An In Vivo-Mimicking In Vitro Testbed for Brain-Computer Interfaces

Abstract

To reduce the amount of animal testing associated with brain-computer interfaces (BCI) therefore reducing the associated monetary cost and cost of animal lives, we report a novel in vitro test setup to validate BCIs. Past validation techniques used stand-alone 50-Ω function generators or function generators connected to high-valued resistors to replicate recording from clinical electrodes; however, neither of these two methods accounts for an electrode's complex, frequency-dependent impedance, and both underestimate the normally high electrode impedance. As such, optimizing and testing via function generators and resistors overestimates BCI sensitivity by at least 250 times and 2 times, respectively. To more closely replicate a clinical recording environment and accurately assess BCI performance, we utilize a function generator to mimic neural signals sent to a human-body-mimicking electrolyte fluid via a Pt electrode. These signals are then recorded via a neural electrode connected to the BCI under test. To validate the proposed method, we compare sensitivity of a previously reported fully passive BCI via the three methods. The presented results demonstrate that recording via the proposed clinically accurate setup when the BCI has been optimized with one of the two other test methods gives the lowest sensitivity in the frequency domain and worst signal stability in the time domain. Overestimating BCI sensitivity during in vitro testing results in the potential inability to record desired neural signals in vivo. The proposed method provides a means by which to accurately assess BCI performance, therefore reducing errors during animal trials and saving time, money, and lives.