Mulitmodal Neuroimaging for a Non-Invasive Biomimetic Brain-Computer Interface

University Of Minnesota, Minneapolis MN

Investigators

Linked publications & trials

Abstract

? DESCRIPTION (provided by applicant): Over 300,000 Americans are currently living with some level of paralysis caused by spinal cord injury (SCI). Over 55% of these cases result in tetraplegia severely diminishing neuromuscular control of the upper extremities. SCI patients therefore rely significantly on specialized care for everyday tasks and often express a strong desire for increased autonomy in their everyday lives. While surgical reconstruction can be used to help some patients regain motor function, brain-computer interface (BCI) technology is beginning to show great promise for increasing the level of independence a patient can achieve. These BCIs provide a synthetic channel to the brain for conveying motor intent such that a patient can control a robotic prosthesis in an intuitive fashion to accomplish common tasks. Nevertheless, successful BCI approaches currently require electrodes to be implanted into the brain to detect useable signals and introduce significant short and long term health risks associated with the surgery. While non-invasive BCIs have achieved successful control of external devices in up to three dimensions, control with higher degrees-of-freedom is significantly limited by the poor spatial resolution of the electroencephalographic (EEG) signals detected on the scalp. The proposed research aims to address this limitation and demonstrate intuitive prosthetic control using EEG-based BCI technology. The main hypothesis of this work is that high spatio-temporal multimodal imaging techniques will allow motor imagery tasks involving dexterous manipulations of the hands to be separated in an online BCI using non-invasive EEG recordings. Three specific aims are proposed to build up to the online demonstration of non-invasive prosthetic BCI control. Firstly, we will investigate the separability of motor imaginations involving right hand flexion, extension, supination, and pronation using functional MRI (fMRI) and EEG source imaging (ESI) techniques. We will identify biomarkers, such as EEG frequencies and time windows, and parameters, such as the fMRI weighting, which optimize discrimination between the four different tasks. Secondly, we will establish an online ESI-based BCI platform for investigating cortical changes involved in motor learning. Using this setup, we will employ fMRI-constrained source imaging to determine if healthy human subjects can be trained to modulate focal frequency-specific activity in their brain. Thirdly, we aim to use the ESI-based BCI to train healthy human subjects in natural and intuitive control of a robotic prosthesis using dexterous motor imaginations of the right hand. We expect that the results of this work will have a significant effect on the BCI community for translating non-invasive technologies toward clinical use and improving the quality of life of SCI patients.