Ventricle Segmentation

The left ventricle segmentation out of the coronary CT angiography (cCTA) images is needed for some research methods such as calculating myocardial blood flow (MBF). In this repo, I developed a pipeline to segment the left ventricle automatically using MONAI label and nnU_Net models on 3D slicer.

The project repository is as below:

Ventricle_Segmentation
|-Ventricle_Segmentation
  |-ImageTools
    |-ConvertVTK2NIFTI.py >>> takes a folder containing vtk/vti images and convert them into nifti.
    |-shuffle_images.py   >>> shuffles between images in a folder and takes a few number of them for manual segmentation
    |-VTI2VTK.py          >>> Takes a folder containing vti images and returns vtk images that are supported in Slicer.
|-Images
|-.gitignore
|-LICENSE
|-README.md >>> this file

The pipeline is available below:

1. Manual Segmentation and Image Processing

( to-do: delibrate the pipeline 1. converting vti images to vtu images and store them in specific folders 2. manually segment 20-30% of the vtk images using the slicer tools on 3D Slicer 3. export segmentation model from slicer and do a manual smoothing using MeshMixer 4. convert smoothed model to vtp. Use scripts to convert the model into a label-map segmentation. 5. save the images and labels using the data format and file system suggested by MONAI label. 6. label the rest of the datasets using either the pre-existing models in MONAI label or generate a model from scratch. 7. generate a NNU-Net model and train it based on the dataset segmented using MONAI Label. )

Use the tools in the ImageTools folder to convert the images into nifti for them to be supported on Slicer. Use the following commands on your terminal:

python ImageTools/ConvertVTK2NIFTI.py -InputFolder /path/to/the/folder/containing/vtk/images -Nformat nii or nii.gz

For manual segmentation shuffle the images and take some of them to apply manual segmentation on:

python ImageTools/shuffle_images.py -InputFolder /path/to/the/folder/containing/nifti/images -OutputFolder /path/to/the/output/folder

Use the manual segmentation tools on 3DSlicer to segment the left ventricle in the folder containing the selected shuffle images. for more information check out this video:

Note: refine the left ventricle segmentation slide by slide in each direction.

In order to save the segmentation in vtp format use the following steps:

1. on Data tab, export segmentations into a model.

2. Export the left ventricle segmentation to file:

3. save it in your desire format such as vtp or stl:

Use the same method to export your segmentation to binary label map and export that into file and finally save it as .nii.gz which is supported by MONAIlabel as segmentation format.