Geometry of spiking patterns in early visual cortex: a topological data analytic approach.

Area of Science:

• Computational Neuroscience
• Neuroscience
• Topological Data Analysis

Background:

• Neural spiking patterns exist in high-dimensional spaces, posing challenges for understanding their intrinsic structure.
• Determining the geometry of these patterns is crucial for comprehending neural coding in the brain.

Purpose of the Study:

• To introduce a novel framework using topological data analysis (TDA) for analyzing spike train data.
• To determine the underlying geometry of spiking patterns in the primate visual cortex.

Main Methods:

• Developed a parametrized family of spike-timing-based distances to quantify neuronal response dissimilarity.
• Applied TDA to single-unit and multi-unit spiking activity recorded from macaque V1 and V2.
• Analyzed data across multiple timescales to reveal geometric properties.

Main Results:

• Identified a common geometry for spiking patterns in both V1 and V2.
• The inferred geometry is best modeled by low-dimensional Euclidean or hyperbolic spaces with modest curvature.
• The geometric structure is timescale-dependent, most evident at timescales relevant for encoding visual features like contrast and orientation.

Conclusions:

• Topological data analysis provides a powerful framework for uncovering the geometric structure of neural activity.
• The visual cortex exhibits a low-dimensional geometric organization of spiking patterns, suggesting efficient coding principles.
• Timescale-specific geometric properties are critical for understanding how the brain processes visual information.