• Link and node hierarchies in the macaque interareal cortical network (accepted for the SfN 2023 conf.): 

Retrograde tracers injected into a cortical area allow us to estimate the density of neurons projecting axons into that area from all the other cortical areas. Injections in multiple areas result in a weighted directed network whose nodes are the cortical areas, the edges are the neuronal connections, and the axonal densities represent the weights. Retrograde tract-tracing data from the macaque and marmoset monkeys and mice reveal that interareal networks are dense, directed, and their connection weights are heterogeneous, with values covering several orders of magnitude. Discovering communities in such complex networks is a challenging problem since such networks are not suited for current state-of-the-art community detection algorithms used in network science. For example, the standard algorithms are based on similarity measures that cannot handle weights varying across multiple scales. For these reasons, we extended the link community algorithm to be applicable to the study of anatomical neural networks, introducing novel methods to handle their directionality, density, and heterogeneous weights. Furthermore, our algorithm identifies link and node hierarchies, allowing us to analyze the cortical network's structure at multiple scales. Using benchmark networks with known community structures, we demonstrate that the node hierarchy encodes the information of the ground-truth partition with high accuracy. Using the extended link community algorithm, we infer the macaque anatomical neural network's areal (nodal) hierarchy with 47 injected areas from an atlas of 106. We validated its statistical significance using appropriate null models (omega index). These statistical tests show that the interareal physical distances partially explain the brain's community structure and information, which is lost if axonal projections are randomly swapped. Finally, using a recently introduced hierarchical entropy measure, we find that node and link hierarchies from the macaque data have a broad distribution, implying that the system comprises multiple branches and subbranches, an unusual property compared to other networked systems. We compare the hierarchy within the visual system to the one obtained by Felleman and Van Essen, and others and show how the new hierarchy addresses some of the shortcomings of the previous hierarchies. Understanding the network connectivity patterns at a deeper level might provide valuable insights into understanding the information processing capabilities of the cortex.

🎥 Watch the video poster here.