Abstract
1201
Objectives: Although assessment of bone lesions with PET/CT is an important clinical application, the accuracy of PET reconstruction in bone has proved hard to assess. PET quantification accuracy is typically measured using water-filled phantoms that do not generally reflect the attenuation characteristics of bone. More elaborate phantoms including bone-equivalent materials do not allow for PET quantification in the bone regions. Here we present results from a series of phantom experiments that included positron-emitting radioactivity distributed within regions with bone-like attenuation properties.
Methods: A phantom with three identical cylindrical insets (25-mm diameter, 38-mm height, and 18.65-cc volume) was used in these experiments. One insert was filled with a solution representing cortical bone (full concentration), one with half the weight percent of the previous compartment (half concentration), and the third with water. The solution for the full concentration compartment had 149g of K2HPO4 dissolved in 100g of water (i.e., at the room temperature solubility), resulting in linear attenuation coefficients of 0.315 and 0.152 cm-1 at 100 and 511 keV, respectively. 18F was carefully added to each insert so that the differences in activity concentration between the two cylindrical inserts with the bone-like solutions were less than 1% different from the water-filled insert. Imaging was performed using two PET/CT systems (A and B) from different vendors, both accurately calibrated using conventional 18F water phantoms. On each scanner, a single PET acquisition was performed using our clinical whole-body protocol, followed by four low-dose CT acquisitions, each with different tube voltages, 80 kVp, 100 kVp, 120 kVp, and 140 kVp. Separate PET images were reconstructed using each of the four CT images, with attenuation and scatter correction enabled, using iterative reconstruction (OSEM with 21 subsets and 2 iterations). The experiment was repeated on three separate occasions (3 phantom preparations × 2 scanners × 4 reconstructions = 24 PET images). Cylindrical volumes of interest (VOIs) (15-mm diameter) were manually placed in each of the three insert regions. Activity concentrations in the VOIs were computed and values for the two compartments with bone-like attenuation were compared to that of the water region.
Results: At a 120 kVp tube voltage, the mean measured Hounsfield Units (HU) from the bone-like solutions were 1074.48 ± 32.91 and 697.69 ± 26.04 for the full- and half-concentration solutions, respectively. For scanner A and a tube voltage of 120 kVp, the average biases in the mean activity concentrations relative to the water compartment were +2.34 and -0.63% for the full- and half-concentration solutions, respectively. The corresponding biases for scanner B were -2.81% and -2.89%, respectively. A decreasing trend in bias was seen with increasing tube voltage, as shown in the table below.
Conclusions: The solution of K2HPO4 produced HUs that approximately matched the values observed in clinical low-dose CT images, suggesting the solution accurately simulated bone’s attenuation properties. The solution allowed for the addition of known amounts of radioactivity, making it possible to assess PET quantification in bone-like materials. PET bias in the simulated bone region was under 3% at 120 kVp CT, providing support for quantitative analysis of PET bone studies.