This page shows the cell segmentation/tracking program for FUNseq pipeline.
Identifying a sparse subset of cancer cells from a large heterogeneous population, based on aggressive phenotypes (like invasive migration, multipolar divisions, or asymmetric lineage development) is challenging. Also, such aberration identification is critical, as cells exhibiting these characteristics are linearly correlated with poor prognosis. A high-throughput screening microscope has been developed in our group to capture these sparse, abnormal cancer cells. A program to identify these cells of interest in an accurate and fast fashion is essential. For this, we developed the mTGMM program, which tracks cells of high-throughput images, and determines the aggresive phenotypes.
Features
- Image pre-processing: We provide methods for adjusting quality for images with heterogeous fluorescence intensity profile between cells
- Parallel nuclei segmentation: We used agglomerative watershed for cell segmentation
- Tracking using Gaussian models: The intensity profile of a nucleus is modelled as a 2D Gaussian distribution. Nuclei tracking is done by forwarding every Gaussian from time point t to (t + 1) using Bayesian inference, with a priori knowledge that the position, shape, overall intensity of nuclei cannot change dramatically between two consecutive time points.
- Feature extraction: This is the post-analysis which we generate a feature table with records of cellular migration, division and intra-cellular intensity.
- Data visualization: We provide options to visualize the cell masks, migration trajectories, cell divisions and morphology detections.
Project Samples
