Modeling Neural Recording

This feature is the first introduction of multiple cuffs within the ASCENT pipeline. Begin setting up a run as usual with ASCENT (refer to Getting Started for guidance and a tutorial). To use our pipeline to model neural recording, two major elements need to be present within the JSON configuration files. To prompt the pipeline to model neural recording of single fiber action potentials (SFAPs), the user should first provide a list of two JSON Objects to the cuff property within the Model configuration file. Each JSON Object within this list should have all the original required cuff properties outlined in Model including the "index" property which assigns the cuff an index identifier that can be referenced to set contact weights and cuff type (stimulation or recording) in Sim. The second major element is defining the electrode contact weights in the Sim file. As with a original single cuff ASCENT run through, the active_srcs property in this file contains key-value pairs of contact weights for each preset cuff (as described in Sim File). Now, for modeling recording, the user should add an additional active_recs property that serves the same purpose for the recording cuff. For these two properties, the user must add an additional key-value pair called cuff_index, which ASCENT uses to matche the weighting properties to the cuff indices we defined in the Model file. Any index defined in "active_srcs" designates a stimulating cuff. Any index defined in "active_recs" designates a recording cuff. Remember to include any cuff preset names the relevant stimulation and recording cuffs in config/system/cuffs. Run the pipeline as usual from the root directory with python run pipeline <insert run index>. Following the completion of the run, change directories to /submit and run python submit.py <insert run index> to submit the NEURON simulations. Following the completion of the NEURON simulations, navigate to the root ascent directory and run python run import_n_sims <insert run index> to transfer NEURON results from the submit directory to the sample directory, where the data can be accessed by the scripts in examples/analysis.

Similar to the tutorial (See Getting Started), we provide example configuration files for modeling stimulation and recording of a rat vagus nerve. The files can be found in examples/cap_tutorial.

NEURON Outputs

When running a model with a recording cuff, in addition to the standard stimulation cuff output files, running NEURON simulations using submit.py will produce an output .dat file for every n_sim generated, called SFAP_time, containing the recorded SFAP collected by the recording cuff. Each SFAP file has a column for the stimulation waveform’s time points (ms) and a column for corresponding recorded SFAP (uV).

Optionally, the user can set the saving > cap_recording > Imembrane_matrix boolean variable in the Simulation file to true to save a transmembrane current matrix for each fiber which contains all the transmembrane currents of every compartment across time. Be careful, as these files can be quite large if there are many fiber simulations.

Current Limitations

The multiple cuff configuration framework is designed to support any number of cuffs, but has only been tested using two cuffs, with one designated as the stimulation cuff and the other designated as the recording cuff.

Analysis Scripts

Plotting Single Fiber Action Potentials (SFAPs)

The examples/analysis/plot_SFAPs.py script allows users to visualize the recorded single fiber action potentials (SFAPs) that resulted from modeling the neural recordings. The user should open the script and update the sample, model, and sim number accordingly in the call to query, as in the other analysis scripts. The user may plot the SFAPs from all the fibers simulated (default behavior), controlled by the all_fibers input boolean argument to the call to the query class’ sfap_data function. Alternatively, the user may choose to indicate a subset of fiber indices to plot using the fiber_indices variable at the top of the script. If both variables are provided, and all_fibers is True, the fiber_indices are ignored and all fibers will be provided. Run the script from the ascent root directory with the command python examples/analysis/plot_SFAPs.py.

Plotting Compound Action Potentials (CAPs)

The examples/analysis/plot_CNAP.py script allows users to visualize the combined action potentials of multiple fibers. The user should open the script and update the sample, model, and sim number accordingly in the call to query, as in the other analysis scripts. As described above, the user may plot the combined SFAPs from all simulated fibers (default behavior), or may choose to only visualize a subset of fibers by indicating their indices in the fiber_indices variable. Run the script from the ascent root directory with the command python examples/analysis/plot_CNAP.py.

Generating Current Templates per Fiber Diameter

In examples/analysis, a script is provided called generate_templates.py which allows users to generate transmembrane current action potential templates for each fiber diameter in a given simulation. These templates require the transmembrane current matrices to be saved, as described above. The templates can be used in conjunction with the following MATLAB repository (https://github.com/eurypt/CAPulator) to simulate whole-nerve fiber populations efficiently with a a speed up of 27,000 - 1,900,000x in terms of CPU hours [Peña et al., 2024]. Rather than simulating every action potential of every nerve fber, the template method uses the action potential from a single point in space and a few discrete nerve fiber diameters to interpolate the action potential at every point in space across all fiber diameters present in the nerve. See the publication for more details.

To generate the templates, open the script and update the sample, model, sim, and fiber values at the top of the file accordingly. The fiber variable refers to the fiber number within each generated n_sim. Then, run the script from the ascent root directory with the command python examples/analysis/generate_templates.py. The action potential templates are expected to differ in latency depending on the stimulation model parameters, and the templates differ across fiber types. Thus, the user should run the script for each distinct set of stimulation parameters or fiber types.