Abstract
Purpose :
Retinal prostheses electrically activate surviving retinal ganglion cells (RGCs) to create artificial vision. Given the incredible neural information transmission from the retina during natural viewing, it is critical to consider how much information RGCs would carry for the prosthetic vision. Recently, we reported ON RGCs transmit richer neural information than OFF RGCs by more heterogeneous spiking responses arising from the same electric stimulus. Here, we quantified neural information changes depending on the levels of heterogeneity across simulated spike trains.
Methods :
We used neural computational methods to generate correlated spike trains and calculate the amount of information of those simulated spiking activities. First, we created five groups of 1-sec-long spike trains using a modified version of ‘Brian 2’, an open-source simulator for spiking neural networks. All groups had the same range of mean firing rate (80±20 Hz) and peak firing rate (200±50 Hz) but different levels of correlations across spike trains in each group. Second, 50 spike trains were randomly chosen until their spike time tiling coefficient (STTC) average reached specific values (i.e., 0.1, 0.3, 0.5, 0.7, and 0.9). Then, the neural information was computed for 200 of random combinations of 15 spike trains.
Results :
The average STTCs of the five groups were 0.09±0.07, 0.31±0.08, 0.50±0.07, 0.69±0.06, and 0.90±0.04 (mean±std). The amounts of neural information transmitted by 15 cells of each group were 7.59±0.06, 7.06±0.09, 6.14±0.11, 4.84±0.12, and 2.33±0.15 bits, respectively. Consistent with our work, more homogeneous spike trains resulted in more reduced information. However, it is notable similar increments in STTCs produced different information reductions. For example, the neural information was slightly decreased by 0.53 bits (p<0.001) for the STTC change from 0.09 to 0.31. In contrast, the information was substantially decreased by 2.51 bits (p<0.001) for the STTC change from 0.69 to 0.90.
Conclusions :
Our results showed the level of cell-to-cell spiking heterogeneities significantly impacts on the information transmission: increase in the correlation level leads to decrease in the quantity of neural information. This suggests optimization of spiking correlation in RGC populations may enhance the performance of the prosthetic vision.
This is a 2021 ARVO Annual Meeting abstract.