MRI-Based Attenuation Correction for PET/MRI Using Multiphase Level-Set Method.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Artificial Intelligence in Medical Imaging

Background:

• Accurate attenuation correction (AC) is crucial for quantitative PET/MRI studies.
• Inaccuracies in MR-based AC (MR-AC) can lead to quantification errors and misinterpretation of brain lesions.
• Existing manufacturer MR-AC methods (MR-AC(mMR)) may underestimate tracer uptake.

Purpose of the Study:

• To develop and evaluate an improved MR-AC method for PET/MRI using an ultrashort echo time sequence.
• To assess a novel multiphase level-set algorithm with B0 inhomogeneity correction for MR-AC.
• To compare the performance of the novel level-set-based MR-AC (MR-AC(level)) against CT-AC and MR-AC(mMR).

Main Methods:

• A multiphase level-set algorithm segmented air, bone, and soft tissue from ultrashort echo time MR images.
• PET data from (18)F-FDG and (18)F-FP-CIT scans (PET/MRI and PET/CT) were reconstructed using CT-AC, MR-AC(mMR), and MR-AC(level).
• Quantitative accuracy was assessed by comparing SUV, SUVr, and specific-to-nonspecific binding ratios using region-of-interest and voxel-based analyses.

Main Results:

• The MR-AC(level) method demonstrated superior segmentation of air cavities and bones compared to MR-AC(mMR).
• MR-AC(level) provided PET quantification nearly equivalent to CT-AC, significantly reducing underestimation of tracer uptake observed with MR-AC(mMR).
• PET quantification error was reduced by a factor of 3 with MR-AC(level) (SUV error <10% or <5%) compared to MR-AC(mMR) (<30% or <15%).

Conclusions:

• The proposed multiphase level-set-based MR-AC method significantly enhances the quantitative accuracy of brain PET imaging within PET/MRI studies.
• This improved MR-AC technique offers a more reliable alternative to existing methods, reducing diagnostic errors.
• The findings support the clinical utility of advanced MR-AC techniques for precise PET/MRI quantification.