Multi-model Transfer Learning and Genotypic Analysis for Seizure Type Classification

Yang, Y; Guo, K; Fang, Z; Lin, H; Grosser, M; Lu, J

Multi-model Transfer Learning and Genotypic Analysis for Seizure Type Classification

Yang, Y Guo, K Fang, Z

Lin, H Grosser, M Lu, J

Permalink

Publisher:: Springer Nature
Publication Type:: Chapter
Citation:: Health Information Science, 2023, 14305 LNCS, pp. 223-234
Issue Date:: 2023-01-01

Closed Access

	Filename	Description	Size
	Multi-model Transfer Learning and Genotypic Analysis for Seizure Type Classification.pdf	Accepted version	382.75 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Yang, Y
dc.contributor.author	Guo, K
dc.contributor.author	Fang, Z https://orcid.org/0000-0003-0602-6255
dc.contributor.author	Lin, H
dc.contributor.author	Grosser, M
dc.contributor.author	Lu, J
dc.date.accessioned	2024-02-28T12:25:57Z
dc.date.available	2024-02-28T12:25:57Z
dc.date.issued	2023-01-01
dc.identifier.citation	Health Information Science, 2023, 14305 LNCS, pp. 223-234
dc.identifier.isbn	9789819971077
dc.identifier.uri	http://hdl.handle.net/10453/175944
dc.description.abstract	The recent progress in phenotypic information and machine learning has led to a remarkable development in the accuracy of binary seizure detection. Yet the performance of classifying specific seizure types remains suboptimal due to the limited availability of annotated data with accurate seizure type labels. Transfer learning is promising to mitigate data scarcity to improve classification accuracy on smaller datasets. However, finding the best transferable model based on the specific training and testing dataset can be a complex and repetitive process, and a single-modelled approach may not fully capture the best feature representation of the input data. Moreover, genotypic data is often neglected in previous AI-based seizure detection studies, where analyses like Polygenic Risk Scores (PRS) could offer insights into genetic predispositions to seizures. To mitigate these challenges, we propose a seizure-type classification framework incorporating a multi-model weighting system designed to assign weights to different models, thus reducing computational complexity and processing time. In addition, we carry out a PRS analysis, aiming to bridge the gap between genotypic and phenotypic data, further enhancing the comprehensiveness and precision of seizure detection. Our model outperformed similar classifiers by more than 13–16% on the Temple University Hospital EEG Seizure Corpus dataset. This study represents a pioneering examination of the multi-source transfer learning framework in the field of type-specific seizure classification.
dc.language	en
dc.publisher	Springer Nature
dc.relation	http://purl.org/au-research/grants/arc/LP210100414
dc.relation.ispartof	Health Information Science
dc.relation.ispartofseries	Lecture Notes in Computer Science
dc.relation.isbasedon	10.1007/978-981-99-7108-4_19
dc.rights	info:eu-repo/semantics/closedAccess
dc.subject.classification	Artificial Intelligence & Image Processing
dc.subject.classification	46 Information and computing sciences
dc.title	Multi-model Transfer Learning and Genotypic Analysis for Seizure Type Classification
dc.type	Chapter
utslib.citation.volume	14305 LNCS
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Strength - AAII - Australian Artificial Intelligence Institute
utslib.copyright.status	closed_access	*
dc.date.updated	2024-02-28T12:25:55Z
pubs.publication-status	Published
pubs.volume	14305 LNCS

Abstract:

The recent progress in phenotypic information and machine learning has led to a remarkable development in the accuracy of binary seizure detection. Yet the performance of classifying specific seizure types remains suboptimal due to the limited availability of annotated data with accurate seizure type labels. Transfer learning is promising to mitigate data scarcity to improve classification accuracy on smaller datasets. However, finding the best transferable model based on the specific training and testing dataset can be a complex and repetitive process, and a single-modelled approach may not fully capture the best feature representation of the input data. Moreover, genotypic data is often neglected in previous AI-based seizure detection studies, where analyses like Polygenic Risk Scores (PRS) could offer insights into genetic predispositions to seizures. To mitigate these challenges, we propose a seizure-type classification framework incorporating a multi-model weighting system designed to assign weights to different models, thus reducing computational complexity and processing time. In addition, we carry out a PRS analysis, aiming to bridge the gap between genotypic and phenotypic data, further enhancing the comprehensiveness and precision of seizure detection. Our model outperformed similar classifiers by more than 13–16% on the Temple University Hospital EEG Seizure Corpus dataset. This study represents a pioneering examination of the multi-source transfer learning framework in the field of type-specific seizure classification.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/175944