Analysis of Evolving Cortical Neuronal Networks Using Visual Informatics

TL;DR

This study introduces a novel framework using Minimum-Distortion Embedding (MDE) to analyze evolving neuronal networks, outperforming traditional methods like PCA and t-SNE in capturing complex dynamics and connectivity changes over time.

Contribution

The paper develops and validates a new MDE-based approach for analyzing high-dimensional neuronal data, demonstrating its superiority over existing techniques in both simulated and real experiments.

Findings

MDE effectively preserves global and local data structures.

MDE captures developmental trajectories of neuronal networks.

Correlation metrics improve embedding quality.

Abstract

Understanding the nature of the changes exhibited by evolving neuronal dynamics from high-dimensional activity data is essential for advancing neuroscience, particularly in the study of neuronal network development and the pathophysiology of neurological disorders. This work examines how advanced dimensionality reduction techniques can efficiently summarize and enhance our understanding of the development of neuronal networks over time and in response to stimulation. We develop a framework based on the Minimum-Distortion Embedding (MDE) methods and demonstrate how MDE outperforms better known benchmarks based on Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) by effectively preserving both global structures and local relationships within complex neuronal datasets. Our \emph{in silico} experiments reveal MDE's capability to capture the evolving connectivity patterns of simulated neuronal networks, illustrating a clear trajectory tracking the simulated network development. Complementary \emph{in vitro} experiments further validate MDE's advantages, highlighting its ability to identify behavioral differences and connectivity changes in neuronal cultures over a 35-day observation period. Additionally, we explore the effects of stimulation on neuronal activity, providing valuable insights into the plasticity and learning mechanisms of neuronal networks. Our findings underscore the importance of metric selection in dimensionality reduction, showing that correlation metrics yield more meaningful embeddings compared to Euclidean distance. The implications of this research extend to various areas, including the potential development of therapeutic intervention strategies for neurological disorders, and the identification of distinct phases of neuronal activity for advancing cortical-based computing devices.