Enhancing transfer performance across datasets for brain-computer interfaces using a combination of alignment strategies and adaptive batch normalization
Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in brain-computer interfaces (BCIs). However, current TL and DL algorithms are usually validated within a single dataset, assuming that data of the test subj...
Uloženo v:
| Vydáno v: | Journal of neural engineering Ročník 18; číslo 4 |
|---|---|
| Hlavní autoři: | , , , , , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
01.08.2021
|
| ISSN: | 1741-2552, 1741-2552 |
| On-line přístup: | Zjistit podrobnosti o přístupu |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Abstract | Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in brain-computer interfaces (BCIs). However, current TL and DL algorithms are usually validated within a single dataset, assuming that data of the test subjects are acquired under the same condition as that of training (source) subjects. This assumption is generally violated in practice because of different acquisition systems and experimental settings across studies and datasets. Thus, the generalization ability of these algorithms needs further validations in a cross-dataset scenario, which is closer to the actual situation. This study compared the transfer performance of pre-trained deep-learning models with different preprocessing strategies in a cross-dataset scenario.Approach. This study used four publicly available motor imagery datasets, each was successively selected as a source dataset, and the others were used as target datasets. EEGNet and ShallowConvNet with four preprocessing strategies, namely channel normalization, trial normalization, Euclidean alignment, and Riemannian alignment, were trained with the source dataset. The transfer performance of pre-trained models was validated on the target datasets. This study also used adaptive batch normalization (AdaBN) for reducing interval covariate shift across datasets. This study compared the transfer performance of using the four preprocessing strategies and that of a baseline approach based on manifold embedded knowledge transfer (MEKT). This study also explored the possibility and performance of fusing MEKT and EEGNet.Main results. The results show that DL models with alignment strategies had significantly better transfer performance than the other two preprocessing strategies. As an unsupervised domain adaptation method, AdaBN could also significantly improve the transfer performance of DL models. The transfer performance of DL models that combined AdaBN and alignment strategies significantly outperformed MEKT. Moreover, the generalizability of EEGNet models that combined AdaBN and alignment strategies could be further improved via the domain adaptation step in MEKT, achieving the best generalization ability among multiple datasets (BNCI2014001: 0.788, PhysionetMI: 0.679, Weibo2014: 0.753, Cho2017: 0.650).Significance. The combination of alignment strategies and AdaBN could easily improve the generalizability of DL models without fine-tuning. This study may provide new insights into the design of transfer neural networks for BCIs by separating source and target batch normalization layers in the domain adaptation process.Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in brain-computer interfaces (BCIs). However, current TL and DL algorithms are usually validated within a single dataset, assuming that data of the test subjects are acquired under the same condition as that of training (source) subjects. This assumption is generally violated in practice because of different acquisition systems and experimental settings across studies and datasets. Thus, the generalization ability of these algorithms needs further validations in a cross-dataset scenario, which is closer to the actual situation. This study compared the transfer performance of pre-trained deep-learning models with different preprocessing strategies in a cross-dataset scenario.Approach. This study used four publicly available motor imagery datasets, each was successively selected as a source dataset, and