Meeting Banner
Abstract #5413

Toward real-time head motion corrections in simultaneous EEG-fMRI: Convolutional neural network classification of EEG-derived motion independent components.

Chung-Ki Wong1, Vadim Zotev1, Raquel Phillips1, Masaya Misaki1, and Jerzy Bodurka1,2,3

1Laureate Institute for Brain Research, Tulsa, OK, United States, 2Stephenson School of Biomedical Engineering, University of Oklahoma, Norman, OK, United States, 3Center for Biomedical Engineering, University of Oklahoma, Norman, OK, United States

In EEG-fMRI, EEG electrodes record head motions with a high temporal resolution (EEG-motion-sensor), which can be utilized for retrospective slice-by-slice fMRI motion correction. EEG motion components derived from independent component (IC) analysis were automatically identified by the common features observed in the IC mean power spectral density, spatial projection topographic map, and signal contribution. For real-time application of the EEG-motion-sensor, pre-trained models are desirable for faster classification. We used convolutional neural network to evaluate performance of motion-IC classification model. High speed and classification accuracy were achieved on a large EEG-fMRI dataset, suggesting the possibility of real-time EEG-motion-sensor applications for fMRI.

This abstract and the presentation materials are available to members only; a login is required.

Join Here