Related Experiment Video
Updated: Dec 17, 2025
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
49.1K
Accurate MR Image Registration to Anatomical Reference Space for Diffuse Glioma
Martin Visser1, Jan Petr2, Domenique M J Müller3
1Department of Radiology and Nuclear Medicine, Amsterdam UMC, Amsterdam, Netherlands.
Frontiers in Neuroscience
|June 26, 2020
Summary
Linear and non-linear transformations accurately register glioma locations in anatomical space. Lower-grade gliomas show better registration than glioblastomas, with similar accuracy for pre- and post-operative scans. Linear transformations are sufficient for summarizing glioma distribution.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Radiology
Background:
- Accurate glioma location summarization requires registering patient MR images to a standard anatomical space.
- Evaluating different registration methods is crucial for reliable glioma distribution analysis.
Purpose of the Study:
- To quantify the accuracy of MR image registration for gliomas using linear and non-linear transformations.
- To compare the performance of six common registration packages (FSL, SPM5, DARTEL, ANTs, Elastix, NiftyReg).
Main Methods:
- Collected pre- and post-operative T1-weighted MR images from 40 patients (20 glioblastoma, 20 lower-grade glioma).
- Used the Montreal Neurological Institute (MNI) brain template as the anatomical reference space.
- Quantified registration accuracy using Dice score, Hausdorff distance for tumors, and landmark distance for general brain alignment.
Main Results:
- Lower-grade gliomas registered more accurately than glioblastomas.
- Registration accuracy was consistent between pre- and post-operative images.
- SPM5 and DARTEL showed the highest tumor registration accuracy; FSL showed the least.
- Non-linear transformations improved general brain alignment but not tumor alignment compared to linear transformations.
Conclusions:
- Linear transformations are adequate for summarizing glioma locations in anatomical reference space.
- The choice of registration package significantly impacts accuracy, with SPM5 and DARTEL being superior for tumor registration.
