CT substitutes derived from MR images reconstructed with parallel imaging.

Area of Science:

• Medical Imaging
• Radiology
• Image Reconstruction

Background:

• Computed tomography (CT) substitute images from ultrashort echo time (UTE) MRI are valuable for PET attenuation correction and radiotherapy.
• Parallel imaging accelerates MRI acquisition by leveraging receiver coil sensitivities.

Purpose of the Study:

• To evaluate if non-Cartesian parallel imaging reconstruction can enhance CT substitutes generated from shorter MRI scan times.
• To assess the impact of subsampled radial k-space data on CT substitute quality.

Main Methods:

• Radial UTE and gradient echo (GE) data from 23 patients were reconstructed using gridding, SPIRiT, and CG-SENSE.
• Images were reconstructed from full and subsampled datasets (1000-10,000 spokes) simulating shorter acquisition times.
• Generated CT substitutes were compared against actual CT images.

Main Results:

• SPIRiT reconstruction slightly reduced the mean absolute deviation between CT substitute and CT numbers compared to gridding.
• CT substitute quality was largely unaffected by moderate subsampling (≥ 5000 spokes).
• Severe subsampling (≤ 3000 spokes) degraded quality, though SPIRiT showed a 25 Hounsfield unit improvement over gridding.

Conclusions:

• Non-Cartesian parallel imaging, specifically SPIRiT, offers marginal improvement in CT substitute quality over gridding.
• The increased reconstruction time for SPIRiT is not well-justified by the observed quality gains.
• CT substitutes can be generated from data acquired in under a minute using gridding without significant quality loss.