Real-time closed loop phase detection

This library adapts a closed-loop phase detection algorithm (from open-ephys modules) to work inside the faster SpikeGadgets environment.

This software detects 4 possible phases in your band-pass of interest (peak, trough, rising zero, falling zero). If band-pass power rises above a threshold on some user-defined set of tetrodes with the correct phase, it triggers a laser to modify brain activity. This was used to pilot exploring a role for theta and beta phase-coding on animal behavior.

A mirror of this code with accompanying trodes software can be found here, in the PhaseDisrupt feature branch: https://bitbucket.org/mkarlsso/trodes/src/PhaseDisrupt/

My implementation rides atop Loren Frank Lab's FSData/FSGui ripple-closed loop detection. This tool is a direct descendent of their closed-loop module and adds the ability to trigger a laser using phase or amplitude of an arbitrary band-pass signal (user-configurable frequencies).

Below is an example of fast beta (20-30hz) phase disurption, with brain samples streaming at 1500hz. I can also acheive reliable theta wave (6-12hz) disruption with this.

Key capabilities of this closed-loop code, old and new:

Neural data acquisition

• Acquires continuous neural data streams and spike waveforms from multiple electrodes
• Configurable selection of electrodes and sampling rates
• UDP data streaming from acquisition module to processing code

Real-time feature detection

• Bandpass filters data to isolate oscillations like theta, beta, ripple bands
• Detects amplitude, phases, and rhythmicity of neural signals
• Algorithms for spatial position, ripple , theta phase targeting

Closed loop stimulation

• Triggers electrical or optogenetic stimulation in real-time based on detected features
• Highly configurable stimulation parameters and patterns
• Precise targeting of stimulation timing based on phase or amplitude detection

Graphical configuration interface

• GUI for selecting electrodes, configuring algorithms, designing stimulation patterns
• Saves/loads system configurations for repeatable experiments

Optimized for real-time performance

• C++ code and libraries optimized for real-time signal processing
• Sub-millisecond latency from data acquisition to stimulation output
• Runs reliably on commodity Linux workstations with minimal dropped data