Related Experiment Video
Updated: Dec 2, 2025
10:46
MRI and PET in Mouse Models of Myocardial Infarction
Published on: December 19, 2013
12.1K
Effect of PET-MR Inconsistency in the Kernel Image Reconstruction Method
Daniel Deidda1, Nicolas Karakatsanis2, Philip M Robson3
1Biomedical Imaging Science Department, Leeds Institute of Cardiovascular and Metabolic Medicine (LICAMM), School of Medicine, and the Department of Statistics, School of Mathematics, University of Leeds, UK.
IEEE Transactions on Radiation and Plasma Medical Sciences
|November 2, 2020
Summary
Spatial inconsistency between PET and MR images significantly impacts positron emission tomography (PET) quantification. The hybrid kernelized expectation maximization (HKEM) method shows improved resilience to these PET-MR inconsistencies compared to KEM.
Area of Science:
- Medical Imaging
- Nuclear Medicine
- Machine Learning
Background:
- Anatomically-driven image reconstruction enhances PET resolution and quantification.
- Spatial inconsistency between MR and PET images is a critical challenge.
- Kernel methods, including hybrid kernelized expectation maximization (HKEM), are used for PET reconstruction.
Purpose of the Study:
- To investigate the impact of spatial inconsistency between MR and PET images on PET quantification.
- To evaluate the performance of the kernel method, specifically HKEM, in the presence of PET-MR spatial inconsistencies.
- To analyze the effects on hot and cold regions in PET images.
Main Methods:
- Applied the kernel method and HKEM to Jaszczak phantom and patient data from a Siemens mMR scanner.
- Investigated the effects of spatial shifts between MR and PET images.
- Quantified changes in activity concentration in hot and cold regions.
Main Results:
- Small spatial shifts between PET and MR images significantly alter activity concentration.
- PET-MR inconsistencies induce partial volume effects, causing 'spill-in' to cold regions and 'spill-out' from hot regions.
- HKEM demonstrated greater robustness, with maximum changes of 37% (cold) and 8% (hot), compared to KEM's 100% (cold) and 10% (hot).
Conclusions:
- Including PET information in the kernel improves reconstruction flexibility when dealing with spatial inconsistencies.
- Accurate registration and appropriate MR image selection are crucial for kernel creation to prevent artifacts, blurring, and bias.
- The study highlights the importance of addressing PET-MR spatial alignment for reliable quantitative PET imaging.
