GulLiver

An image analysis package for whole slide liver immunostainings.

The goal is to segment and classify Sox9+ structures, as well as identify and classify veins. Output is quantification of distance between different types of veins, area covered by each vein type and number of each class of Sox9+ structure per portal vein.

Details on how the pipeline has been developed can be found here.

Installation

You can first install anaconda through the webpage.

Second you can start an Anaconda Prompt (by looking for it in the start menu) and create a new environment with the following:

conda create -n gulliver -c conda-forge python=3.11

After accepting and creating this environment, you will have to activate it and install GulLiver:

conda activate gulliver
pip install git+https://github.com/acorbat/gulliver.git

Usage

Once installed in a specific environment, you should always first activate the environment and then run gulliver.

conda activate gulliver
gulliver

Segmentation

The first module is in charge of processing scenes from CZI files and will output an OME Zarr file that contains all three channels, and labeled images corresponding to Sox9+ structures, liver regions, vessels or ducts and mesenchyma or not well stained areas.

You can either process a single scene from a single file.

gulliver segment path/to/file/czi --scene 0

Or you can point to a folder and gulliver will process every scene from every file.

gulliver segment path/to/folder

Another optional parameters is --chunk-multiplier which will select how big is the chunk to be sent to GPU in number of (1024, 1024) chunks. The default is 7 so $(71024, 71024)$ chunks will be used.

Quantification

Once images are segmented, everything is saved in OME-Zarr files. The quantify command receives a folder where the zarr images reside.

gulliver quantify path/to/folder

It will generate six Excel files where each tab will correspond to each image found in the folder.

portal_regions.xlsx contains all the Sox9+ structures detected in portal regions and information about their area (in $\mu ^2$), their class, to which portal vein they belong to and their labels in the segmented zarr image (see Visualization).

sox9_positive.xlsx contains all the Sox9+ structures detected, even if they are far away from portal veins, and information about their area (in $\mu ^2$), their class, to which portal vein they belong to (0 if the belong to no portal region) and their labels in the segmented zarr image (see Visualization).

portal_veins.xlsx contains all portal veins detected and information about their area (in $\mu ^2$), their distance to the nearest central vein (in $\mu$) and their labels in the segmented zarr image (see Visualization).

central_veins.xlsx contains all central veins detected and information about their area (in $\mu ^2$), their distance to the nearest portal vein (in $\mu$) and their labels in the segmented zarr image (see Visualization).

lumen_regions.xlsx contains all lumens detected irrespective of size and information about their area (in $\mu ^2$), their morphometric properties and their labels in the segmented zarr image (see Visualization).

liver_parts.xlsx contains all liver regions detected and information about their area (in $\mu ^2$), their morphometric properties and their labels in the segmented zarr image (see Visualization).

Visualization

To visualize results you start it activating the environment and running napari.

conda activate gulliver
napari

Once Napari starts, you can drag and drop .zarr files (they look like folders) into napari. If a prompt appears asking what reader to use, choose napari-ome-zarr.

Rationale

During Gulliver's first part of his trips, he arrives to Lilliput and finds himself a prisoner of a race of tiny people. Whole slide imaging looks similar as one is prisoner of an image filled with small details and not too much memory to load everything at the same time. Additionally, Gulliver can be separated into Gul and Liver, Gul in swedish means yellow and some of the diseases this pipeline was developed for end up in yellow livers.