A unified model of heading and path perception in primate MSTd

Area of Science:

• Neuroscience
• Computational Vision
• Perception

Background:

• Human navigation relies on perceiving both straight-line heading and path curvature.
• Existing models of heading perception in the dorsal medial superior temporal area (MSTd) do not fully explain the neural coding of curved paths.
• MSTd neurons often respond best to spiral patterns, not just radial expansion patterns associated with heading.

Purpose of the Study:

• To propose a computational model explaining how MSTd neurons can simultaneously encode path curvature and heading.
• To investigate the neural mechanisms underlying accurate human perception of curved trajectories.
• To propose a functional role for spiral-tuned cells in MSTd.

Main Methods:

• Developed a computational model of MSTd neural population activity.
• Simulated optic flow processing to analyze coding of curvature and heading.
• Compared model performance (curvature and heading errors) to human navigation behavior.

Main Results:

• The model demonstrates that a continuum of MSTd cell tuning (radial to circular) can simultaneously represent curvature and heading.
• Curvature is encoded by the spirality of the most active cell; heading by the receptive field center location.
• Model errors in curvature and heading estimation align with human performance.

Conclusions:

• MSTd's primary function may be trajectory estimation, not solely heading estimation, by simultaneously representing curvature and heading.
• Temporal dynamics in optic flow representation are crucial and challenge instantaneous vector field assumptions.
• Spiral motion patterns in optic flow are vital for complex trajectory guidance, with MSTd cells specialized for extracting this information.