Localization and Tracking in 4D Fluorescence Microscopy Imagery

Shahira Abousamra, Shai Adar, Natalie Elia, Roy Shilkrot; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2018, pp. 2290-2298

3D fluorescence microscopy continues to pose challenging tasks with more experiments leading to identifying new physiological patterns in cells' life cycle and activity. It then falls on the hands of biologists to annotate this imagery which is laborious and time-consuming, especially with noisy images and hard to see and track patterns. Modeling of automation tasks that can handle depth-varying light conditions and noise, and other challenges inherent in 3D fluorescence microscopy often becomes complex and requires high processing power and memory. This paper presents an efficient methodology for the localization, classification, and tracking in fluorescence microscopy imagery by taking advantage of time sequential images in 4D data. We show the application of our proposed method on the challenging task of localizing and tracking microtubule fibers' bridge formation during the cell division of zebrafish embryos where we achieve 98% accuracy and 0.94 F1- score.