Search

The analysis of brain networks is central to neurobiological research. In this context the following tasks often arise: (1) understand the cellular composition of a reconstructed neural tissue volume to determine the nodes of the brain network; (2) quantify connectivity features statistically; and (3) compare these to predictions of mathematical models. We present a framework for interactive, visually supported accomplishment of these tasks. Its central component, the stratification matrix viewer, allows users to visualize the distribution of cellular and/or connectional properties of neurons at different levels of aggregation. We demonstrate its use in four case studies analyzing neural network data from the rat barrel cortex and human temporal cortex., Visual Analytics, Artificial Intelligence, Philipp Harth, Sumit Vohra, Daniel Udvary, Marcel Oberlaender, Hans-Christian Hege, and Daniel Baum

The mammalian neocortex is one of the most complex biological tissues and the center of higher brain functions. Currently, the specific distribution of neurons and neurites, as well as their intricate wiring within an entire neocortical area that emerge during development and are then refined throughout life, are not accessible. Here, I present a reverse engineered model of one neocortical area, the rat barrel cortex. First, I created a model of its structural composition constraint by measurements of cortex geometry, neuron distributions, and morphological reconstructions. This model provided anatomically realistic and robust estimates of the area's neuron and neurite distributions and captured the structural principles preserved across individuals. Second, I used the model's distribution of neurites to constrain synapse formation. Specifically, I introduced a stochastic synapse formation strategy that predicts the area's wiring diagrams if they were solely shaped by the area's structural composition in the absence of any learning or plasticity rules. I find that the predicted wiring diagrams are sparse, heterogeneous, correlated, and structured unlike random networks --- all of which are either observed or speculated properties of neocortical wiring. A systematic comparison between predicted and empirical wiring properties on the subcellular, cellular, and network level revealed a high degree of consistency. This demonstrates that the structural organization of the neuropil provides strong constraints for synapse formation. For the consistently predicted wiring properties, such as connection probabilities, it cannot be ruled out that they were shaped by the area's structural composition, i.e., implicitly by the developmental mechanisms that positioned neurons and neurites within the neuropil. A more sophisticated synapse formation strategy is not necessarily required. In contrast, such a sophisticated strategy might underlie the inconsistently predicted wiring properties, e.g., the frequency of certain circuit motifs. The herein presented approach can hence act as a starting point to identify wiring correlates of sensory experience or learning and provide a foundation to explore the relationship between synapse formation, an area's structural composition, and network architecture.