Modeling Physiological Sources of Heading Bias from Optic Flow

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Human heading perception relies on optic flow, with accuracy decreasing in the visual periphery.
• The dorsal medial superior temporal (MSTd) area is crucial for heading perception, receiving input from the medial temporal (MT) area.
• Existing models propose feedforward integration of MT motion signals to explain MSTd neuron heading sensitivity.

Purpose of the Study:

• To investigate how physiological properties of MT and MSTd neurons shape heading perception signals within a feedforward model.
• To determine which neural tuning characteristics optimize heading estimation accuracy and precision.

Main Methods:

• Simulated a feedforward computational model integrating MT and MSTd neural properties.
• Quantified the impact of varying physiological parameters (e.g., peripheral overrepresentation, receptive field size, direction tuning, noise tolerance) on heading signal accuracy and precision.

Main Results:

• Known physiological tuning characteristics generally supported heading estimation accuracy.
• A weak-to-moderate overrepresentation of peripheral headings in MSTd yielded the highest accuracy and precision.
• The model demonstrated robustness to noise in optic flow vectors when peripheral MSTd units were strongly direction tuned and had varied receptive field sizes.
• MT direction tuning biases toward the radial direction aided heading estimation, but MT speed and receptive field scaling with eccentricity did not.

Conclusions:

• Peripheral overrepresentation in MSTd is a key factor for accurate and precise heading perception.
• Specific neural properties, including receptive field size diversity and strong direction tuning, enhance heading estimation, particularly under noisy conditions.
• Understanding these physiological influences clarifies the neural basis of visual self-motion perception.