Scala vestibuli prostheses - Comsol, python, and Neuron model

Cochlear implants (CI)s belong to the most efficient neuroprostheses, although their artificially generated neural patterns miss many details seen in the natural code. Therefore, improvement of the quality of the generated neural pattern is a permanent challenge in CI research.

Traditionally, the stimulating electrode arrays are inserted into the scala tympani, a cochlear cavity, which enables simple access for surgery. However, often deep insertion is blocked, e.g., by ossification, and the auditory nerve fibers (ANF)s of lower frequency regions cannot be stimulated causing severe restrictions in speech understanding. As an alternative, the electrode array can be inserted into the scala vestibuli, the other large cochlear cavity.

Some of the key mechanisms to generate a spiking pattern in the auditory nerve with a CI are known from single ANF recordings in cats. In contrast to cat, human neurons of the auditory nerve have other stimulation characteristics, mainly because of longer dendrites and non-myelinated somata. As ethical reasons prevent human single ANF recordings, several biophysically based computer simulation studies supported our knowledge about differences between feline and human ANF excitation with CIs.

Target of the project was to improve the state of the art of such simulation studies in two ways. (i) The first available three-dimensional pathways of human ANFs, reconstructed from micro-CTs, are not as simple in shape as assumed in previous studies. Noteworthy, first investigations show that the pathway irregularities have a considerable impact on the excitation pattern. (ii) Results from the plenty of scala tympani studies are not directly applicable for other electrode placings. This causes a lack of knowledge as not a single simulation study exists for scala vestibuli implants.

Context and methodology

3D Comsol model:

simulates the electrical potential inside the cochlea established by CIs placed in the scala vestibuli and scala thypani

python model:

controls and runs the Comsol model by calling it with distinctive parameters

manages Comsol model results by storing and reusing them

simulate nerve responses by using Neuron software (python interface)

allows parameter sweeps for systematic investigations

creates model results for further use

A finite element model (FEM) of the human cochlear together with a CI array implemented in Comsol Multiphysics 6.1 (https://www.comsol.com) was used to compute the extracellular potential along realistic auditory nerve fibers (ANFs). Then the spiking behavior of the ANFs was simulated in Neuron 8.2 with a well-established compartment model of cochlear neurons where the calculated extracellular potential distributions from the FEM served as an input. The whole pipeline was controlled by a custom-written python script that ran on a server with 64 parallel threads with 256 GB RAM. To control and access the Comsol model, the python library MPh 1.2.3 (https://github.com/MPh-py/MPh) was used in combination with Comsol batch commands. The Neuron model was fully implemented and controlled by using Neuron’s python interface.

Technical details

Provided as 7z files

Do not change the folder structure of python.7z, as results are stored and reused when running the python code

Includes: 3D cochlea geometry, Comsol model, python source code, Neuron extensions

Runs in Windows and Linux environments

Requires internet access for Comsol authentification during model execution

Requires a Comsol lincence including Electric Currents modul (ec), python 3 enviroment including common (scientific) libraries, MPh library, Neuron (Yale University) installation

Further details

see readme.txt

Code files are MIT licensed