Related Experiment Video
Updated: Apr 7, 2026
Author Spotlight: An Efficient and Robust Software for Automated Fusion of Multiple Preclinical Imaging Modalities
Published on: October 27, 2023
Deformation field correction for spatial normalization of PET images.
Murat Bilgel1, Aaron Carass2, Susan M Resnick3
1Image Analysis and Communications Laboratory, Johns Hopkins University School of Engineering, Baltimore, MD, USA; Dept. of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Laboratory of Behavioral Neuroscience, National Institute on Aging, National Institutes of Health, Baltimore, MD, USA.
This study introduces a novel method for spatial normalization of positron emission tomography (PET) images. The technique enhances anatomical alignment by correcting deformable registration, improving accuracy for population studies.
Area of Science:
- Medical Imaging
- Neuroimaging
- Computational Anatomy
Background:
- Spatial normalization is crucial for analyzing positron emission tomography (PET) data in population studies.
- Current PET-to-PET registration methods rely on deformable registration algorithms designed for structural images, limiting accuracy.
- Improved anatomical alignment in PET imaging is needed for more reliable population-based analyses.
Purpose of the Study:
- To develop and evaluate a novel method for spatial normalization of PET images that enhances anatomical alignment compared to existing techniques.
- To improve the accuracy of PET image registration for population studies by leveraging structural image information during model training.
- To provide a more precise PET-to-PET registration method that does not require structural images during the application phase.
Main Methods:
- A novel approach was developed to correct deformable registration results for PET images.
- A generalized ridge regression model was trained using voxel-wise PET intensities and locations from training data with known structural image registrations (ground truth).
- The trained model corrects PET-to-PET registration without needing a structural image at the time of registration.
Main Results:
- The proposed method demonstrated superior anatomical alignment of PET images compared to standard deformable PET-to-PET registration.
- Cross-validation on 79 subjects confirmed the improved accuracy through visual inspection, reduced deformation field errors, and enhanced overlap with ground truth segmentations.
- The method successfully refines PET image registration, offering better alignment for population-level analysis.
Conclusions:
- The developed method significantly improves spatial normalization of PET images, offering enhanced anatomical alignment.
- This technique provides a more accurate and reliable approach to PET-to-PET registration for population studies.
- The findings suggest a valuable advancement in neuroimaging analysis, particularly for large-scale PET data studies.
