A closed-loop Brain-Machine Interface simulator based on computer mouse control

Abstract

Brain Machine Interface (BMIs), which delivers information from the brain to external devices, can provide independence to the disabled who cannot control their body voluntarily. Over the last decade, BMIs have been developed rapidly by the support of a number of non-human and human studies to provide the proof-of-concept of a BMI. BMI development draws upon the findings and insights from neuroscience, engineering, mathematics and computer science, typically testing new approaches using the pre-existing neural database due to practical issues such as time limit, subject training and on-site system maintenance. For those studies which involve repetitive changes of parameters and models, the cost of performing closed-loop experiments would exponentially increase. However, testing in an offline condition may not assure that the concept confirmed in an open-loop (OL) analysis also applies to closed-loop (CL) applications. To provide both OL and CL testing environments, we propose a new software-based tool that simulates closed-loop neural control through BMIs. In this tool, neural data are generated by the user's computer mouse input through a linear tuning function whose parameters are configured from the real neural data. The decoding algorithm as well as the control interface can be easily customized and one can execute a closed-loop experiment in a similar way to real BMI control environment. We present a case study of using this BMI simulator to investigate the effect of closed-loop training on control performance. It is anticipated that the BMI simulator may be able to allow researchers to investigate the entire process of a closed-loop BMI study and address the issues which cannot be resolved in an open loop data analysis.