Analysis of an open source, closed-loop, realtime system for hippocampal sharp-wave ripple disruption

Abstract

Objective: The ability to modulate neural activity in a closed-loop fashion enables causal tests of hypotheses which link dynamically-changing neural circuits to specific behavioral functions. One such dynamically-changing neural circuit is the hippocampus, in which momentary sharp-wave ripple (SWR) events-≈ 100 ms periods of large 150-250 Hz oscillations-have been linked to specific mnemonic functions via selective closed-loop perturbation. The limited duration of SWR means that the latency in systems used for closed-loop interaction is of significant consequence compared to other longer-lasting circuit states. While closed-loop SWR perturbation is becoming more wide-spread, the performance trade-offs involved in building a SWR disruption system have not been explored, limiting the design and interpretation of paradigms involving ripple disruption.

Approach: We developed and evaluated a low-latency closed-loop SWR detection system implemented as a module to an open-source neural data acquisition software suite capable of interfacing with two separate data acquisition hardware platforms. We first use synthetic data to explore the parameter space of our detection algorithm, then proceed to quantify the realtime in vivo performance and limitations of our system.

Main results: We evaluate the realtime system performance of two data acquisition platforms, one using USB and one using ethernet for communication. We report that signal detection latency decomposes into a data acquisition component of 7.5-13.8 ms and 1.35-2.6 ms for USB and ethernet hardware respectively, and an algorithmic component which varies depending on the threshold parameter. Using ethernet acquisition hardware, we report that an algorithmic latency in the range of ≈20-66 ms can be achieved while maintaining <10 false detections per minute, and that these values are highly dependent upon algorithmic parameter space trade-offs.

Significance: By characterizing this system in detail, we establish a framework for analyzing other closed-loop neural interfacing systems. Thus, we anticipate this modular, open-source, realtime system will facilitate a wide range of carefully-designed causal closed-loop experiments.