Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies
PLOS ONE · 2024
DOI: 10.1371/journal.pone.0307978 ↗Kurz gefasst
Multizentrische Studie in der Zeitschrift PLOS ONE, 2024. Verfasst von Shinnosuke Sawano und 24 weiteren Autor:innen. Beteiligte Einrichtung: Department of Cardiovascular Medicine, The University of Tokyo Hospital, Tokyo, Japan. Hauptthemen laut MEDLINE: Electrocardiography, Neural Networks, Computer. Die Arbeit wurde bisher 19-mal zitiert (Crossref) und nennt 35 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Japan Agency for Medical Research and Development.
Abstract (Originalsprache)
The generalization of deep neural network algorithms to a broader population is an important challenge in the medical field. We aimed to apply self-supervised learning using masked autoencoders (MAEs) to improve the performance of the 12-lead electrocardiography (ECG) analysis model using limited ECG data. We pretrained Vision Transformer (ViT) models by reconstructing the masked ECG data with MAE. We fine-tuned this MAE-based ECG pretrained model on ECG-echocardiography data from The University of Tokyo Hospital (UTokyo) for the detection of left ventricular systolic dysfunction (LVSD), and then evaluated it using multi-center external validation data from seven institutions, employing the area under the receiver operating characteristic curve (AUROC) for assessment. We included 38,245 ECG-echocardiography pairs from UTokyo and 229,439 pairs from all institutions. The performances of MAE-based ECG models pretrained using ECG data from UTokyo were significantly higher than that of other Deep Neural Network models across all external validation cohorts (AUROC, 0.913–0.962 for LVSD, p < 0.001). Moreover, we also found improvements for the MAE-based ECG analysis model depending on the model capacity and the amount of training data. Additionally, the MAE-based ECG analysis model maintained high performance even on the ECG benchmark dataset (PTB-XL). Our proposed method developed high performance MAE-based ECG analysis models using limited ECG data.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Department of Cardiovascular Medicine, The University of Tokyo Hospital, Tokyo, Japan.
- Division of Cardiology, Mitsui Memorial Hospital, Tokyo, Japan.
- Department of Cardiovascular Medicine, Asahi General Hospital, Chiba, Japan.
- Department of Cardiology, Sakakibara Heart Institute, Tokyo, Japan.
- Division of Cardiovascular Medicine, Saitama Medical Center, Jichi Medical University, Omiya, Japan.
- Department of Cardiology, Tokyo Bay Medical Center, Urayasu, Japan.
ORCID-Profile: Shinnosuke Sawano, Satoshi Kodera, Norifumi Takeda, Masao Daimon
Förderung
- Japan Agency for Medical Research and Development
Zitieren
APA
Sawano, S., Kodera, S., Setoguchi, N., Tanabe, K., Kushida, S., Kanda, J., Saji, M., Nanasato, M., Maki, H., Fujita, H., Kato, N., Watanabe, H., Suzuki, M., Takahashi, M., Sawada, N., Yamasaki, M., Sato, M., Katsushika, S., Shinohara, H., …, & Komuro, I. (2024). Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. PLOS ONE. https://doi.org/10.1371/journal.pone.0307978
Vancouver
Sawano S, Kodera S, Setoguchi N, Tanabe K, Kushida S, Kanda J, et al. Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. PLOS ONE. 2024. doi:10.1371/journal.pone.0307978
BibTeX
@article{Sawano2024,
author = {Sawano, Shinnosuke and Kodera, Satoshi and Setoguchi, Naoto and Tanabe, Kengo and Kushida, Shunichi and Kanda, Junji and Saji, Mike and Nanasato, Mamoru and Maki, Hisataka and Fujita, Hideo and Kato, Nahoko and Watanabe, Hiroyuki and Suzuki, Minami and Takahashi, Masao and Sawada, Naoko and Yamasaki, Masao and Sato, Masataka and Katsushika, Susumu and Shinohara, Hiroki and Takeda, Norifumi and Fujiu, Katsuhito and Daimon, Masao and Akazawa, Hiroshi and Morita, Hiroyuki and Komuro, Issei},
title = {Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies},
journal = {PLOS ONE},
year = {2024},
doi = {10.1371/journal.pone.0307978}
}
EUVIMED