LOCUS – A PRECISE AND TRAINABLE 2D BRAIN MAPPING TOOL

Comprehensive anatomical data on brain circuits provide the foundation for understanding how neuronal circuits are organized both spatially and functionally. To uncover the complexity of neuronal circuits —ranging from local to brain-wide connections—it is essential to precisely identify their fundamental components and map them to specific regions across the entire mouse brain. However, reconstructing whole-brain datasets from serial sections is time-consuming and challenging because tissue-acquisition methods can result in sections with tilted angles, deformations, artifacts, or partial sections. Moreover, quantifying features such as neurons or synaptic puncta across sections requires extensive manual annotation. To address these challenges, we created LOCUS, an open-source, all-in-one software with a graphical user interface. It leverages machine learning to automatically assign entire and partial sections to the Allen Mouse Brain CCF, segment labeled features, and generate exportable quantitative 2D and 3D visualizations. The LOCUS plate assignment and registration strongly correlate with ratings from three anatomists, indicating it operates at a human expert level while being 100 times faster. During assignment and registration corrections, segmentation of features such as somata or synaptic puncta can be initiated in parallel. The segmentation accuracy closely aligns with features identified by experimenters, but is much faster. Lastly, at each step, the user can manually perform corrections. These refinements can be used to improve existing trained models, offering maximum flexibility to meet individual needs. LOCUS provides an intuitive and efficient method for mapping experimental 2D mouse brain sections into a 3D atlas, accelerating the analysis of anatomical mouse brain data.