An Unsupervised Approach To Detect Microglia Tip In Volumetric Fluorescence Imaging Data

Microglia play a key role in maintaining brain health, and the detection of microglia tips is essential for analyzing their motility. However, current tip detection methods either rely on deep neural networks, which require time-consuming annotations, or are unsupervised methods analyzing local patterns, which are significantly influenced by the microglia morphology change. In this paper, we propose an unsupervised, multi-scale microglia tip detection approach. Our approach measures the distance between the candidate tip and the convex hull of adjacent pixels to eliminate the influence of morphology variation. We demonstrated the new approach on volumetric fluorescence imaging data and achieved superior performance compared to peer algorithms.