Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events
Adverse events after surgery not only affect the patient’s recovery but also increase the burden on doctors and patients due to prolonged hospitalization. Predicting adverse events from patient data before surgery with a machine learning method is highly expected. It is difficult to colle...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2023-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/10124969/ |
_version_ | 1797820038566117376 |
---|---|
author | Yuki Yamasaki Chiaki Doi Shiori Kitagawa Hiroyuki Seki Hiroshi Shigeno |
author_facet | Yuki Yamasaki Chiaki Doi Shiori Kitagawa Hiroyuki Seki Hiroshi Shigeno |
author_sort | Yuki Yamasaki |
collection | DOAJ |
description | Adverse events after surgery not only affect the patient’s recovery but also increase the burden on doctors and patients due to prolonged hospitalization. Predicting adverse events from patient data before surgery with a machine learning method is highly expected. It is difficult to collect a large amount of patient data since the number of surgeries in a year is limited and predict the occurrence of adverse events accurately since patient data are imbalanced data. To improve the accuracy of adverse event prediction, this paper proposes data generation with Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE for the preoperative prediction of adverse events. Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE has filters by the reconstruction error and by a machine learning method. After <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE generates minority class data, the two layers of filtering are used to remove low-quality minority class data that have little contribution to the adverse event prediction. In the evaluations, patient data obtained from Tokyo Dental University Ichikawa General Hospital were used. The proposed method can predict adverse events with a recall of 0.848, which is 5.6% more accurate than existing methods. The effects of filtering in Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE are visualized, and the reasons for the improvement in prediction accuracy are clarified. Furthermore, this paper shows that <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE can generate arbitrary patient data even in table data, corresponding to the distribution of the original patient data. |
first_indexed | 2024-03-13T09:32:27Z |
format | Article |
id | doaj.art-7d56c990d8814218bc30767fed027364 |
institution | Directory Open Access Journal |
issn | 2169-3536 |
language | English |
last_indexed | 2024-03-13T09:32:27Z |
publishDate | 2023-01-01 |
publisher | IEEE |
record_format | Article |
series | IEEE Access |
spelling | doaj.art-7d56c990d8814218bc30767fed0273642023-05-25T23:00:29ZengIEEEIEEE Access2169-35362023-01-0111486674867610.1109/ACCESS.2023.327678310124969Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse EventsYuki Yamasaki0https://orcid.org/0000-0002-7046-3049Chiaki Doi1Shiori Kitagawa2Hiroyuki Seki3https://orcid.org/0000-0001-5311-5088Hiroshi Shigeno4https://orcid.org/0000-0001-6320-8835Department of Information and Computer Science, Keio University, Kanagawa, JapanDepartment of Information and Computer Science, Keio University, Kanagawa, JapanDepartment of Information and Computer Science, Keio University, Kanagawa, JapanDepartment of Anesthesiology, Tokyo Dental College Ichikawa General Hospital, Chiba, JapanDepartment of Information and Computer Science, Keio University, Kanagawa, JapanAdverse events after surgery not only affect the patient’s recovery but also increase the burden on doctors and patients due to prolonged hospitalization. Predicting adverse events from patient data before surgery with a machine learning method is highly expected. It is difficult to collect a large amount of patient data since the number of surgeries in a year is limited and predict the occurrence of adverse events accurately since patient data are imbalanced data. To improve the accuracy of adverse event prediction, this paper proposes data generation with Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE for the preoperative prediction of adverse events. Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE has filters by the reconstruction error and by a machine learning method. After <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE generates minority class data, the two layers of filtering are used to remove low-quality minority class data that have little contribution to the adverse event prediction. In the evaluations, patient data obtained from Tokyo Dental University Ichikawa General Hospital were used. The proposed method can predict adverse events with a recall of 0.848, which is 5.6% more accurate than existing methods. The effects of filtering in Filtered <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE are visualized, and the reasons for the improvement in prediction accuracy are clarified. Furthermore, this paper shows that <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-VAE can generate arbitrary patient data even in table data, corresponding to the distribution of the original patient data.https://ieeexplore.ieee.org/document/10124969/Machine learningdata generationmedicaladverse events |
spellingShingle | Yuki Yamasaki Chiaki Doi Shiori Kitagawa Hiroyuki Seki Hiroshi Shigeno Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events IEEE Access Machine learning data generation medical adverse events |
title | Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events |
title_full | Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events |
title_fullStr | Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events |
title_full_unstemmed | Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events |
title_short | Data Generation With Filtered <italic>β</italic>-VAE for the Preoperative Prediction of Adverse Events |
title_sort | data generation with filtered italic x03b2 italic vae for the preoperative prediction of adverse events |
topic | Machine learning data generation medical adverse events |
url | https://ieeexplore.ieee.org/document/10124969/ |
work_keys_str_mv | AT yukiyamasaki datagenerationwithfiltereditalicx03b2italicvaeforthepreoperativepredictionofadverseevents AT chiakidoi datagenerationwithfiltereditalicx03b2italicvaeforthepreoperativepredictionofadverseevents AT shiorikitagawa datagenerationwithfiltereditalicx03b2italicvaeforthepreoperativepredictionofadverseevents AT hiroyukiseki datagenerationwithfiltereditalicx03b2italicvaeforthepreoperativepredictionofadverseevents AT hiroshishigeno datagenerationwithfiltereditalicx03b2italicvaeforthepreoperativepredictionofadverseevents |