ITK 3D Image Registration

This repository contains python script, dataset, and results of image registration performed using ITK python library.

Dataset

T1-Weighted MRI and T2-Weighted MRI volumes from a same subject acquired as a part of Female data set of Visible Human Project.

Initially, there is small different in orientation between the two volumes.

Red = T1-Weighted MRI

Blue = T2-Weighted MRI

3D Image Registration

Registration is performed in order to discover the best transformation that aligns the regions of interest in the supplied images. Translation registration is carried out to register two MRI volumes using the registration methods of ITK python package. In this assignment, T2-Weighted MRI volume is registered to T1-Weighted MRI volume.

Fixed Image = T1-Weighted MRI
Moving Image = T2-Weighted MRI
Initial transformation = Translation transform
Optimizer = Gradient descent
Interpolator = Nearest neighbor
Similarity metric = Mutual information

Results

Red = Fixed Image

Blue = Resampled Moving Image

Impact of Image Registration - Comparision

The fixed, moving and moving resampled images are compared using 3D Slicer software tool.

Evalution

Image registration is validated by visual insepection and results of automatic registration feature of ITK-SNAP software.

First row, First column = Fixed Image
First row, Second column = Resampled moving Image by ITK-SNAP
Second row, First column = Overlaid fixed and moving image
Second row, Second column = Overlaid resampled moving image by itk and ITKSNAP