A tool for automatic neurite outgrowth and cell viability estimation from microscopy images using deep learning and graph theory. It is optimized for Neuroblastoma cells in low power (20X) magnification and a dual FITC & DAPI channels setup. It can be used for large scale high-throughput drug screening and validation experiments.
- Estimate neurite outgrowth and toxicity for large scale experiments from microscopy images
- Outlier removal algorithms for cleaner results
- Graph representation of cell cultures and novel connectivity based features for neurite outgrowth
- Neurite semantic segmentation
- Nuclei instance segmentation from https://github.com/Lopezurrutia/DSB_2018
- Cell instance segmentation
- Cell foreground segmentation
- Two neurite segmentaion datasets for live and fixed cells staining.
-
Experiment_Demo.ipynb Is the easiest way to start analyzing experiment data (e.g. high throuput screening data from a 96 wells plate).
-
Computer_Vision_Pipeline_Demo.ipynb Displays the computer vision models in this repository and the different steps in the neurite outgrowth analysis.
-
(optional) To fully understand the computer vision pipeline, feature extraction, graph representation, novel connectivity features and outlier removal algorithms please refer to the Methods section in Thesis or inspect outlier_removal.py, feature_extraction.py, graph_representation.py, experiment_inference_utils.py.
-
To train your own neurite segmentation model you can use our neurite segmentation dataset which includes a live cells staining dataset and a fixed staining dataset.
-
Clone this repository
-
Install dependencies
pip install -r requirements.txt
- Download the Mask RCNN weights (too large for Github) and place them in repository root directory.
To be writen :)