{"title":"用于跨数据集运动图像脑电图转移学习的多源深度域自适应集合框架。","authors":"Minmin Miao, Zhong Yang, Zhenzhen Sheng, Baoguo Xu, Wenbin Zhang, Xinmin Cheng","doi":"10.1088/1361-6579/ad4e95","DOIUrl":null,"url":null,"abstract":"<p><p><i>Objective</i>. Electroencephalography (EEG) is an important kind of bioelectric signal for measuring physiological activities of the brain, and motor imagery (MI) EEG has significant clinical application prospects. Convolutional neural network has become a mainstream algorithm for MI EEG classification, however lack of subject-specific data considerably restricts its decoding accuracy and generalization performance. To address this challenge, a novel transfer learning (TL) framework using auxiliary dataset to improve the MI EEG classification performance of target subject is proposed in this paper.<i>Approach</i>. We developed a multi-source deep domain adaptation ensemble framework (MSDDAEF) for cross-dataset MI EEG decoding. The proposed MSDDAEF comprises three main components: model pre-training, deep domain adaptation, and multi-source ensemble. Moreover, for each component, different designs were examined to verify the robustness of MSDDAEF.<i>Main results</i>. Bidirectional validation experiments were performed on two large public MI EEG datasets (openBMI and GIST). The highest average classification accuracy of MSDDAEF reaches 74.28% when openBMI serves as target dataset and GIST serves as source dataset. While the highest average classification accuracy of MSDDAEF is 69.85% when GIST serves as target dataset and openBMI serves as source dataset. In addition, the classification performance of MSDDAEF surpasses several well-established studies and state-of-the-art algorithms.<i>Significance</i>. The results of this study show that cross-dataset TL is feasible for left/right-hand MI EEG decoding, and further indicate that MSDDAEF is a promising solution for addressing MI EEG cross-dataset variability.</p>","PeriodicalId":20047,"journal":{"name":"Physiological measurement","volume":" ","pages":""},"PeriodicalIF":2.3000,"publicationDate":"2024-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Multi-source deep domain adaptation ensemble framework for cross-dataset motor imagery EEG transfer learning.\",\"authors\":\"Minmin Miao, Zhong Yang, Zhenzhen Sheng, Baoguo Xu, Wenbin Zhang, Xinmin Cheng\",\"doi\":\"10.1088/1361-6579/ad4e95\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p><i>Objective</i>. Electroencephalography (EEG) is an important kind of bioelectric signal for measuring physiological activities of the brain, and motor imagery (MI) EEG has significant clinical application prospects. Convolutional neural network has become a mainstream algorithm for MI EEG classification, however lack of subject-specific data considerably restricts its decoding accuracy and generalization performance. To address this challenge, a novel transfer learning (TL) framework using auxiliary dataset to improve the MI EEG classification performance of target subject is proposed in this paper.<i>Approach</i>. We developed a multi-source deep domain adaptation ensemble framework (MSDDAEF) for cross-dataset MI EEG decoding. The proposed MSDDAEF comprises three main components: model pre-training, deep domain adaptation, and multi-source ensemble. Moreover, for each component, different designs were examined to verify the robustness of MSDDAEF.<i>Main results</i>. Bidirectional validation experiments were performed on two large public MI EEG datasets (openBMI and GIST). The highest average classification accuracy of MSDDAEF reaches 74.28% when openBMI serves as target dataset and GIST serves as source dataset. While the highest average classification accuracy of MSDDAEF is 69.85% when GIST serves as target dataset and openBMI serves as source dataset. In addition, the classification performance of MSDDAEF surpasses several well-established studies and state-of-the-art algorithms.