Feasibility Analysis of Genetically-Encoded Calcium Indicators as a Neural Signal Source for All-Optical Brain-Machine Interfaces

Abstract

Optical techniques such as two-photon (2p) calcium imaging have the potential to transform the way we interrogate neural circuits, both in the realm of basic neuroscience and in the development of brain-machine interfaces (BMIs). This may be possible by overcoming some of the limitations of electrophysiological methods. Here we ask if optical imaging signals, in particular 2p calcium imaging signals from GCaMP6, can benefit BMIs despite their relatively long activity-response time constants, low signal-to-noise ratios (SNRs), and slow acquisition frame rates. We employed motor cortical electrode array recordings as the basis for generating synthetic 2p GCaMP signals. We then decoded movement kinematics from these surrogate data using a state-of-the-art BMI decoder algorithm. We found that it was possible to decode the position and velocity of the hand from synthetic imaging signals. We quantified the decoder performance using standard mean squared errors (MSEs) and Pearson's correlation coefficient (r) measures. Decode quality varied considerably as a function of SNR and the frame rate of data acquisition. Future computational and experimental research is required to quantify SNR more accurately and to increase the imaging frame rate while maintaining high SNR, in order to improve all-optical BMI (o-BMI) performance. This study should help establish the feasibility and design space of o-BMIs.