Respiratory motion compensated pulmonary imaging using simultaneous PET-MR

Objectives To explore the feasibility of using motion-compensated simultaneous PET-MR for improved pulmonary lesion detection.

Methods We have developed a maximum a posteriori (MAP) reconstruction framework for PET that uses a uniform quadratic penalty, incorporates MR-derived deformation fields that map multiple respiratory phases/gates to a reference gate, and computes a single reference gate image using all gated data. We have designed a radial FLASH pulse sequence for fast acquisition of high-quality gated volumetric MR images. List-mode PET data acquired is retrospectively binned into 8 gated sinograms using phase information from an MR navigator signal. Deformation fields are obtained using diffeomorphic demons based non-rigid registration of gated MR images (5 mm slice thickness, 24 coronal slices with 4096 radial lines and 256 samples per line, 3.3 ms TR per slice, 1kHz/pixel bandwidth, and an excitation angle of 30°). Attenuation correction is performed using dynamic MR-derived attenuation maps. We scanned a volunteer injected with 5 mCi [18F]FDG for preliminary evaluation of pulmonary PET/MR image quality achieved with our motion compensated MAP reconstruction approach.

Results Our results demonstrate the following: 1) Although the lungs appear relatively featureless in MR due to low proton density, blood vessels in our generated MR images suffice as landmarks for reliable computation of deformation fields by means of non-rigid registration. 2) By comparing uncorrected and motion-corrected PET images, we observe significant improvement in image quality due to the elimination of motion artifacts in the proximity of the diaphragm.

Conclusions Although standalone MR is not the traditional choice for lung scans, simultaneous PET-MR, unique in its capability of combining structural information from MR with functional information from PET, shows significant promise in pulmonary imaging.

Research Support This work was funded in part by NIH grants R21EB012326, R21CA149587, and R01CA165221.