cells-from-leaves

Introduction

This repository contains code to aid the annotation of cells in leafs with vectors from the centroid to a marker in the cell membrane.

The code segments leafs into cells, producing a number of cell images to check the quality of the segmentation and for identifying the marker of interest.

In order to make the analysis less biased the orientation of cell images are randomly rotated by 0, 90, 270 degrees. This, along with the fact that the cell images are cropped from the original leaf image, obfuscates the direction of the vector with regards to the context of the whole leaf.

The centroid of the cells are determined programatically from the segmentation and the user can use the cells-from-leaves-tagger to manually annotate the location of the marker of interest in the cell membrane. The manual step is also used to validate the correctness of the cell segmentation.

After manually annotating the segmented cells there is a post processing step for converting the annotations into a CSV file for further analysis.

Technology dependencies

This image analysis project has been setup to take advantage of a technology known as Docker.

This means that you will need to:

1. Download and install the Docker Toolbox
2. Build a docker image

Before you can run the image analysis in a docker container.

Build a Docker image

Before you can run your analysis you need to build your docker image. Once you have built the docker image you should not need to do this step again.

A docker image is basically a binary blob that contains all the dependencies required for the analysis scripts. In other words the docker image has got no relation to the types of images that we want to analyse, it is simply a technology that we use to make it easier to run the analysis scripts.

$ cd docker
$ bash build_docker_image.sh
$ cd ..

Run the image analysis in a Docker container

The image analysis will be run in a Docker container. The script run_docker_container.sh will drop you into an interactive Docker session.

$ bash run_docker_container.sh
[root@048bd4bd961c /]#

Now you can run the image analysis. Initially this will involve several rounds of finding suitable parameters.

For this specific data analysis the parameters found have been saved in the parameters directory.

Below is an example command.

[root@048bd4bd961c /]# python scripts/analysis.py --debug data/ output/
python scripts/analysis.py data/leaf.tif data/mask.tif parameters/params.yml output/ --debug

Post processing: manual point picking

Post process the data in the output/annotated-cells directory using the cells-from-leaves-tagger.

Post processing: generate summary data

Run the scripts/post_tagging_processing.py script.

Post processing: generate figures

Figures for the paper were generated using the Matlab scripts stored in the matlab_scripts directory.