the others were used as target datasets. EEGNet and ShallowConvNet with four preprocessing strategies, namely channel normalization, trial normalization, Euclidean alignment, and Riemannian alignment, were trained with the source dataset. The transfer performance of pre-trained models was validated on the target datasets. This study also used adaptive batch normalization (AdaBN) for reducing interval covariate shift across datasets. This study compared the transfer performance of using the four preprocessing strategies and that of a baseline approach based on manifold embedded knowledge transfer (MEKT). This study also explored the possibility and performance of fusing MEKT and EEGNet.Main results. The results show that DL models with alignment strategies had significantly better transfer performance than the other two preprocessing strategies. As an unsupervised domain adaptation method, AdaBN could also significantly improve the transfer performance of DL models. The transfer performance of DL models that combined AdaBN and alignment strategies significantly outperformed MEKT. Moreover, the generalizability of EEGNet models that combined AdaBN and alignment strategies could be further improved via the domain adaptation step in MEKT, achieving the best generalization ability among multiple datasets (BNCI2014001: 0.788, PhysionetMI: 0.679, Weibo2014: 0.753, Cho2017: 0.650).Significance. The combination of alignment strategies and AdaBN could easily improve the generalizability of DL models without fine-tuning. This study may provide new insights into the design of transfer neural networks for BCIs by separating source and target batch normalization layers in the domain adaptation process. |
|---|---|
| AbstractList | Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in brain-computer interfaces (BCIs). However, current TL and DL algorithms are usually validated within a single dataset, assuming that data of the test subjects are acquired under the same condition as that of training (source) subjects. This assumption is generally violated in practice because of different acquisition systems and experimental settings across studies and datasets. Thus, the generalization ability of these algorithms needs further validations in a cross-dataset scenario, which is closer to the actual situation. This study compared the transfer performance of pre-trained deep-learning models with different preprocessing strategies in a cross-dataset scenario.Approach. This study used four publicly available motor imagery datasets, each was successively selected as a source dataset, and the others were used as target datasets. EEGNet and ShallowConvNet with four preprocessing strategies, namely channel normalization, trial normalization, Euclidean alignment, and Riemannian alignment, were trained with the source dataset. The transfer performance of pre-trained models was validated on the target datasets. This study also used adaptive batch normalization (AdaBN) for reducing interval covariate shift across datasets. This study compared the transfer performance of using the four preprocessing strategies and that of a baseline approach based on manifold embedded knowledge transfer (MEKT). This study also explored the possibility and performance of fusing MEKT and EEGNet.Main results. The results show that DL models with alignment strategies had significantly better transfer performance than the other two preprocessing strategies. As an unsupervised domain adaptation method, AdaBN could also significantly improve the transfer performance of DL models. The transfer performance of DL models that combined AdaBN and alignment strategies significantly outperformed MEKT. Moreover, the generalizability of EEGNet models that combined AdaBN and alignment strategies could be further improved via the domain adaptation step in MEKT, achieving the best generalization ability among multiple datasets (BNCI2014001: 0.788, PhysionetMI: 0.679, Weibo2014: 0.753, Cho2017: 0.650).Significance. The combination of alignment strategies and AdaBN could easily improve the generalizability of DL models without fine-tuning. This study may provide new insights into the design of transfer neural networks for BCIs by separating source and target batch normalization layers in the domain adaptation process.Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in brain-computer interfaces (BCIs). However, current TL and DL algorithms are usually validated within a single dataset, assuming that data of the test subjects are acquired under the same condition as that of training (source) subjects. This assumption is generally violated in practice because of different acquisition systems and experimental settings across studies and datasets. Thus, the generalization ability of these algorithms needs further validations in a cross-dataset scenario, which is closer to the actual situation. This study compared the transfer performance of pre-trained deep-learning models with different preprocessing strategies in a cross-dataset scenario.Approach. This study used four publicly available motor imagery datasets, each was successively selected as a source dataset, and the others were used as target datasets. EEGNet and ShallowConvNet with four preprocessing strategies, namely channel normalization, trial normalization, Euclidean alignment, and Riemannian alignment, were trained with the source dataset. The transfer performance of pre-trained models was validated on the target datasets. This study also used adaptive batch normalization (AdaBN) for reducing interval covariate shift across datasets. This study compared the transfer performance of using the four preprocessing strategies and that of a baseline approach based on manifold embedded knowledge transfer (MEKT). This study also explored the possibility and performance of fusing MEKT and EEGNet.Main results. The results show that DL models with alignment strategies had significantly better transfer performance than the other two preprocessing strategies. As an unsupervised domain adaptation method, AdaBN could also significantly improve the transfer performance of DL models. The transfer performance of DL models that combined AdaBN and alignment strategies significantly outperformed MEKT. Moreover, the generalizability of EEGNet models that combined AdaBN and alignment strategies could be further improved via the domain adaptation step in MEKT, achieving the best generalization ability among multiple datasets (BNCI2014001: 0.788, PhysionetMI: 0.679, Weibo2014: 0.753, Cho2017: 0.650).Significance. The combination of alignment strategies and AdaBN could easily improve the generalizability of DL models without fine-tuning. This study may provide new insights into the design of transfer neural networks for BCIs by separating source and target batch normalization layers in the domain adaptation process. |
| Author | Xu, Minpeng Ming, Dong Jung, Tzyy-Ping Wang, Kun Xu, Lichao Ma, Zhen |
| Author_xml | – sequence: 1 givenname: Lichao surname: Xu fullname: Xu, Lichao – sequence: 2 givenname: Minpeng surname: Xu fullname: Xu, Minpeng – sequence: 3 givenname: Zhen surname: Ma fullname: Ma, Zhen – sequence: 4 givenname: Kun surname: Wang fullname: Wang, Kun – sequence: 5 givenname: Tzyy-Ping surname: Jung fullname: Jung, Tzyy-Ping – sequence: 6 givenname: Dong surname: Ming fullname: Ming, Dong |
| BookMark | eNpNTTtPwzAQtlCRaAs7o0eWQGw3TjKiqjykSiwwVxf73Bol5xA7DPwT_i0pIMRy3-l7LtiMAiFjlyK_FnlV3YhyJTJZFPIGjEArT9j8j5r9-8_YIsbXPFeirPM5-9zQAch42vM0AEWHA-9xcGHoJho5mCHEyC0kiJginwTeDOApM6HrxzTZPU3XgcHIx3gsAj5pjSdIPhAPjkPr99QhJR6nkYR7P3mBLAcLffLvyBtI5sDpuNr6j-_gOTt10Ea8-MUle7nbPK8fsu3T_eP6dpsZJVTKTClkqSWWTSlVLbTUoAoHWhVW1s6uxMq4phLWaSO1sLmpCqhsU5gKyxpcI5fs6qe3H8LbiDHtOh8Nti0QhjHuZKFlpUSdK_kFf6JyhQ |
| CitedBy_id | crossref_primary_10_1109_TNSRE_2022_3207494 crossref_primary_10_1088_1741_2552_ad0a01 crossref_primary_10_3390_brainsci13020268 crossref_primary_10_3389_fpsyg_2022_839440 crossref_primary_10_1109_TNSRE_2022_3150007 crossref_primary_10_1109_JSEN_2025_3528009 crossref_primary_10_1109_TIM_2022_3181276 crossref_primary_10_1016_j_compbiomed_2023_107806 crossref_primary_10_1088_1741_2552_addd49 crossref_primary_10_1088_1741_2552_ac857d crossref_primary_10_1109_TCSS_2024_3462823 crossref_primary_10_1155_2022_1603104 crossref_primary_10_1186_s12938_023_01129_4 crossref_primary_10_3390_s23052750 crossref_primary_10_1016_j_neucom_2024_128577 crossref_primary_10_3390_app12031695 crossref_primary_10_1016_j_jneumeth_2022_109535 crossref_primary_10_1088_1741_2552_ac8dc5 crossref_primary_10_1088_1741_2552_acfe9c crossref_primary_10_3389_fncom_2022_909553 crossref_primary_10_1016_j_neunet_2025_107511 |
