سامانه همایش ها

English

صفحه اصلی / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

EJES: A Diverse Estimator Bank Framework for High-Resolution EEG/MEG Source Localization

نویسندگان :

Reza Khajehsarvi¹ Sayed Mahmoud Sakhaei² Sadegh Jamshidpour³

1- دانشگاه صنعتی نوشیروانی بابل 2- دانشگاه صنعتی نوشیروانی بابل 3- دانشگاه صنعتی نوشیروانی بابل

کلمات کلیدی :

Brain Source Reconstruction،Inverse Problem،Localization،Estimator Bank،MUSIC،LCMV

چکیده :

Brain source reconstruction from electroencephalography (EEG) and magnetoencephalography (MEG) signals is a central inverse problem in neuroscience. Classical localization algorithms, however, are highly sensitive to realistic conditions such as limited data length, low signal-to-noise ratios, and structured interference, which greatly restricts their reliability in clinical and research applications. To address this limitation, we introduce the Ensemble of Joint Estimation Strategy (EJES), a novel framework for robust source localization. EJES leverages algorithmic diversity by constructing a heterogeneous bank of estimators drawn from two distinct families: subspace-based approaches, implemented as weighted Multiple Signal Classification (MUSIC) estimators, and spatial filtering approaches, implemented as beamformers operating on different powers of the data covariance matrix. A final, stable source estimate is obtained by selectively integrating the outputs of these estimators through a robust consensus mechanism. The performance of the EJES framework was quantitatively evaluated against standard, single-algorithm approaches through extensive Monte Carlo simulations. Results consistently demonstrate that EJES provides significantly more accurate and stable localization than conventional single-algorithm methods, particularly under challenging scenarios combining short data segments, low signal quality, and high interference. These findings underscore the potential of ensemble strategies to improve the robustness of neuroelectromagnetic source reconstruction, providing a more reliable tool for noninvasive brain imaging.