Towards Feasible Deep Learning Approach Using EEG for Neurodegenerative Diseases

In this study, we propose the phase space reconstruction (PSR)-based deep learning framework for enhanced detection of frontotemporal dementia (FTD) utilizing electroen-cephalography (EEG). Our experimental results demonstrate significant improvements over conventional screening methods, particularl...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:The ... International Winter Conference on Brain-Computer Interface s. 1 - 5
Hlavní autoři: Park, Dogeun, Kim, Ho-Jung, Ju, Young-Gi, Kim, Keun-Tae, Wont, Dong-Ok
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 24.02.2025
Témata:
Accuracy
Brain-computer interfaces
Deep learning
Dementia
EEG
Electroencephalography
false- negative
frontotemporal dementia
MMSE
Neurological diseases
phase-space reconstruction
Protocols
Reliability
ISSN:2572-7672
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this study, we propose the phase space reconstruction (PSR)-based deep learning framework for enhanced detection of frontotemporal dementia (FTD) utilizing electroen-cephalography (EEG). Our experimental results demonstrate significant improvements over conventional screening methods, particularly the mini-mental state examination (MMSE), which exhibits a considerable false-negative rate of 39.1 % in FTD screening. The proposed framework achieved a remarkable 88.89 % accuracy in identifying previously misclassified FTD cases. The evaluation was conducted using a public dataset comprising 23 FTD patients and 29 healthy controls, wherein our framework consistently outperformed both traditional MMSE and the EEGNet-based approach. Through the integration of PSR techniques with advanced deep learning architectures, our findings suggest that EEG-based computational approaches could serve as robust complementary diagnostic tools in clinical FTD assessment protocols. These results underscore the potential of machine learning applications in enhancing the accuracy and reliability of neurological disorder diagnosis.
ISSN:2572-7672
DOI:10.1109/BCI65088.2025.10931330