A simulation study on decoding algorithms for brain-machine interfaces with the non-stationary neuronal ensemble activity

Abstract

Intracortical brain-machine interfaces (BMIs) aim to provide motor functions via high-performance neural prosthetics to patients with tetraplegia. BMI simulation can be useful in certain circumstances to evaluate functional environments of neuronal conditions, avoiding clinical issues such as infection or tissue damage. In this study, we performed a stimulation study on an intracortical BMI for the reconstruction of three dimensional arm movements. We compared two key decoding algorithms, the Kalman filter (KF) and the optimal linear estimator (OLE), when neuronal preferred directions (PDs) varied over time. The simulation results revealed that the KF decoded the movement direction better than the OLE for non-stationary neuronal ensemble activity. These results may provide a guidance for selecting an appropriate decoding algorithm for various environments of intracortical BMIs.