Related Experiment Videos
A metric for testing the accuracy of cross-modality image registration: validation and application
K J Black1, T O Videen, J S Perlmutter
1Department of Radiology Washington University School of Medicine, St. Louis, MO 63110-1093, USA.
Journal of Computer Assisted Tomography
|September 1, 1996
Summary
This study developed a new metric to assess image registration accuracy and found that automated image registration (AIR) of baboon brain MR and PET images resulted in a modest average error of 2.20 mm.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Primate Neuroscience
Background:
- Accurate image registration is crucial for multimodal neuroimaging studies.
- Existing methods may lack sensitivity for precise alignment, particularly between different imaging modalities.
Purpose of the Study:
- To develop and validate a sensitive metric for evaluating image registration accuracy across and within modalities.
- To quantify the accuracy of the Automated Image Registration (AIR) algorithm for aligning baboon brain MR and PET images.
Main Methods:
- Five baboons with implanted fiducial markers were imaged using MRI and PET (H2 15O blood flow).
- Anatomical MRI and PET blood flow images were aligned using the AIR algorithm, with fiducials used for validation.
- A novel metric was developed to assess registration accuracy based on fiducial point localization and distances.
Main Results:
- Fiducial localization was precise (approx. one-tenth voxel size), with distance accuracy within 0.2% (MRI) and 1.4% (PET).
- AIR resulted in a mean registration error of 2.20 mm (maximum 3.57 mm) within the baboon brain.
- Registration error was influenced by PET image smoothing, indicating sensitivity to image properties.
Conclusions:
- A reliable and sensitive metric for testing image registration techniques has been established.
- The AIR algorithm demonstrates modest accuracy for aligning baboon brain PET and MR images, with potential for improvement.