Purpose: The use of magnetic resonance (MR) radiofrequency (RF) surface coils is a prerequisite for high-quality positron emission tomography (PET)/MR imaging. In lack of in-gantry transmission (TX) sources, the exact position of the RF coils is unknown in PET/MR, and may, therefore, lead to false attenuation correction (AC) of the emission (EM) data. The authors assess lesion and background quantification in AC-PET by mimicking different PET/MR imaging situations using a whole-body (WB) PET-only tomograph.
Methods: Phantom experiments were performed on a PET tomograph with 68 Ge-rod TX sources. First, a 15-cm plastic cylinder was filled uniformly with [18F]-FDG to simulate a head study. Second, a NEMA NU-2001 image quality phantom (35 x 25 x 25 cm3) was filled uniformly with [18F]-FDG to simulate torso imaging. The phantom contained six lesions (10-38 mm diameter, lesion-to-background ratio 6:1) centred around a 5 cm diameter lung insert. EM and TX measurements were acquired with and without MR head (cylinder) and surface (NU-2001 phantom) RF coils in place. The following imaging situations were mimicked in both head and torso phantom studies: (1) PET scan without MR coils in EM and TX for reference, (2) PET scan with coils in both EM and TX, and (3) PET scan with coils in EM but without coils in TX. Two more set-ups were performed for the torso phantom: (4) PET scan with coils in EM only and phantom shifted slightly compared to (3), and (5) PET scan with coils in EM and TX following local displacement of the surface coils. PET EM data (1)-(4) were corrected for attenuation and scatter using cold TX data. Imaging situations (1)-(3) were repeated with the cylinder phantom and head coil in a combined PET/MR prototype system employing template-based AC.
Results: Head phantom: In case the MR head coils were not accounted for during AC (3), central and peripheral background activity concentration was underestimated by 13%-19% when compared to the reference setup (1). The effects of MR coil omission during AC was replicated in the repeat study with the combined PET/MR prototype. Torso phantom: All lesions were equally visible on all AC-PET images. The effects of disregarding MR surface RF coils during AC [(3) vs (1)] were 4%, or less. A slightly higher bias was observed when accounting for the RF surface coils that were shifted between EM and TX (5). The effect of coil misalignment and neglect during AC on the quantification of the simulated lungs was insignificant compared to the noise levels in AC-PET.
Conclusions: Unaccounted attenuation from MR surface coils causes a regional bias of AC-PET data in body regions near the MR coils. Bias of central regions was more noticeable in smaller-size objects. In torso studies with body surface coils, the visibility of central lesions on PET was unaffected by MR coils following incomplete AC. Coil misalignment of several cm between emission and attenuation images causes an error that was comparable to that arising from unaccounted MR coil attenuation but small compared to the average standard deviation of the activity concentration levels.