔记

问答

领域