Introduction: CochlearModel

A computational model is customized with real patient data by the neural response telemetry (NRT) amplitude.

The following framework let the user to predict the behaviour of auditory nerve stimulated by a CI (Cochlear Implant).

The computational model is divided into two types of FEM models:

• Electrode mode calculates the current densities that reach to the virtual neurons (VNs) when an electrode is stimulating with 1mA (stationary problem).

• Neuron mode calculates the potential that reach the electrode when a membrane current intensity is propagated along each neuron.

After that, the differential evolution (DE) algorithm adjusted the parameters (FVr_x = [α, β, Jmin, Jmax]) that minimize the error between real and simulated data.

Flowchart

The next flowchart shows a summary of the model:

Getting started

Prerequisites

This framework needs COMSOL Multiphysics®, Matlab® and the COMSOL MATLAB api (LiveLink™). The code was tested for the following version of the software aforementioned.

• COMSOL Multiphysics® version 5.6
• Matlab® R2021a
• COMSOL plugging LiveLink™ Matlab® for COMSOL 5.6

Installation

Clone the repo

git clone https://github.com/IUSIANI/CochlearModel.git

How to use

The main file is Rundeopt.m. The initial parameters for the model or the differential evolution (DE) are in the Modelconfig.m and DEconfing.m files respectively.

Folders explanation

• ./bases/: save the potential V reaching the electrodes (from Neuron Mode) and/or the current density at neurons J (from Electrode Mode) when the electrode is stimulating with 1mA, in a .mat file. It allows us to save time by loading the model results.

• ./comsol/: contains usefull matlab functions .m used in this framework for creating the comsol model (geometry, material, mesh, phisics, solution, results ...), the differential evolution (DE), make custom plots or some utils functions. The init.m file initilize the aforementioned folders, define global variables and assign output files in folders.

• ./currentInputs/: contains the current membrane (WB, bEIF...) in .csv for different inputs. This is used in Neuron Mode.

• ./data/: contains '.m' where clinical data of the patient (Stimulating Electrode, Stimulating Intensity, Neuronal Response) is stored. Moreover, an example of transimpedance matrix from patient 1 is also included.

• ./results/: Different outputs is stored such as images, comsol models (.mph), optimizated params (.mat) (FVr_x = [α, β, Jmin, Jmax]), tabular data with the NRT real and simulated or the weights (.csv|.xlsx) or the full matlab workspace (.mat).

Tested

The code was tested over the following conditions:

• OS: Ubuntu 18.04.6 LTS

• CPU: Intel (®) Core (™) i9-10900X

• DRAM: 128Gb DDR4

• GPU: NVIDIA Quadro P400

Elapsed Times:

• Neuron Mode (Comsol) ≈ 1h 15 min

• Electrode Mode (Comsol) ≈ 15 min

• Differential Evolution (DE) Algorithm ≈ 2 min

• Total elapsed time ≈ 1h 30min

Cite as

Ramos-de-Miguel, Á., Escobar, J. M., Greiner, D., Benítez, D., Rodríguez, E., Oliver, A., Hernández, M., & Ramos-Macías, Á. (2022). A phenomenological computational model of the evoked action potential fitted to human cochlear implant responses. In W. Nogueira (Ed.), PLOS Computational Biology (Vol. 18, Issue 5, p. e1010134). Public Library of Science (PLoS). https://doi.org/10.1371/journal.pcbi.1010134

Bibtex

@article{RamosdeMiguel2022,
  doi = {10.1371/journal.pcbi.1010134},
  url = {https://doi.org/10.1371/journal.pcbi.1010134},
  year = {2022},
  month = may,
  publisher = {Public Library of Science ({PLoS})},
  volume = {18},
  number = {5},
  pages = {e1010134},
  author = {{\'{A}}ngel Ramos-de-Miguel and Jos{\'{e}} M. Escobar and David Greiner and Domingo Ben{\'{\i}}tez and Eduardo Rodr{\'{\i}}guez and Albert Oliver and    Marcos Hern{\'{a}}ndez and {\'{A}}ngel Ramos-Mac{\'{\i}}as},
  editor = {Waldo Nogueira},
  title = {A phenomenological computational model of the evoked action potential fitted to human cochlear implant responses},
  journal = {{PLOS} Computational Biology}
}

License

Distributed under the MIT License. See LICENSE.txt for more information.

Acknowledgments

This work has been supported by Ministerio de Ciencia, Innovación y Universidades, Gobierno de España, grant contract: PID2019-110185RB-C22 and Agencia Canaria de Investigación, Innovación y Sociedad de la Información, Consejería de Economía, Conocimiento y Empleo del Gobierno de Canarias and European Regional Development Funds (ERDF/FEDER), grant contract: PROID2020010022.