Blind source separation of neural recordings and control signals

Neural signals recorded in parts of the body where voluntary movement has been retained can be used to control prosthetic devices that assist patients to regain lost function. The number of signals recorded to control these devices can be increased by using a single multi-contact electrode placed over a multi-fasciculated peripheral nerve. Recordings made using these electrodes can then be separated using blind signal separation (BSS) methods to recover individual fascicular neural activity. In this study, we investigate the feasibility of separating peripheral neural recordings, obtained using a multi-contact electrode, to recover individual fascicular signals. We implement BSS through independent component analysis (ICA) and investigate the effects of the number of contacts used and electrode layout on separation. Peripheral neural signals were simulated using a finite element model of the hypoglossal nerve of adult beagle dogs with a multi-contact cuff electrode placed around it. FastICA was then used to separate simulated neural signals. The separated and post-ICA processed neural signals were then compared to the original signals in the fascicles that caused them through correlation coefficient (CC) calculations. For n=50 trials, the CC values obtained were all higher than 0.9 indicating that BSS can be used to recover linearly mixed independent fascicular neural signals recorded using a multi-contact cuff electrode.