Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies.
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 EC...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2024-01-01
|
| Series: | PLoS ONE |
| Online Access: | https://doi.org/10.1371/journal.pone.0307978 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846098985398304768 |
|---|---|
| author | Shinnosuke Sawano Satoshi Kodera Naoto Setoguchi Kengo Tanabe Shunichi Kushida Junji Kanda Mike Saji Mamoru Nanasato Hisataka Maki Hideo Fujita Nahoko Kato Hiroyuki Watanabe Minami Suzuki Masao Takahashi Naoko Sawada Masao Yamasaki Masataka Sato Susumu Katsushika Hiroki Shinohara Norifumi Takeda Katsuhito Fujiu Masao Daimon Hiroshi Akazawa Hiroyuki Morita Issei Komuro |
| author_facet | Shinnosuke Sawano Satoshi Kodera Naoto Setoguchi Kengo Tanabe Shunichi Kushida Junji Kanda Mike Saji Mamoru Nanasato Hisataka Maki Hideo Fujita Nahoko Kato Hiroyuki Watanabe Minami Suzuki Masao Takahashi Naoko Sawada Masao Yamasaki Masataka Sato Susumu Katsushika Hiroki Shinohara Norifumi Takeda Katsuhito Fujiu Masao Daimon Hiroshi Akazawa Hiroyuki Morita Issei Komuro |
| author_sort | Shinnosuke Sawano |
| collection | DOAJ |
| description | 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. |
| format | Article |
| id | doaj-art-204ccd9b977e45ee8257d4026be4bf06 |
| institution | Kabale University |
| issn | 1932-6203 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | Public Library of Science (PLoS) |
| record_format | Article |
| series | PLoS ONE |
| spelling | doaj-art-204ccd9b977e45ee8257d4026be4bf062025-01-01T05:31:07ZengPublic Library of Science (PLoS)PLoS ONE1932-62032024-01-01198e030797810.1371/journal.pone.0307978Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies.Shinnosuke SawanoSatoshi KoderaNaoto SetoguchiKengo TanabeShunichi KushidaJunji KandaMike SajiMamoru NanasatoHisataka MakiHideo FujitaNahoko KatoHiroyuki WatanabeMinami SuzukiMasao TakahashiNaoko SawadaMasao YamasakiMasataka SatoSusumu KatsushikaHiroki ShinoharaNorifumi TakedaKatsuhito FujiuMasao DaimonHiroshi AkazawaHiroyuki MoritaIssei KomuroThe 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.https://doi.org/10.1371/journal.pone.0307978 |
| spellingShingle | Shinnosuke Sawano Satoshi Kodera Naoto Setoguchi Kengo Tanabe Shunichi Kushida Junji Kanda Mike Saji Mamoru Nanasato Hisataka Maki Hideo Fujita Nahoko Kato Hiroyuki Watanabe Minami Suzuki Masao Takahashi Naoko Sawada Masao Yamasaki Masataka Sato Susumu Katsushika Hiroki Shinohara Norifumi Takeda Katsuhito Fujiu Masao Daimon Hiroshi Akazawa Hiroyuki Morita Issei Komuro Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. PLoS ONE |
| title | Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. |
| title_full | Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. |
| title_fullStr | Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. |
| title_full_unstemmed | Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. |
| title_short | Applying masked autoencoder-based self-supervised learning for high-capability vision transformers of electrocardiographies. |
| title_sort | applying masked autoencoder based self supervised learning for high capability vision transformers of electrocardiographies |
| url | https://doi.org/10.1371/journal.pone.0307978 |
| work_keys_str_mv | AT shinnosukesawano applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT satoshikodera applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT naotosetoguchi applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT kengotanabe applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT shunichikushida applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT junjikanda applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT mikesaji applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT mamorunanasato applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hisatakamaki applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hideofujita applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT nahokokato applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hiroyukiwatanabe applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT minamisuzuki applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT masaotakahashi applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT naokosawada applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT masaoyamasaki applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT masatakasato applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT susumukatsushika applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hirokishinohara applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT norifumitakeda applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT katsuhitofujiu applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT masaodaimon applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hiroshiakazawa applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT hiroyukimorita applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies AT isseikomuro applyingmaskedautoencoderbasedselfsupervisedlearningforhighcapabilityvisiontransformersofelectrocardiographies |