Tensor-based epilepsy prediction system with incomplete multi-way electroencephalogram

Autor: Nguyễn Thị Ngọc Anh, Võ Trung Hùng
Jazyk: English<br />Vietnamese
Rok vydání: 2024
Předmět:
Zdroj: Tạp chí Khoa học và Công nghệ, Pp 83-89 (2024)
Druh dokumentu: article
ISSN: 1859-1531
Popis: In this paper, our overarching goal is to propose a novel technology that will facility to analyze multi-way arrays electroencephalogram (EEG) brain signals to forecast epileptic seizure activity in the presence of missing entries while most previous conventional techniques commonly are restricted to perform the forecast epilepsy through 2D-based noninvasive EEG with complete channels. As such, the proposed method can forecast future trends of epilepsy activity while simultaneously dealing with missing data within one framework automatically. The key novelty of the proposed method demonstrates (1) exploiting both latent states and time series dynamics for detecting patterns trends information in missing reconstruction as well as prediction, (2) preserving the nature of tensor structure of multiway EEG brain signals. The proposed method performs its robustness via demonstrating high seizure forecasting accuracy through the comparative study with other techniques on the real public dataset with different scenarios of corrupted data.
Databáze: Directory of Open Access Journals