Abstract

This project aims to classify cell states (mitosis and interphase) using 2D time-lapse imaging data, integrating advanced segmentation and tracking techniques. The dataset consists of fluorescence-labeled (FUCCI) and DIC images acquired over 48 hours at 10-minute intervals, capturing cell behavior in multiple wells. While the primary goal is to identify cell cycle stages, challenges such as uneven fluorescence labeling, out-of-focus cells, and staining leakage necessitate robust preprocessing and analysis pipelines.

We used tools like CellProfiler, StarDist, and TrackMate for segmentation and tracking, revealing promising results despite observed limitations. A Fiji macro integrating StarDist for segmentation and TrackMate for tracking enabled cell trajectory generation and classification. Complementary approaches, including ilastik and Imaris, were considered for comparison. The project also highlights potential machine learning applications to improve classification accuracy using existing labeled datasets.