Related Experiment Video
Updated: Apr 20, 2026
05:07
Author Spotlight: Optimized Lung MRI Protocol with Computationally Efficient Reconstruction Methods
Published on: September 6, 2024
917
PET attenuation correction using synthetic CT from ultrashort echo-time MR imaging
Snehashis Roy1, Wen-Tung Wang2, Aaron Carass3
1Center for Neuroscience and Regenerative Medicine, Henry Jackson Foundation, Bethesda, Maryland snehashis.roy@gmail.com.
Summary
A novel patch-matching method synthesizes CT images from MR scans, improving PET reconstruction accuracy. This approach enhances attenuation correction for PET/MR imaging, outperforming existing segmentation and registration techniques.
Area of Science:
- Medical Imaging
- Radiology
- Biophysics
Background:
- Integrated PET/MR systems are increasingly used in clinical and research settings.
- Accurate quantitative PET reconstruction relies on attenuation coefficient maps (μ maps) derived from electron density.
- A key challenge in PET/MR is the precise computation of μ maps, as MR imaging does not directly measure electron density.
Purpose of the Study:
- To develop a patch-based method for generating whole-head μ maps from ultrashort echo-time (UTE) MR imaging.
- To create synthetic CT images from UTE MR data without requiring image registration or segmentation.
- To evaluate the accuracy of PET reconstructions using these synthesized μ maps compared to traditional methods.
Main Methods:
- A patch-matching technique was employed using dual-echo UTE MR images and coregistered CT scans from a reference dataset.
- Patches from the reference CT were combined via a Bayesian framework to generate synthetic CT images for target UTE MR scans.
- No image registration or segmentation was necessary for μ map generation.
Main Results:
- PET reconstructions using the synthetic CT-based μ maps showed higher correlation (average ρ = 0.996) and minimal bias (regression slope, 0.990) compared to original CT-based PET.
- The synthetic CT approach yielded a peak signal-to-noise ratio of 35.98 dB in reconstructed PET activity, significantly higher than segmentation- and registration-based methods.
- The method demonstrated improved PET reconstruction accuracy, even in subjects with pathological findings.
Conclusions:
- A patch-matching approach effectively synthesizes CT images from UTE MR data, leading to significantly improved PET reconstruction accuracy.
- This method offers superior performance over segmentation- and registration-based techniques for attenuation correction in PET/MR.
- The developed technique provides a robust solution for accurate quantitative PET imaging in integrated PET/MR systems.
