Evaluation of the Reproducibility of a Modified Aapm TG-126 Spatial Resolution and PET/CT Registration Test
Abstract
Purpose
To evaluate the reproducibility of a PET/CT QA test for assessing spatial resolution and image registration.
Methods
A modified AAPM TG-126 spatial resolution and registration test was performed using capillary tubes filled with a mixture of CT contrast and FDG (3.33 mCi/mL). Four tubes acting as point sources were affixed to a Jaszczak phantom: three parallel to the phantom axis (centered at isocenter, 10 cm x-offset, and 10 cm y-offset) and one perpendicular at the phantom center. The phantom was imaged on a Siemens Biograph TruePoint PET/CT using TrueX reconstruction (2 iterations, 14 subsets, 0.63-mm slice thickness). SUV-based line profiles were used to calculate FWHM values. Line profiles were generated in the sagittal and coronal directions for each parallel tube and an axial profile for the perpendicular tube. Misregistration was determined by comparing the slice with the peak SUV to the CT slice with the most contrast; slice separation multiplied by slice thickness yielded misregistration. Nine datasets were acquired and analyzed for the min, max and COV (coefficient of variation).
Results
Y-offset sagittal (4.78mm-9.99mm, COV 25.52%), y-offset coronal (4.71mm-9.16mm, COV 23.52%), center sagittal (4.60mm-9.34mm, COV 24.53%), center coronal (4.52mm-7.96mm, COV 22.52%), x-offset sagittal (4.58mm-10.12mm, COV 27.01%), x-offset coronal (4.07mm-9.98mm, COV 29.94%), perpendicular axial (4.32mm-8.94mm, COV 26.74%). This same analysis was repeated for image registration: Y-offset sagittal (0.0mm-1.89mm, COV 73.78%), y-offset coronal (0.0mm-1.89mm, COV 66.17%), center sagittal (0.63mm-1.89mm, COV 47.42%), center coronal (0.0mm-1.89mm, COV 86.6%), x-offset sagittal (0.0mm-3.15mm, COV 97.98%), x-offset coronal (0.0mm-3.78mm, COV 74.04%), perpendicular axial (0.0mm-1.26mm, COV 98.43%).
Conclusion
The coefficients of variation indicate poor reproducibility of the current methodology. COV for misregistration might be artificially high as small variations in slice misregistration reflects as large variation in misregistration distance. Further investigation is needed to assess sources of user introduced variability: point source dimensions, angulation and phantom positioning.