Related Experiment Video
Updated: Jun 13, 2026
10:02
Whole-body PET/MRI of Pediatric Patients: The Details That Matter
Published on: December 19, 2017
Retrospective evaluation of PET-MRI registration algorithms
Zuyao Y Shan1, Sara J Mateja, Wilburn E Reddick
1Division of Translational Imaging Research, Department of Radiological Sciences, St. Jude Children's Research Hospital, 262 Danny Thomas Place, Memphis, TN 38105, USA. Zuyao.Shan@stjude.org
Journal of Digital Imaging
|May 4, 2010
Summary
This study assessed registration accuracy for positron emission tomography (PET) head images to MRI brain atlases. Linear and affine registration using ratio image uniformity (RIU) yielded the highest gray matter concordance.
Area of Science:
- Medical Imaging
- Neuroscience
- Radiology
Background:
- Accurate registration of positron emission tomography (PET) to MRI is crucial for multimodal brain imaging analysis.
- Standardized brain atlases aid in spatial normalization and comparison of neuroimaging data.
Purpose of the Study:
- To evaluate the accuracy of various registration algorithms for aligning PET head images to an MRI-based brain atlas.
- To compare the performance of different objective functions and transformation models in PET-MRI registration.
Main Methods:
- Utilized [(18)F]fluoro-2-deoxyglucose PET and MRI brain atlas data.
- Applied nine registration algorithms combining objective functions (RIU, NMI, CC) and transformation models (rigid, linear, affine, nonlinear).
- Assessed accuracy via visual inspection and quantified gray matter (GM) concordance.
Main Results:
- Linear and affine registration methods using ratio image uniformity (RIU) demonstrated the highest gray matter concordance (average similarity index of 0.71).
- Both linear and affine registration outperformed rigid and nonlinear transformations in terms of GM concordance.
- RIU objective function proved effective for accurate PET-MRI registration.
Conclusions:
- Linear and affine registration algorithms, particularly with the RIU objective function, are highly accurate for normalizing PET head images to MRI brain atlases.
- These findings support the use of specific registration strategies for improved precision in neuroimaging studies.
- The study provides valuable insights into optimizing image registration techniques for clinical and research applications.
