Evaluation of PET/CT alignment accuracy of the Discovery MI PET/CT according to the NEMA NU2-2018 standard

Objectives: Recent, evaluations of the GE Discovery MI (DMI) PET/CT, and its upgradeable digital ready (DR) version, have been performed using the NEMA NU-2 2012 standard (1-4). Newer performance standard, NEMA NU2-2018, while largely similar, now includes a PET/CT alignment accuracy test that was absent in these prior studies (5). This test is important because of PET’s expanding role in radiation oncology and interventional radiology, where co-registered PET/CT images are often used to guide radiation delivery or intervention. This study assesses the PET/CT alignment accuracy for three GE PET/CT scanners, two DMI and one DMI DR, and compares them to the NEMA and AAPM TG-126 tolerance guidance (6).

Methods: The NEMA NU2-2018 method was followed using two of GE’s VQC phantoms to provide transaxially distributed sources. Each phantom contained ten 0.7 MBq Ge-68 fiducial markers with 10 mm diameter. The PET scans (10-min total scanning time in 4 bed positions for the DMI scanners, and an 18.3-min 10-bed-position PET scan for the DMI DR scanner), were followed by a spiral CT scan acquired with and without 117.9 kg (260 lb) weight distributed on the scanner table as per NEMA. PET images were reconstructed without attenuation correction with time-of-flight (VPFX) using Q.Clear and OSEM (3 iteration ×16 subsets with a 5.0-mm transaxial post-filter and “standard” axial filter) in a 384×384 matrix with 1.32×1.32×2.17 mm and 1.8×1.8×3.3 mm voxels for the DMI and the DMI DR scanners respectively. The PET/CT co-registration error (CE) was determined as the distance between the centroids of the fiducial marker images in the PET and CT scans. The R_{max CT} and R_{max PET} ratios were calculated as described in NEMA by dividing the largest voxel value of each fiducial marker by the total counts along each of the 3 dimensions and taking the maximum of all 10 fiducial markers.

Results: The R_{max CT} and R_{max PET} of all 3 scanners were <0.3 and satisfied the NEMA requirement. Without weights (NW), the mean and max CEs were 2.7±0.6 mm and 3.3mm, 2.1±0.3 mm and 2.6 mm, and 0.9±0.3 mm and 1.2 mm for the two DMI and the DR scanners respectively. However, with the weights (WW), the mean and max CEs increased to 3.1±0.9 mm and 4.7 mm, 2.5±0.3 mm and 3.0 mm, and 2.9±0.6 mm and 3.7mm. Detailed analysis of the CEs in the patient lateral (X), anterior-posterior (Y) and superior-inferior (Z, scanner axial) direction for the three scanners will be presented (Table 1).

Conclusions: When loaded with a heavy patient equivalent, the measured max CEs are marginal to the ±1 PET voxel tolerance suggested by the AAPM TG126 report. Variability in the response of the scanners to weight is observed. This suggests that correcting for PET/CT misregistration may be necessary for particularly sensitive clinical applications for heavy patients.