Real-time contextual feedback for close-loop control of navigation

Abstract

Objective: Close-loop control of brain and behavior will benefit from real-time detection of behavioral events to enable low-latency communication with peripheral devices. In animal experiments, this is typically achieved by using sparsely distributed (embedded) sensors that detect animal presence in select regions of interest. High-speed cameras provide high-density sampling across large arenas, capturing the richness of animal behavior, however, the image processing bottleneck prohibits real-time feedback in the context of rapidly evolving behaviors.Approach: Here we developed an open-source software, named PolyTouch, to track animal behavior in large arenas and provide rapid close-loop feedback in~5.7 ms, ie. average latency from the detection of an event to analog stimulus delivery, e.g. auditory tone, TTL pulse, when tracking a single body. This stand-alone software is written in JAVA. The included wrapper for MATLAB provides experimental flexibility for data acquisition, analysis and visualization.Main results: As a proof-of-principle application we deployed the PolyTouch for place awareness training. A user-defined portion of the arena was used as a virtual target; visit (or approach) to the target triggered auditory feedback. We show that mice develop awareness to virtual spaces, tend to stay shorter and move faster when they reside in the virtual target zone if their visits are coupled to relatively high stimulus intensity (≥49dB). Thus, close-loop presentation of perceived aversive feedback is sufficient to condition mice to avoid virtual targets within the span of a single session (~20min).Significance: Neuromodulation techniques now allow control of neural activity in a cell-type specific manner in spiking resolution. Using animal behavior to drive closed-loop control of neural activity would help to address the neural basis of behavioral state and environmental context-dependent information processing in the brain.

Prominent Inhibitory Projections Guide Sensorimotor Computation: An Invertebrate Perspective

“…[30][31][32][33] Accordingly, as C. elegans moves through a chemical gradient, it navigates its environment using a biased random walk, [34,35] just like bacteria navigate a chemical gradient, [36] using multiprotein receptor complexes for temporally integrating information. [41] Consequently, the integration of sensorimotor information generates highly context-specific actions, [43] whether in terms of a quick escape from a predator or navigation in response to food availability. [38][39][40] In C. elegans, the behavior of the nematode in the absence of food will change the navigational state from "searching" to "dispersal," each of which requires a distinct set of sensory neurons.…”

PASER for automated analysis of neural signals recorded in pulsating magnetic fields

“…The massive size of the subsequent raw voltage traces will make it infeasible to store them for offline processing. Moreover, real-time spike sorting allows for experimental conditions that adapt according to the neural responses that are observed which makes room for wider scientific investigation including but not limited to the neural basis of adaptive sensorimotor computation [78,79], contextual information processing [80][81][82][83] and storage [84][85][86][87], navigation [88][89][90], and information transfer in neural circuits [46,48]. Brain-machine-interfaces (BMI), like limb prosthetics, also necessitate that spike sorting is performed in real-time on a time-scale of hundreds of milliseconds [65], as these are usually controlled by direct neuronal signalling that is measured invasively by an array of electrodes [91].…”

Prominent Inhibitory Projections Guide Sensorimotor Computation: An Invertebrate Perspective

“…Consequently, the integration of sensorimotor information generates highly context specific actions [43] , whether in terms of a quick escape from a predator, or navigation in response to food perceptual and motor learning [41,[44][45][46][47][48][49] . Neural circuits have thus evolved to execute sensorimotor integration with ever increasing involvement from mnemonic contributions and executive control, for example in the mammalian brain, the circuits that plan, control, execute and monitor action generation [50][51][52][53] .…”