Evaluation of 2D and 3D UTE for PET/MR attenuation correction in the head

Objectives MR sequences capable of imaging tissues with short T2 relaxation times have been proposed in the past as a means to account for bone attenuation in hybrid PET/MR scanners. In this study we test two and three-dimensional Ultra-short Echo Time (UTE) sequences with the goal of defining an optimized acquisition protocol for MR-based attenuation correction.

Methods Data was acquired from ten healthy volunteer subjects using a tri-modality setup consisting of a GE Discovery 690 PET/CT in an adjacent room to a Discovery 750w MR. Scan times were restricted to between 2 and 4 minutes, compatible with typical PET acquisition times. The echo time was 30 μs for the 3D and 50 μs for the 2D sequence. Dual echo techniques were tested with a birdcage transmit/receive head coil and a 32-channel brain receive array coil. Matlab was used for background and skin removal, histogram-based segmentation and long TE tissue subtraction.

Results Both 2D and 3D acquisition led to usable bone maps with acquisition times compatible with integrated PET/MR imaging. Partial volume effects were found to be a source of false positive skull assignments, requiring an image resolution of below 2mm. This is further aggravated when using 3D acquisition with coils that have a long axial field-of-view. Artifacts were appreciated around the teeth, but these remained moderate, even in the presence of metallic implants. Undesired structures with very short relaxation times, such as cartilage (nose, ears) and connective tissue, could not be discerned from bone tissue. Suitable post-processing steps relying on anatomical landmarks must be implemented to prevent these structures from being incorrectly assigned in the attenuation map.

Conclusions Short echo time sequences are suitable for bone tissue detection, provided that sufficient scan time is available to prevent partial volume artifacts. Adequate post-processing is required to account for non-osseous short TE structures.