DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals
Objective: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects a significant percentage of the elderly. EEG has emerged as a promising tool for the timely diagnosis and classification of AD or other dementia types. This paper proposes a novel approach to AD E...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10179900/ |
| _version_ | 1827895877846433792 |
|---|---|
| author | Andreas Miltiadous Emmanouil Gionanidis Katerina D. Tzimourta Nikolaos Giannakeas Alexandros T. Tzallas |
| author_facet | Andreas Miltiadous Emmanouil Gionanidis Katerina D. Tzimourta Nikolaos Giannakeas Alexandros T. Tzallas |
| author_sort | Andreas Miltiadous |
| collection | DOAJ |
| description | Objective: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects a significant percentage of the elderly. EEG has emerged as a promising tool for the timely diagnosis and classification of AD or other dementia types. This paper proposes a novel approach to AD EEG classification using a Dual-Input Convolution Encoder Network (DICE-net). Approach: Recordings of 36 AD, 23 Frontotemporal dementia (FTD), and 29 age-matched healthy individuals (CN) were used. After denoising, Band power and Coherence features were extracted and fed to DICE-net, which consists of Convolution, Transformer Encoder, and Feed-Forward layers. Main results: Our results show that DICE-net achieved an accuracy of 83.28% in the AD-CN problem using Leave-One-Subject-Out validation, outperforming several baseline models, and achieving good generalization performance. Significance: Our findings suggest that a convolution transformer network can effectively capture the complex features of EEG signals for the classification of AD patients versus control subjects and may be expanded to other types of dementia, such as FTD. This approach could improve the accuracy of early diagnosis and lead to the development of more effective interventions for AD. |
| first_indexed | 2024-03-12T22:27:55Z |
| format | Article |
| id | doaj.art-afff6252bb2d45c9bf34d0cde25c909e |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-12T22:27:55Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-afff6252bb2d45c9bf34d0cde25c909e2023-07-21T23:00:46ZengIEEEIEEE Access2169-35362023-01-0111718407185810.1109/ACCESS.2023.329461810179900DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG SignalsAndreas Miltiadous0https://orcid.org/0000-0003-0675-9088Emmanouil Gionanidis1Katerina D. Tzimourta2https://orcid.org/0000-0001-9640-7005Nikolaos Giannakeas3https://orcid.org/0000-0002-0615-783XAlexandros T. Tzallas4https://orcid.org/0000-0001-9043-1290Department of Informatics and Telecommunications, University of Ioannina, Kostakioi, Arta, GreeceIndependent Researcher, Atlanta, GA, USADepartment of Informatics and Telecommunications, University of Ioannina, Kostakioi, Arta, GreeceDepartment of Informatics and Telecommunications, University of Ioannina, Kostakioi, Arta, GreeceDepartment of Informatics and Telecommunications, University of Ioannina, Kostakioi, Arta, GreeceObjective: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that affects a significant percentage of the elderly. EEG has emerged as a promising tool for the timely diagnosis and classification of AD or other dementia types. This paper proposes a novel approach to AD EEG classification using a Dual-Input Convolution Encoder Network (DICE-net). Approach: Recordings of 36 AD, 23 Frontotemporal dementia (FTD), and 29 age-matched healthy individuals (CN) were used. After denoising, Band power and Coherence features were extracted and fed to DICE-net, which consists of Convolution, Transformer Encoder, and Feed-Forward layers. Main results: Our results show that DICE-net achieved an accuracy of 83.28% in the AD-CN problem using Leave-One-Subject-Out validation, outperforming several baseline models, and achieving good generalization performance. Significance: Our findings suggest that a convolution transformer network can effectively capture the complex features of EEG signals for the classification of AD patients versus control subjects and may be expanded to other types of dementia, such as FTD. This approach could improve the accuracy of early diagnosis and lead to the development of more effective interventions for AD.https://ieeexplore.ieee.org/document/10179900/Alzheimer’s diseasedeep learningdetectionEEGFrontotemporal dementiatransformers |
| spellingShingle | Andreas Miltiadous Emmanouil Gionanidis Katerina D. Tzimourta Nikolaos Giannakeas Alexandros T. Tzallas DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals IEEE Access Alzheimer’s disease deep learning detection EEG Frontotemporal dementia transformers |
| title | DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals |
| title_full | DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals |
| title_fullStr | DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals |
| title_full_unstemmed | DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals |
| title_short | DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals |
| title_sort | dice net a novel convolution transformer architecture for alzheimer detection in eeg signals |
| topic | Alzheimer’s disease deep learning detection EEG Frontotemporal dementia transformers |
| url | https://ieeexplore.ieee.org/document/10179900/ |
| work_keys_str_mv | AT andreasmiltiadous dicenetanovelconvolutiontransformerarchitectureforalzheimerdetectionineegsignals AT emmanouilgionanidis dicenetanovelconvolutiontransformerarchitectureforalzheimerdetectionineegsignals AT katerinadtzimourta dicenetanovelconvolutiontransformerarchitectureforalzheimerdetectionineegsignals AT nikolaosgiannakeas dicenetanovelconvolutiontransformerarchitectureforalzheimerdetectionineegsignals AT alexandrosttzallas dicenetanovelconvolutiontransformerarchitectureforalzheimerdetectionineegsignals |