Neuroimaging with Python

Overview

This project focuses on implementing fundamental Statistical Parametric Mapping (SPM) features in Python for neuroimaging data processing and analysis. By providing a collection of essential tools and techniques, including image registration, smoothing, normalization, and brain extraction, this project aims to make neuroimaging data processing more accessible and efficient for researchers and practitioners in the field. With an easy-to-follow structure and well-documented examples, this project serves as a valuable resource for those interested in exploring the potential of Python in the domain of neuroimaging.

Table of Contents

• Format
• Requirements
• Feature
  • Rotate DICOM Format
  • Display Dicom Format
  • DICOM to NIFTI
  • Image Registration
  • Smoothing
  • Normalization
  • Brain Extraction
  • Brain Resize
• Little SPM Manual
• References
• Tips

Format

• DICOM (.dcm) - Digital Imaging and Communications in Medicine
• NIfTI (.nii) - Neuroimaging Informatics Technology Initiative

Requirements

• SimpleITK - Open Source Insight Segmentation and Registration Toolkit
• Pydicom - Python package for working with DICOM files
• Nibabel - Successor of PyNIfTI
• dicom2nifti - Python library for converting dicom files to nifti
• OpenCV - Open Source Computer Vision Library
• NumPy

pip install -r requirements.txt

Feature

1. Rotate Dicom Format

here : rotation/rotate_image.py
Rotation Angle - Positive(+) value - ACW / Negative(-) value - CW

python rotation/rotateimage.py 'INPUT DIRECTORY' 'OUTPUT DIRECTORY' ROTATION-ANGLE

2. Display Dicom Format

here : display/display_dicom.ipynb
Run Jupyter Notebook and Change 'folder_path'
or

python display/display_dicom.py 'DICOM FOLDER PATH'

3. DICOM to NIFTI

for handling single sample

python dcm_to_nii_date.py 'SAMPLE_FOLDER'

for handling multiple samples

python dcms_to_nii_date.py 'DIRECTORY_OF_DICOM_SAMPLES'

4. Image Registration

here : image_registration/image_registration.ipynb
Input - Fixed Image & Moving Image (.nii, .mha, ...)
Output - Moved Image & Transformation File (.tfm)
or

python image_registration/image_registration_BS.py 'MOVING IMAGE' 'FIXED IMAGE' 

5. Smoothing

here : smoothing/smoothing.ipynb
Input - NIfTI Image (.nii)
Output - Smoothed Image (.nii)
or

python smoothing/smoothing.py 'INPUT NIfTI' 'OUTPUT NIfTI' SIGMA(optional)

6. Normalization

pip install intensity-normalization

Individual time-point normalization methods

# Z-score normalization   
zsocre-normalize --image IMAGE --output-dir OUTPUT_DIR ([--brain-mask BRAIN_MASK] [--single-img])
# Fuzzy C-means (FCM)-based tissue-based mean normalization  
fcm-normalize --image IMAGE --brain-mask BRAIN_MASK --tissue-mask TISSUE_MASK --output-dir OUTPUT_DIR ([--tissue-type {wm,gm,csf}] [--single-img])
# Gaussian Mixture Model (GMM)-based WM mean normalization   
gmm-normalize --image IMAGE --brain-mask BRAIN_MASK --output-dir OUTPUT_DIR ([--contrast {t1,t2,flair}] [--single-img])
# Kernel Density Estimate (KDE) WM mode normalization   
kde-normalize --image IMAGE --brain-mask BRAIN_MASK ([--output-dir OUTPUT_DIR] [--contrast {t1,t2,flair,md,largest,first,last}] [--single-img])
# WhiteStripe   
ws-normalize --img-dir IMG_DIR --mask-dir MASK_DIR --output-dir OUTPUT-DIR ([--contrast {t1,t2,flair,md}])

Sample-based normalization methods

# Least squares (LSQ) tissue mean normalization   
lsq-normalize --img-dir IMG_DIR --output-dir OUTPUT_DIR ([--mask-dir MASK_DIR])
# Piecewise Linear Histogram Matching (Nyúl & Udupa)   
nyul-normalize --img-dir IMG_DIR --output-dir OUTPUT_DIR ([--mask-dir MASK_DIR])
# RAVEL   
ravel-normalize --img-dir IMG_DIR --mask-dir MASK_DIR ([--output-dir OUTPUT_DIR] [--contrast {t1,t2,flair}])

7. Brain Extraction a.k.a. Skull-Stripping

ROBEX(Robust Brain Extraction) for Linux (Windows is not supported)

pip install pyrobex
robex path/to/t1w_image.nii -os path/to/stripped.nii -om path/to/mask.nii

import nibabel as nib
from pyrobex.robex import robex
image = nib.load('path/to/t1w_image.nii')
stripped, mask = robex(image)

Multi Stripping

python multi_brain_extraction.py path/to/original -os path/to/stripped -om path/to/mask

Little SPM Manual

Input Folder

• Dicom to Nifti
  python little_spm.py --rotate --directory <sample directory> --angle <degree>
ex) python little_spm.py --rotate -d 15819775_T1 -a 3

• Rotate Dicoms
  python little_spm.py --convert --directory <sample directory>
ex) python little_spm.py --convert -d 15819775_T1

Input Nifti

• Image Registration
  python little_spm.py --registration --input <dicom image> --template <dicom image> --iterations <numberOfIterations>
ex) python little_spm.py --registration -i 15819775.nii -t brain_atlas.nii

• Brain Smoothing
  python little_spm.py --smoothing --input <nifti file> --fwhm <fwhm>
ex) python little_spm.py --smoothing -i 15819775_T1.nii -f 8

• Brain Extraction (Only run in Linux)
  python little_spm.py --extract --input <nifti file>
ex) python little_spm.py --extract -i 15819775_T1.nii

• Normalization
  python little_spm.py --normalize --input <nifti file>
ex) python little_spm.py --normalize -i 15819775_T1.nii

• Resize
  python little_spm.py --resize --input <nifti file> -x <x> -y <y> -z <z>
ex) python little_spm.py --normalize -i 15819775_T1.nii -x 160 -y 190 -z 224

References

• Image Registraion
  • InsightSoftwareConsortium/SimpleITK-Notebooks
  • SuperElastix/SimpleElastix
  • voxelmorph/voxelmorph
• Brain Extraction
  • iitzco/deepbrain
• Normalization
  • ANTsX/ANTsPy
  • jcreinhold/intensity-normalization
  • sergivalverde/MRI_intensity_normalization

Tips

• Download Brain Templates and Atlases here (MNI Space)
• Download DICOM Viewer here (MicroDicom)
• Download NIfTI to DICOM here
• Download (X)MedCon - Open Source Toolkit for Medical Image Conversion
• NIfTI Viewer : https://socr.umich.edu/HTML5/BrainViewer/
• DICOM or NIfTI to JPEG or PNG

  pip install med2image
  med2image -i 'INPUT DICOM OR NIfTI' -d 'OUTPUT DIRECTORY`

• Brain Extraction Tools
  • HD-BET for Linux
    hd-bet -i INPUT_FOLDER -o OUTPUT_FOLDER