UTE-mDixon-based thorax synthetic CT generation.

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Biomedical Engineering

Background:

• Accurate photon attenuation assessment in thoracic MRI is challenging due to tissue heterogeneity and lung imaging difficulties.
• Existing approximations (water-equivalent, soft-tissue-only) can lead to significant errors in attenuation correction (AC).
• This is critical for PET/MRI attenuation correction and MR-only radiation treatment planning (RTP).

Purpose of the Study:

• To develop a method for voxel-wise thoracic synthetic CT (sCT) generation from MRI data.
• To enable accurate attenuation correction (AC) for PET/MRI.
• To support MR-only radiation treatment planning (RTP).

Main Methods:

• Optimized a radial stack-of-stars UTE-mDixon sequence for thoracic MRI acquisition.
• Extracted seven MR features (UTE, Echo1, Echo2, Dixon water, Dixon fat, Water fraction, R2*) for classification.
• Utilized fuzzy c-means for automatic 6-tissue classification (air, lung, fat, soft tissue, low-density bone, dense bone) and voxel-wise attenuation coefficient assignment.

Main Results:

• The UTE-mDixon sequence provided comparable image quality to traditional mDixon, with added lung and bone information.
• A combination of Dixon water, Dixon fat, and Water fraction proved optimal for robust 6-tissue classification.
• Generated thoracic sCT showed a mean absolute difference of <50 HU compared to reference CT, outperforming traditional Dixon-based methods.

Conclusions:

• Established a method for MR thoracic acquisition and analysis to automatically identify six distinct tissue types.
• Enables the generation of synthetic CT (sCT) for MR-based AC in PET/MR.
• Facilitates MR-only radiation treatment planning (RTP).