سامانه همایش‌ها TSTA | ثبت‌نام و ارسال مقاله کنفرانس‌ها و کنگره‌های ملی

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

An Automatic Pipeline for Simultaneous EEG-fMRI Artifact-removal (SEFA)

نویسندگان :

Farid Hosseinzadeh¹ Amin Mohammad Mohammadi² Mehrdad Anvarifard³ ُSasan Keshavarz⁴ Elias Ebrahimzadeh⁵ Hamid Soltanian-Zadeh⁶

1- دانشگاه تهران 2- دانشگاه تهران 3- دانشگاه تهران 4- دانشگاه تهران 5- دانشگاه تهران 6- دانشگاه تهران

کلمات کلیدی :

Simultaneous EEG-fMRI،EEG،preprocessing،artifact removal،automation،pipeline،ٍَُّSEFA

چکیده :

Simultaneous EEG–fMRI provides complementary temporal and spatial information about brain function, but its utility is hindered by severe scanner-induced artifacts such as gradient and ballistocardiographic (BCG) noise. Manual artifact correction is effective but labor-intensive, inconsistent, and difficult to scale. We introduce SEFA, a fully automated two-stage preprocessing pipeline for simultaneous EEG–fMRI that integrates MRI-specific artifact correction (average artifact subtraction, optimal basis set, and PCA/OBS modeling) with state-of-the-art EEG cleaning techniques adapted from a previous popular standard EEG preprocessing pipeline, HAPPE, including automated independent component classification (MARA and ICLabel), bad-channel detection, multitaper regression for line noise, and segment-level quality control. Validation against manually corrected datasets from a reward-based decision-making task demonstrated that SEFA achieves near-perfect equivalence with expert preprocessing. Event-related potentials (ERPs) from both approaches exhibited indistinguishable morphology, latency, and amplitude, with mean channel-wise correlations of r = 0.91 ± 0.14, and 72% of electrodes exceeding r > 0.90. Signal-to-noise ratio (SNR) improved from ~0.8 dB in raw data to 6.7 dB with SEFA, matching manual performance (6.9 dB). Statistical testing confirmed no significant differences in ERP amplitude or latency between automated and manual methods (all p > 0.1). By reducing operator bias and cutting processing time from hours to minutes, SEFA enables reproducible, scalable, and clinically feasible preprocessing of simultaneous EEG–fMRI data.