Preprocessing¶

Preprocessing is fundamental to provide standardized high quality images to the segmentation algorithm, and thus obtain successful segmentation.

In pyKNEEr preprocessing acts on:

Spatial image characteristics: Images are transformed to the same orientation (RAI, i.e. right, anterior, posterior), knee laterality (left), and origin (0,0,0)
Image intensities: In all images, the homogeneous magnetic field is corrected, intensities are rescaled to the fix range [0,100], and cartilage contours are enhanced [1]

Note

Standardization of spatial characteristics is recommended for all images
Standardization of image intensities is recommended only for images that are going to be segmented

Input: Image folder list¶

For the demo images, the input file is image_list_preprocessing.txt, which contains:

[1] ./original/
[2] 01/DESS/01
[3] right
[4] 01/DESS/02
[5] right
[6] 01/cubeQuant/01
[7] right
[8] 01/cubeQuant/02
[9] right
[10] 01/cubeQuant/03
[11] right
[12] 01/cubeQuant/04
[13] right

where:

Line 1: Path of the folder original, containing the image dicom folders
Even lines from 2 to 12: Folder names of the dicom stacks to preprocess
Odd lines from 3 to 13: Knee laterality

Tip

When using your own data:
- Create a folder original and add your dicom folders
- Customize image_list_preprocessing.txt with the paths of your own dicom folders
- Specify knee laterality using pyKNEEr coordinate system
- Tip: For images that are not directly involved in intersubject segmentation, you can run a separate notebook where you can set intensity_standardization = 0 to save computational time

Executing `preprocessing.ipynb`¶

To preprocess data:

Launch Jupyter notebook
In File Browser, navigate to preprocessing.ipynb, open it, and:
- Customize the input variables:
  - n_of_cores (How do I choose the number of cores?)
  - intensity_standardization (0 for only spatial preprocessing, 1 for spatial and intensity preprocessing)
- Follow the instructions in the notebook
Save your notebook at the end of the process

Output: Preprocessed images

The preprocessed images are in the folder ./preprocessed. For each dicom folder in the original folder, the outputs are:

*_orig.mha (e.g. 01_DESS_01_orig.mha): Images with the same intensities as the volume in the original dicom folder, but different orientation, knee laterality (if right), and image origin. These images can be used to compute relaxation maps
*_prep.mha (e.g. 01_DESS_01_orig.mha): Images with the same spacial characteristics as *_orig.mha but different intensities, because the constant magnetic field was corrected, the intensities were rescaled to a fix range, and the cartilage contours were enhanced. These images can be used to segment femoral knee cartilage
*_orig.txt (e.g. 01_DESS_01_orig.txt): Text files containing the header of the .dcm files. They can be used to extract acquisition information such as echo time, flip angle, etc.

Note

Both *_orig.mha and *_prep.mha are anonymized images, while *_orig.txt contains all the information of the dicom header (including subject name, etc.) if the dicom was not anonymized

Visualization: Original and preprocessed images¶

For a qualitative check, for each subject you can see a 2D slice of *_orig.mha and *_prep.mha, similarly to this one:

For images that were only spatially standardized, you will see only one 2D slice of *_orig.mha.

For 3D visualization, consider using a medical image software such as ITK-SNAP, which allows comparing images in the same coordinate system

References¶

[1] Shan L., Zach C., Charles C., Niethammer M. Automatic Atlas-Based Three-Label Cartilage Segmentation from MR Knee Images. Med Image Anal. Oct;18(7):1233-46. 2014.