Clustering Electrical Retinal Responses Based on Averaged Peri-stimulus Time Histograms


ÇELİK M. E., Jalligampala A., Rathbun D. L.

International Symposium on Fundamentals of Electrical Engineering (ISFEE), Bucharest, Romania, 1 - 03 November 2018 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/isfee.2018.8742413
  • City: Bucharest
  • Country: Romania
  • Gazi University Affiliated: Yes

Abstract

Visual information is delivered to retinal ganglion cells (RGCs) by the retinal circuit and then transmitted from RGCs to the brain. Among these RGCs, there are over 30 different types and each type carries specialized information about the visual world. The question of whether and how RGCs also show different types of responses to electrical stimulation has been much debated for a long time. This work aims to classify RGC responses to pulsed electrical stimulation based on the peri-stimulus time histogram (PSTH). Both hard and soft partitioning methods were used to perform unsupervised clustering. The results of the clustering algorithms were evaluated using cluster validity indices with no prior information. The result was that the k-means as a hard partitioning method provided about 17 clusters, while the fuzzy Gustafson Kessel algorithm gave about 25 clusters. After manual inspection, it was concluded that the clusters are both diverse and specific enough to likely correlate to known RGC types. Therefore, performing clustering based on the electrical pulse-elicited PSTH could be a powerful tool and can contribute in differentiating RGCs and stimulating them selectively.