This DBP repo includes materials for segmenting fibers segmented by migrating cells in images provided by Tom Gaborski's lab at RIT.
This repo contains two approaches to solving the central problem posed by this data: how to segment deposited fibers accurately.
These images are challenging to segment because:
- The fibers are heterogeneous in brightness
- Most tools are built to identify cells or other round-ish shapes. We need to be able to segment fibers, which form very irregular shapes
- We'd like one solution that can work well across multiple fiber types, which often look different from each other
The first approach (Initial_solution_CellProfiler) was created by Pearl Ryder and uses a CellProfiler pipeline to segment fibronectin images. The second approach (Revised_solution_CellProfiler_ilastik) was created by Melissa Gillis and Rebecca Senft and involves combining a CellProfiler pipeline with a machine learning model built in ilastik.
More details can be found for both approaches on their respective blogposts:
- Thinking like an Image Analyst by Pearl Ryder
- Customizing a Model for Fiber Segmentation (coming soon) by Melissa Gillis