See globally, spike locally: oscillations in a retinal model encode large visual features.

Area of Science:

• Computational Neuroscience
• Visual System Modeling
• Neural Coding

Background:

• Traditional models of visual processing rely on average firing rates of retinal ganglion cells.
• Global stimulus properties, like object size, are challenging to infer solely from local, time-averaged firing rates.

Purpose of the Study:

• To investigate if coherent oscillations among retinal ganglion cells encode global topological properties, such as object size.
• To determine if this encoding is detectable at physiologically realistic scales.

Main Methods:

• Analysis of artificial spike trains with experimentally measured oscillatory correlations.
• Extraction of upper gamma-band oscillatory power from multi-unit spike trains.
• Comparison of oscillatory power and coincidence detection for size discrimination tasks.
• Examination of retinal activity transitions using random binary images and percolation theory.

Main Results:

• High-frequency gamma-band oscillations effectively discriminate between small and large spots.
• Size discrimination performance improves with increased cell numbers and analysis duration.
• Synchrony (coincidence detection) shows poorer performance compared to oscillatory power.
• A sharp transition in gamma-band activity was observed at the percolation threshold, indicating a response to large connected clusters.

Conclusions:

• Coherent neural oscillations in the retina encode global properties like size, which are not unambiguously represented by firing rates.
• Gamma-band oscillations offer a robust mechanism for downstream neurons to infer global stimulus features.
• The findings are consistent with reanalyzed data from cat retinal ganglion cells.