The role of cortico-thalamic feedback for visual information processing

Abstract

In the early visual system, neurons in the dorsal lateral geniculate nucleus (dLGN) not only relay visual inputs from the retina to the primary visual cortex (V1), but also receive cortico-thalamic (CT) feedback inputs, which anatomically outnumber feedforward inputs. Most experimental and theoretical work so far has focused on the modulatory influence of CT feedback on relay cells in the dLGN. However, the role of feedback for the activity of V1 neurons, which are mainly driven by visual input meditated by dLGN neurons, is still unclear. To address this question, we devised a spiking neural network model of the thalamo-cortical and cortico-thalamic loop, where the CT feedback can be turned on and off to study its effect. We identified three main phenomena of the effect of CT feedback on V1 neurons: (i) orientation selectivity of cortical responses is increased for all contrasts, (ii) preferred orientations of input and output are better aligned with each other, (iii) neuronal population activity represents the stimulus orientation with higher confidence. Our analysis showed that CT feedback generally improves the signal-to-noise ratio (SNR) of the responses of dLGN neurons and therefore also increases the SNR of thalamic inputs to V1 neurons. Moreover, we found that the modulatory effects on visual processing depend on the geometry of the CT feedback projections. We conclude that CT feedback increases the selectivity of V1 neurons particularly at low contrasts by modulating the activity of dLGN neurons. Author summaryExtensive experimental and theoretical studies have been conducted to understand the functional role of cortico-thalamic (CT) feedback on relay cells in the dorsal lateral geniculate nucleus (dLGN) during visual processing. However, it remains unclear how CT feedback affects the neuronal responses of V1 neurons. To address this question, we developed a computational model of the thalamo-cortico-thalamic loop circuit based on biophysical parameters. Our results indicate that CT feedback modulates the orientation preference of V1 neurons in a contrast-dependent manner: The increase of orientation selectivity (OS) is stronger and the input preferred orientation (PO) is better preserved at lower contrasts. Our analysis suggests a mechanistic explanation: The CT feedback attenuates the untuned part of the thalamic input, while it amplifies the tuned part. This effectively increases the signal-to-noise ratio (SNR) in particular for weak contrasts. Our network model for the first time allows us to study the effect of feedback on cortical neurons and understand the impact of the number and the geometry of feedback connections.