We are creating a concurrent data structure for 2D/3D points that supports efficient storage, range queries, and merging of disjoint datasets, motivated by highly-parallel algorithms for reconstructing neuron connections in the brain.
Connectomics is a rapidly growing field with a need for high-performance parallel algorithms for analyzing data from brain images and reconstructing the graph of connections between neurons. In order to facilitate storing and searching the large amount of spatial data, we are building a concurrent data structure that supports parallel merging of disjoint datasets, opening the door for connectomics pipelines that analyze the entire dataset in blocks (maximizing parallelism) before merging the local results into a shared output for global analysis.
