Welcome to our CNN Annotator repository where you can find our recent Methods in Molecular Biology (MiMB) protocol on how to use a convolutional neural network (CNN) to classify time-lapse microscopy single-cell images patches according to live-cell chromatin morphology.
Wondering what such pipeline could be used for? Check out our recent publications where we made use of CNNs for classifying cells in live-cell imaging:
- Ulicna et al., bioRxiv, 2020 - Automated deep lineage tree analysis using a Bayesian single cell tracking approach
- Bove et al., MBoC, 2017 - Local cellular neighbourhood controls proliferation in cell competition
We provide a detailed walk-through for annotating live-cell microscopy images and training CNN model to infer classification labels on previously unseen images. Here is an overview of the entire process:
For more detailed instructions on how to annotate your microscopy data, train the CNN classifier and infer labels on previously unseen images, please refer to this step-wise manual.
Please use these links to proceed with the training and inference of your CNN models in the Google Colab environment:
| Notebook | Description | Link |
|---|---|---|
| Training | Train the CNN using annotated image patches |  |
| Inference | Use the trained CNN to perform predictions and clustering | |
Clone the repo locally and create a clean conda environment with all needed packages to run notebooks A & B on your local machine using the following commands:
git clone https://github.com/lowe-lab-ucl/cnn-annotator.git
cd cnn-annotator
conda env create -f ./environment.yml
conda activate cnn-annotatorHappy coding!
... Your CellX team