Visual depth encoding in populations of neurons with localized receptive fields

Area of Science:

• Neuroscience
• Computational Neuroscience
• Vision Science

Background:

• Stereopsis enables 3D perception from retinal image disparities.
• Linking neural correlates of binocular disparity to stereopsis qualities remains challenging.
• Bridging electrophysiology and psychophysics is crucial for understanding stereopsis.

Purpose of the Study:

• To simulate disparity-sensitive neurons in V1 using the energy-neuron model.
• To evaluate neural responses to various stimuli and relate them to psychophysical data.
• To compare phase-type and position-type neural models for disparity selectivity.

Main Methods:

• Simulated populations of V1 disparity-sensitive neurons with the energy-neuron model.
• Employed an efficient statistical estimator to analyze neural responses.
• Compared model performance against known psychophysical findings.

Main Results:

• Disparity representation in simulated populations was highly robust, enabling good depth discrimination with small neuron populations.
• Phase-type coding proved more robust than position-type, explaining zero-disparity tendencies and high-pass stimulus limitations.
• High variance in disparity representation for contrast-inverted stereograms suggests a mechanism for absent depth perception in large stimuli.

Conclusions:

• The energy-neuron model effectively simulates disparity representation and links to psychophysical observations.
• Phase-type neural coding offers a robust mechanism for stereopsis.
• Nonlocal interactions may disrupt depth perception in specific stimulus conditions, like large contrast-inverted stereograms.