gms | German Medical Science

Artificial Vision 2019

The International Symposium on Visual Prosthetics

13.12. - 14.12.2019, Aachen

Concept of a retinal closed-loop system with an on-chip fire-rate-detection algorithm

Meeting Abstract

Suche in Medline nach

  • Andreas Erbslöh - University of Duisburg-Essen, Electronic Components and Circuits, Duisburg/D
  • R. Viga - University of Duisburg-Essen, Electronic Components and Circuits, Duisburg/D
  • K. Seidl - University of Duisburg-Essen, Electronic Components and Circuits, Duisburg/D; Fraunhofer Institute for Microelectronic Circuits and Systems, Duisburg/D
  • R. Kokozinski - University of Duisburg-Essen, Electronic Components and Circuits, Duisburg/D; Fraunhofer Institute for Microelectronic Circuits and Systems, Duisburg/D

Artificial Vision 2019. Aachen, 13.-14.12.2019. Düsseldorf: German Medical Science GMS Publishing House; 2019. Doc19artvis14

doi: 10.3205/19artvis14, urn:nbn:de:0183-19artvis140

Veröffentlicht: 10. Dezember 2019

© 2019 Erbslöh et al.
Dieser Artikel ist ein Open-Access-Artikel und steht unter den Lizenzbedingungen der Creative Commons Attribution 4.0 License (Namensnennung). Lizenz-Angaben siehe http://creativecommons.org/licenses/by/4.0/.


Gliederung

Text

Objective: Presentation of a CMOS implementation for simultaneous stimulation of retinal cells, recording the cell activity and determining the fire-rate for each electrode independently.

Concept and methods: Today’s commercially available retina implants only perform an open-loop stimulation and the stimulation efficiency decreases because of the changes within the retinal cells during the operation. Therefore, a bi-directional link between the retinal cells and the electronic front-end allows to understand the signal processing of the retina and to adapt the stimulation. This is especially useful to compensate for aging and other effects on the quality of the metal-tissue-interface which may influence the stimulation efficiency. With this goal in mind an ASIC for closed-loop stimulation has been developed. Each of the eight electrodes are driven by a selectable current controlled or charge-controlled stimulation circuit and record the evoked response. By use of an on-chip algorithm, the fire-rate of the recorded cell response can be determined for each electrode. This is based on a two stage operation. At the input stage, a modified non-linear energy operator amplifies the action potential und suppressed the influence of the local field potential. The second stage performs an adaptive threshold comparator and counts the detected spikes with an upper limit of up to 256 spikes/sec. In addition, a programmable waveform generator is integrated to create the stimulation patterns. The chip can be controlled using an SPI Interface and the protocol allows multiple units to be cascaded.

Results: The ASIC has a total power consumption of 6.92 mW (0.87 mW per electrode) with active stimulation. The algorithm for the fire-rate detection has been tested with a VerilogA model of the Hodgkin-Huxley-model and it achieves an accuracy rate of 100% with a higher peak amplitude of 50 µV. The pre-amplifier has an effective input noise of 4.3 µV in his corner bandwidth (1.2 – 4.2 kHz).