| ContentType | Journal Article |
| Copyright | 2021 IOP Publishing Ltd. |
| Copyright_xml | – notice: 2021 IOP Publishing Ltd. |
| DBID | 7X8 |
| DOI | 10.1088/1741-2552/ac1ed2 |
| DatabaseName | MEDLINE - Academic |
| DatabaseTitle | MEDLINE - Academic |
| DatabaseTitleList | MEDLINE - Academic |
| Database_xml | – sequence: 1 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | no_fulltext_linktorsrc |
| Discipline | Anatomy & Physiology |
| EISSN | 1741-2552 |
| GroupedDBID | --- 1JI 4.4 53G 5B3 5GY 5VS 5ZH 7.M 7.Q 7X8 AAGCD AAJIO AAJKP AATNI ABHWH ABJNI ABQJV ABVAM ACAFW ACGFS ACHIP ADEQX AEFHF AEINN AENEX AFYNE AKPSB ALMA_UNASSIGNED_HOLDINGS AOAED ASPBG ATQHT AVWKF AZFZN CEBXE CJUJL CRLBU CS3 DU5 EBS EDWGO EMSAF EPQRW EQZZN F5P IHE IJHAN IOP IZVLO KOT LAP M45 N5L N9A P2P PJBAE RIN RO9 ROL RPA SY9 W28 XPP |
| ID | FETCH-LOGICAL-c313t-c712762e7b72391626a35fa635d29fd414cfb81df6c261d0c85a8db5c8e79afb2 |
| IEDL.DBID | 7X8 |
| ISICitedReferencesCount | 23 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000692006100001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 1741-2552 |
| IngestDate | Sun Aug 24 03:02:59 EDT 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Issue | 4 |
| Language | English |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c313t-c712762e7b72391626a35fa635d29fd414cfb81df6c261d0c85a8db5c8e79afb2 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
| PQID | 2562831903 |
| PQPubID | 23479 |
| ParticipantIDs | proquest_miscellaneous_2562831903 |
| PublicationCentury | 2000 |
| PublicationDate | 2021-08-01 |
| PublicationDateYYYYMMDD | 2021-08-01 |
| PublicationDate_xml | – month: 08 year: 2021 text: 2021-08-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationTitle | Journal of neural engineering |
| PublicationYear | 2021 |
| SSID | ssj0031790 |
| Score | 2.4646852 |
| Snippet | Objective. Recently, transfer learning (TL) and deep learning (DL) have been introduced to solve intra- and inter-subject variability problems in... |
| SourceID | proquest |
| SourceType | Aggregation Database |
| Title | Enhancing transfer performance across datasets for brain-computer interfaces using a combination of alignment strategies and adaptive batch normalization |
| URI | https://www.proquest.com/docview/2562831903 |
| Volume | 18 |
| WOSCitedRecordID | wos000692006100001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1LSwMxEA5qPXjxVcU3I4i30N3sK3uSIi0epHhQ6a1k82gFzdbuKvhT_LdO0m178CJ43SzsI8l832S-mSHkSiVcszQRlAmODkouGEWWwamMA52lRSRiX774-T4bDPhwmD80B25VI6tc2ERvqFUp3Rl5B6EZkRDhK7qZvlPXNcpFV5sWGuukFSGVcZKubLiMIkSu-tQ8ITKkSJ1ZE6bEjdVZXusIGWrFfplijy_9nf--2S7ZbpgldOdLYY-sabtP2l2LXvXbF1yD13r6Q_Q2-e7ZiSu0YcdQe-aqZzBdpRCA8OAJTj5a6boCHIDCNZOgsukCAa7OxMw4QRc47fwYBOAY-tl-qqE0gAx_7LUGUNWLghQgrAKhxNRZWSgQByZg3VNfm3zQA_LU7z3e3tGmSQOV-MtrKrOQoUHVWZExl8TLUhElRiCPUSw3Kg5jaQokxSaV6KypQPJEcFUkkussF6Zgh2TDllYfEfy2SAU6D3DMxCLhuUT6wk0RcgTyKAuOyeViAka4CVxkQ1hdflSj1RSc_OGeU7LFnDLFy_jOSMvgRtfnZFN-1i_V7MKvoR_i8tWI |
| linkProvider | ProQuest |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Enhancing+transfer+performance+across+datasets+for+brain-computer+interfaces+using+a+combination+of+alignment+strategies+and+adaptive+batch+normalization&rft.jtitle=Journal+of+neural+engineering&rft.au=Xu%2C+Lichao&rft.au=Xu%2C+Minpeng&rft.au=Ma%2C+Zhen&rft.au=Wang%2C+Kun&rft.date=2021-08-01&rft.issn=1741-2552&rft.eissn=1741-2552&rft.volume=18&rft.issue=4&rft_id=info:doi/10.1088%2F1741-2552%2Fac1ed2&rft.externalDBID=NO_FULL_TEXT |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1741-2552&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1741-2552&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1741-2552&client=summon |