<i>Significance</i>. The results of this study show that cross-dataset TL is feasible for left/right-hand MI EEG decoding, and further indicate that MSDDAEF is a promising solution for addressing MI EEG cross-dataset variability.</p>\",\"PeriodicalId\":20047,\"journal\":{\"name\":\"Physiological measurement\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":2.3000,\"publicationDate\":\"2024-06-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Physiological measurement\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://doi.org/10.1088/1361-6579/ad4e95\",\"RegionNum\":4,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"BIOPHYSICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Physiological measurement","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1088/1361-6579/ad4e95","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"BIOPHYSICS","Score":null,"Total":0}
引用次数: 0
摘要
目的:脑电图(EEG)是测量大脑生理活动的一种重要的生物电信号,运动意象脑电图(MI)具有重要的临床应用前景。卷积神经网络(CNN)已成为运动图像脑电图分类的主流算法,但由于缺乏特定对象的数据,其解码精度和泛化性能受到很大限制。为应对这一挑战,本文提出了一种新型迁移学习(TL)框架,利用辅助数据集提高目标受试者的 MI EEG 分类性能:方法:我们为跨数据集 MI 脑电图解码开发了多源深度域自适应集合框架(MSDDAEF)。所提出的 MSDDAEF 由三个主要部分组成:模型预训练、深域适配和多源集合。此外,还对每个组件进行了不同的设计,以验证 MSDDAEF 的鲁棒性:在两个大型公共 MI EEG 数据集(openBMI 和 GIST)上进行了双向验证实验。当 openBMI 作为目标数据集,GIST 作为源数据集时,MSDDAEF 的最高平均分类准确率达到 74.28%。而当 GIST 作为目标数据集,openBMI 作为源数据集时,MSDDAEF 的最高平均分类准确率为 69.85%。此外,MSDDAEF 的分类性能还超过了几项成熟的研究和最先进的算法:本研究的结果表明,跨数据集 TL 对左/右手 MI 脑电图解码是可行的,并进一步表明 MSDDAEF 是解决 MI 脑电图跨数据集变异性的一种有前途的解决方案。
Multi-source deep domain adaptation ensemble framework for cross-dataset motor imagery EEG transfer learning.
Objective. Electroencephalography (EEG) is an important kind of bioelectric signal for measuring physiological activities of the brain, and motor imagery (MI) EEG has significant clinical application prospects. Convolutional neural network has become a mainstream algorithm for MI EEG classification, however lack of subject-specific data considerably restricts its decoding accuracy and generalization performance. To address this challenge, a novel transfer learning (TL) framework using auxiliary dataset to improve the MI EEG classification performance of target subject is proposed in this paper.Approach. We developed a multi-source deep domain adaptation ensemble framework (MSDDAEF) for cross-dataset MI EEG decoding. The proposed MSDDAEF comprises three main components: model pre-training, deep domain adaptation, and multi-source ensemble. Moreover, for each component, different designs were examined to verify the robustness of MSDDAEF.Main results. Bidirectional validation experiments were performed on two large public MI EEG datasets (openBMI and GIST). The highest average classification accuracy of MSDDAEF reaches 74.28% when openBMI serves as target dataset and GIST serves as source dataset. While the highest average classification accuracy of MSDDAEF is 69.85% when GIST serves as target dataset and openBMI serves as source dataset. In addition, the classification performance of MSDDAEF surpasses several well-established studies and state-of-the-art algorithms.Significance. The results of this study show that cross-dataset TL is feasible for left/right-hand MI EEG decoding, and further indicate that MSDDAEF is a promising solution for addressing MI EEG cross-dataset variability.
期刊介绍:
Physiological Measurement publishes papers about the quantitative assessment and visualization of physiological function in clinical research and practice, with an emphasis on the development of new methods of measurement and their validation.
Papers are published on topics including:
applied physiology in illness and health
electrical bioimpedance, optical and acoustic measurement techniques
advanced methods of time series and other data analysis
biomedical and clinical engineering
in-patient and ambulatory monitoring
point-of-care technologies
novel clinical measurements of cardiovascular, neurological, and musculoskeletal systems.
measurements in molecular, cellular and organ physiology and electrophysiology
physiological modeling and simulation
novel biomedical sensors, instruments, devices and systems
measurement standards and guidelines.