Computational Modeling of Retinal Neurons for Visual Prosthesis Research - Fundamental Approaches

This protocol helps predict the response of neural cells to different stimuli, allowing us to explore novel ideas in neural stimulation without needing physical prototype or life tissue. It can be used as a preliminary study to test concepts that are challenging to emulate physically, and it is cost and time-efficient, enabling testing of large numbers of parameters. This method can predict neural response in other neural stimulation systems besides the retina.

It is also not limited to electrical stimulations, but can be used for light stimuli as well. To begin, run the FEM software and click on model wizard, then 3D. In the select physics list box, expand AC/DC and select electric fields and current.

Click on study and add a stationary study under the general studies option, and then click on done. In the geometry settings, change the unit of length from meter to micrometer. Right-click on geometry one, then click block to create a block domain.

Repeat this step two more times to create three blocks in total. For all blocks, set both the depth and width to 5, 000 micrometers and assign the height value for the block. Change the base option to center and assign Z-values for each block.

To create a work plane for adding an electrode to the model, right-click on the geometry one in the model tree and choose work plane. Click on work plane one and change the plane type to face parallel. Click on the activate selection button below the plane type and choose the bottom surface of block one.

Click on parameters one and define the value for the radius of the electrode. To draw a disk electrode on the work plane, click on plane geometry under work plane one and click on sketch in the main toolbar. Select circle, click anywhere inside the rectangle in the graphics tab, and drag to create a disk electrode.

Change the radius to a predefined value in micrometer, xw and yw to zero micrometer, and then click on build all. In the model tree, right-click on material, then click blank material, and then click on material one and change the selection to manual. Click on the domains in the graphics window so that only domain one is chosen.

Choose material properties, basic properties, then click electrical conductivity and click on the add to material button. Change the electrical conductivity value to 0.043 siemens per meter. Repeat the steps for domains two and three with the electrical conductivity values of 0.7 and 1.55 siemens per meter, respectively.

To mesh a 3D model, go to the model tree and right-click on mesh one, then click free tetrahedral. Click on free tetrahedral one and choose build all. To apply physics to FEM, expand electric currents one in model tree and check whether current conservation one, electric insulation one, and initial values one are listed.

Then right-click on the electric currents. Then click ground and apply this to the surface furthest from the electrode. Next, right-click on the electric currents.

Then click floating potential assigned to the disk electrode and change the I0 value to one microampere to apply a unitary current. To run the simulation with a parametric sweep, in the model tree, right-click on study one, then click parametric sweep. Click on parametric sweep, and in the study setting table, click on add, and then choose elec_rad for the parameter name.

Type 50, 150, 350, 500 for the parameter value list and micrometer for the parameter unit. Then click on compute to run the study. To import the morphology using the CellBuilder feature, run nrngui from the NEURON Computational Suite's installation folder.

Next, click on tools, then click miscellaneous, import 3D, and then tick the choose a file box. Locate the downloaded SWC file and click on read. Once the geometry has been imported, click on export, then CellBuilder.

To create a HOC file of the imported cell morphology, go to the subsets tab and observe the subsets that have been predefined in the model. Tick the continuous create box, go to management, then click export and export the morphology as rgc.hoc. To view the morphology of the cell, click on tool, model view, one real cells.

Then click root soma zero on the toolbar. Right-click on the appearing window and click on access type and view access. By visual inspection, this model's dendritic field diameter should be around 250 micrometers.

Close the NEURON windows for now. Open the FEM software. Go to the Application Builder.

Right-click on methods in the Application Builder tree. Choose new method and click on okay. Go to file, then click preferences and methods.

Tick the view all codes box and click on okay. Write the HOC file that loads the coordinates of the neurons'segments to a text file. Use the FEM method script to shift the values to match the desired location and save a text file containing the coordinate values for the new location of the cell.

Open the COMSOL method and save the shifted coordinate and voltage values. To run the automated steps in the FEM software, switch to the model builder, developer, run method, and click method one. This will produce DAT files with the appropriate voltage values.

Loop the simulations in a general-purpose programming language by opening the chosen IDE and click on new file to make a new script, as determined in the text manuscript. Finally, click on run or press F5 to run the script, which will also open the NEURON Computational Suite's GUI. Graph the response of the NEURON model to the extracellular stimulation in the NEURON Computational Suite's GUI.

To do this, run stimulation. hoc, click on graph, then click voltage access from the toolbar, and on the graph window, right-click anywhere and choose, plot what? Type in axon.

v1 in the variable to graph field, which means that it will plot the transmembrane potential of the last segment of the axon per time step. The model described here confirmed that increasing the suprachoroidal electrode size at a 0.25 millisecond pulse width increased the activation threshold of the model neuron. The action potential characteristics were observed to validate the model.

The latency or the time between the stimulus onset and the peak of the action potential spike ranged from 1 to 2.2 milliseconds. This corresponded to the short latency spiking due to non-network-mediated retinal activation. The spike width of this model was one millisecond and this is in the same range as the spike widths of rabbit RGCs measured in vitro.

The model showed the lowest threshold when the narrow segment of the axon was immediately above the disk electrode and it increased as the X distance became larger. Moving the electrode further toward the distal axon produced a lower threshold as compared to moving the electrode toward the dendrites due to the presence of the axon initial segment and the narrow segment where the sodium channels are more prevalent. The described method is easy to apply and it accelerates researchers in producing proofs of concept for novel stimulation methods or neuro-electro designs.