Year: 2026
Type: preprint
Abstract: Advancements in spatial bioimaging analytics locate regions of interest as mask or vector geometries from raster images that are associated with molecular measurements, connectivity, and anatomical labels. A major challenge is to connect and jointly query these regions and their associated metadata at scale. The MicroData Model (muDM) is a spatial data model for associating 2D and 3D vector geometry with its annotated records, including identities, measurements, and hierarchies, as a precomputed multi-resolution pyramid. The data model has built-in integrations that emit target formats depending on use case, for consumption by machine learning and visualization applications. An open-source Python implementation is available on the Python Package Index, with a public catalog of over 100 datasets. In two Drosophila connectomes, the coarsest level retained at least 87% of finest-level neuron-typing accuracy with about 1% of the stored geometry. Linking synapse counts to the meshes improved accuracy beyond geometry alone.
Authors: Bengt Ljungquist, Nathan Hotaling, Nick Schaub
Language: English
Subfield: Biophysics
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7222955483