Prostate cancer imaging with PET/MRI: A study on data-driven bulk patient motion detection and correction

Objectives: Prostate cancer (PCa) assessment, and particularly biopsy planning protocols used to target PCa lesions and aid ultrasound in the ensuing biopsy, has greatly benefited from PSMA PET/MRI. Oftentimes, however, final PET images produced in lengthy scans where bulk patient motion spontaneously occurred feature artefacts leading to reduced spatial resolution that could be to blame for some biopsy mismatches with, presumably, very small lesions. The objective of this study was thus to detect and correct for bulk patient motion in a prostate biopsy planning PET/MR protocol.

Methods: Fifteen PCa patients injected with 328±39 MBq of F-18-PSMA were scanned 15 min in a single bed position in the PET/MR scanner. Axial motion was detected in the single-slice rebinned sinograms using a technique based on centre-of-mass (CoM) tracking every 250 ms. After filtering the CoM signal for noise and respiration artefacts, axial shifts were detected and divided into rapid and gradual motion. The entire PET raw data were accordingly divided into motion frames, separately reconstructed and co-registered, thus generating a motion-corrected PET image. SUVmean was computed from the lesions before and after motion correction.

Results: Ten patient scans featured rapid or gradual movements. On average, 3.5±1.5 frames were considered for each patient, resulting in mean frame duration of 3.60 min (to ensure a sufficiently high SNR, 3 min was the minimum). A mean SUV change of 10% was found, going as high up as 19% in one patient scan. Maximal detected axial displacement was 1.3 mm.

Conclusions: Patient bulk motion is common in prostate PET examinations and can affect PET quantitation and final representation of lesion size and location. To better evaluate the full scope of bulk patient motion in prostate PET/MRI, further work including transversal motion assessment is in progress.