Integrating Terabyte-scale 3D Cell Detection
Cellfinder is a highly efficient 3D cell detection algorithm originally developed for the analysis of whole-brain microscopy datasets (e.g. light sheet microscopy in cleared samples). Unlike the majority of approaches that focus on computationally expensive cell segmentation, this novel approach combines classical computer vision with deep learning that allows extremely rapid analysis (e.g. 90 minutes for accurate 3D cell detection in ~200GB images). As mesoscale microscopy becomes routine, automated mapping of labelled cells across an entire organ will be key to leverage these new datasets. This team will improve upon the existing cellfinder-napari plugin and expand access to the broader community, extending into other fields acquiring comparably large datasets.
The project will increase ease of access for new users by removing data-specific barriers, providing real-time feedback on the functions of the algorithm, leveraging existing data and models, and identifying potential new applications of the software. These improvements will advance cell detection in large datasets and break down barriers for new users across fields.