Impact of respiratory motion compensation in patients with solid tumors undergoing integrated Time of Flight PET/MR

Objectives: For chest and abdominal imaging, respiratory motion can create blurring and suboptimal quantification in PET imaging, leading to apparent increase of lesion size, reduction of measured SUV and poor conspicuity of small lesions. The high sensitive SIGNA PET-MR (GE) system allows for short PET acquisition time with preserved high image quality and calls for a reassessment of the clinical value of respiratory gating during routine protocols. We evaluated the impact of bellow-based respiratory compensation to mitigate motion both in MR and PET images during oncologic routine PET-MRI.

Methods: Forty patients underwent a FDG PET/MR (107±15 min after FDG injection, 4MBq/kg, 4-6 beds, 3-4 min/bed position, time-of-flight OSEM reconstruction, 2 iterations/28 subsets, no post reconstruction filter) for solid tumor staging. Respiratory signal was recorded using a respiratory bellow. PET data were reconstructed using 2 approaches: the quiescent phase gating (PET frame with around 50% of the counts; Q.Static), and gated data set registered to a single time frame using a non-rigid registration technique based on a motion field derived from the PET data (Q.Freeze). All lesions located in thorax and superior abdominal areas that were clearly distinct from adjacent lesions or surrounding background were included for analysis. Motion blur was analyzed visually, before and after correction, and using SUVpeak. Noise in reconstructed PET images was estimated by the SUV coefficient of variation CV in a right hepatic lobe ROI.

Results: 92 lesions were included in the analysis: 23 mediastinal lymph nodes (LN), 41 lung nodules and 28 abdominal lesions. Motion blur was observed in 30/92 lesions. Q.Freeze visually compensated for motion blur in 97% (29/30) of lesions, whereas Q.Static did in 27% (8/30). SUVpeak values increased significantly more in lesions with motion blur than in lesions without apparent motion blur using Q.Freeze, but not using Q.Static: +9.4% vs 3.5% (p=0.003) and +4.1% vs. 1.3% (p=0.5), respectively. Noise level was significantly higher in motion corrected PET images than in static images: CV [range] without correction, with Q.Freeze and with Q.Static were 0.21 [0.19; 0.25], 0.24 [0.22; 0.29], and 0.29 [0.26; 0.34], respectively (p<10^-4).

Conclusion: Q.Static and Q.Freeze are easy to implement during PET/MR routine protocols. Motion correction was more effective with Q.Freeze compared to Q.Static, with a median increase in SUVpeak of 10% in lesions presenting motion blurring (33% of our lesions). The impact of such SUVpeak enhancement when using PERCIST-based criteria for patient monitoring should be investigated. Research Support: ANR-11-IDEX-0003-02, France Life Imaging.