Article
Investigation of spatial selectivity using blind source separation algorithm for electrical retinal stimulation
Search Medline for
Authors
-
Mahmut E. Celik
- Electrical and Electronics Engineering Department, Gazi University, Ankara/TR
-
D. Nguyen
- L’Institut de la Vision, Paris/F
-
E. Scorsone
- CEA Saclay, Paris, France
-
L. Rousseau
- ESIEE Paris/F
-
S. Picaud
- L’Institut de la Vision, Paris/F
| Published: | December 10, 2019 |
|---|
Outline
Text
Objective: It is aimed to develop a tool to check spatially-selective activation capability of the electrical stimulation by quantitatively differentiating electrically evoked potentials (EEPs) recorded by a diamond implant using Blind Source Separation Algorithm. The provided activation maps of the visual cortex give insight into spatial selectivity of the retinal electrical stimulation.
Method: EEPs are recorded using diamond implant from brain surface while electrical stimulation is applied to the retina through an implant with gold electrodes placed to the subretinal space. As one of the most commonly used Blind Source Separation methods, Independent Component Analysis is used to project recorded data into the weighted sum of original data as new independent data channels, which is called Independent Components (ICs). A classification process for ICs is performed to extract ICs with artifact, noise and no peaks. The activation maps of the visual cortex are created based on each selected ICs.
Results: Electrical stimulation causes artifacts on all recording channels. The amplitude of EEPs is directly related to the stimulus intensity applied on the retinal implant. For each stimulation electrode, i.e. different retinal stimulation points, activation maps of the visual cortex are obtained using selected ICs, showing distinct activation areas. Preliminary results show that the effect of individual stimulation electrode on selective activation of the visual cortex can be evaluated by the suggested method.
Discussion: The proposed approach is used both to separate linearly mixed information from the recorded channels and to obtain activation maps using the inverse of transformation matrix with selected statistically ICs. More experimental data can confirm that this approach reveals spatial selectivity of individual stimulation channels of an array.
Acknowledgment: This work is supported by the programs of the Scientific and Technological Research Council of Turkey (TUBITAK) 2219 and NEURODIAM under European Research Council (ERC).
