Fixational Eye Movements Enhance the Precision of Visual Information Transmitted by the Primate Retina.

The retina transmits visual signals to the brain in the spiking activity of retinal ganglion cells (RGCs). This signal is necessarily imperfect: some visual information is lost in phototransduction and retinal processing. To quantify the transmitted visual signal, we developed a Bayesian method to reconstruct images from the simultaneously recorded spikes of hundreds of macaque RGCs of the four dominant types. The algorithm combines a stochastic likelihood model for RGC light responses that is fitted to spiking data, with a prior model for natural images implicitly embedded within an artificial neural network trained for image denoising. When applied to retinal population responses to both flashed images and images jittered to emulate fixational eye movements, the method provided reconstruction performance exceeding or matching all previous reconstruction algorithms, in an interpretable analytical framework that provided insight into the neural code. Reconstructions improved with increasing jitter amplitude over a behaviorally relevant range (even when the jitter trajectory was unknown), revealing that fixational eye movements improve rather than degrade the retinal signal. Reconstructions were degraded by artificial perturbation of spike times as small as 5 ms, revealing a temporal encoding precision finer than expected from previous studies. Ablating cell-to-cell interactions in the encoding model substantially reduced reconstruction quality, indicating the importance of stimulus-evoked correlations in representing the visual scene. Thus, fixational eye movements contribute to highly precise retinal population activity, enabling more accurate transmission of visual signals to the brain.