SA-PET/CT data acquisition and reconstruction parameters
SA-PET/CT examinations were performed on an Inveon SA-PET/CT (Siemens Medical Solutions, Knoxville, TN, USA). First, SA-CT images were acquired in approximately 10 min using 80 keV and 500 μA. Then PET acquisitions were performed using energy and coincidence timing windows of 350 to 650 keV and 3.4 ns, respectively. Emission scan duration was 20 min for phantom acquisitions, as recommended by the NEMA NU 4–2008 standards, and 15 min for animal scans. When the animal was imaged more than once, the duration of the subsequent acquisition was increased to account for 18F decay.
Reconstructions were performed using a three-dimensional maximum a posteriori (MAP) reconstruction with a 128 × 128 transaxial image matrix size. Three-dimensional ordered subset expectation maximization (OSEM-3D)/MAP was used with 2 OSEM-3D iterations and 18 MAP iterations with the β parameter set to 0.2. For NEMA NU 4 phantom studies, reconstructions were performed (a) with attenuation and scatter corrections, (b) with attenuation correction but without scatter correction, and (c) with neither attenuation nor scatter correction. For animal studies, data were corrected for attenuation and scatter events.
Phantom studies
Phantom studies were carried out with the NEMA NU 4–2008 image quality phantom. This phantom has the following features: a main fillable cylindrical chamber of 30-mm diameter and 30-mm length; a solid part with five fillable rods drilled through (at 7 mm from the center) with diameters of 1, 2, 3, 4, and 5 mm, respectively, and 20 mm in length; and a part with two cold cylindrical chambers 15 mm in length and 8 mm in diameter (one filled with non-radioactive water and the other with air). A more detailed description can be found elsewhere [17]. The image quality phantom was filled either with an 18F-FDG solution (diluted with pure water) or with an 18F-FDG solution containing iohexol at a concentration of 100 mg iodine (I)/mL, representing the highest concentration from our preclinical protocols.
Radioactivity at the beginning of the emission scan was 3.7 MBq ± 5%. The NEMA NU 4–2008 phantom was scanned twice for each situation.
Moreover, a homemade phantom was used to evaluate the impact of high Hounsfield densities on the accuracy of quantitative values for a pertinent-sized target. This phantom was designed to mimic tumors surrounded by water or intraperitoneal contrast media: small tubes (volume 2 mL, diameter 10 mm) were filled with an 18F-FDG-containing solution and placed at the center of a 20-mL syringe (diameter 18 mm) filled either with water or with a solution of iohexol (100 mg I/mL) inserted into a 60-mL syringe (diameter 27 mm). Syringes were consecutively scanned four times, with 18F-FDG concentrations ranging from 0.38 to 0.87 MBq/mL.
For each acquisition with contrast media, the phantom and vials were gently shaken immediately before the start of the CT scan in order to keep the iodine solution homogeneous. Preliminary studies (data not shown) have demonstrated that no sedimentation of iodine contrast media occurred within 45 min following the preparation of a phantom containing contrast material. Thus, it was not necessary to shake the phantom a second time after the CT had been performed, since overall, the acquisition time for a phantom SA-PET/CT acquisition was not more than 35 min.
Animal experiments
The regional ethics committee approved the experiments. A total of 16 mice and 6 rats were used. Four-week-old nude mice and nude rats were intraperitoneally injected with human ovarian cancer cell lines (SKOV-3 and OVR cell lines, purchased from American Type Culture Collection). Animals were kept under pathogen-free conditions and fed and watered ad libitum except on the day of SA-PET examination when a 6-h fasting period was maintained prior to the tracer injection.
Animals were injected with 18F-FDG in the tail vein under general anesthesia. Mean administered activity was 10 MBq for mice and 40 MBq for rats. Before, during, and after the 18F-FDG injections, animals were kept under an infrared light to minimize brown fat visualization. For general anesthesia, heated inhaled isoflurane was administered with an anesthesia device dedicated to small animals (Minerve, France). Mice were imaged in groups of four using a customized bed scanner. Technologic issues related to multiple mice imaging have been previously reported [18, 19] and have been recently discussed in detail and are beyond the scope of this article [20].
Figure 1 summarizes the animal experiments that were performed. In the first experiment, designed to study the impact of various contrast media protocols on the accuracy of quantitative values as compared to ex vivo counting, 11 animals were sacrificed immediately after the PET acquisition. In the second experiment, aimed at comparing the diagnostic performance and quantitative values of CEPET/CT and UEPET/CT acquisitions, 11 animals were sacrificed after the tracer uptake period to stop tracer uptake that may have occurred in between the two series of SA-PET/CT examinations.
Ex vivo counting
In the first experiment, immediately after the PET examination, tumors and organs were harvested. Tissue samples were weighed with a precision scale (±0.01 mg). Tumor radioactivity was counted for 2 min in a cylinder-well counter (Cobra II, Packard, GMI, Inc., MN, USA) and corrected for instrument efficiency and decay. Counts per minute were converted to becquerel and normalized for sample weight, assuming a density of 1 g/mL.
Calibration and cross calibration
The SA-PET/CT system was calibrated according to the manufacturer's guidelines, by imaging a 68Ge cylinder phantom. A cross calibration among the SA-PET/CT system, the dose calibrator, and the gamma counter was performed. A 10 MBq 18F-FDG solution (as assessed by the dose calibrator) was used to fill a vial with an exact known volume, which resulted in a solution with an exactly known concentration. This solution was used to fill a cylinder phantom that was scanned for 20 min on the SA-PET/CT scanner. Three samples of this solution (0.5 mL) were also drawn up with a calibrated pipette to be counted in the gamma counter. A large volume of interest (VOI) was used to determine the mean activity concentration as assessed by the SA-PET/CT scanner. Cross-calibration factors were then derived and used to synchronize counts/measurements for the three pieces of equipment.
Data analysis
Phantom studies
The following NEMA NU 4–2008 parameters were determined for each configuration and each reconstruction: (1) image noise was defined by the percentage standard deviation (%SDunif) in a VOI (22.5-mm diameter, 10-mm length) drawn over the center of the uniform region, and (2) spillover ratios in water (SORwat) and in air (SORair) were determined by the ratio of the mean activity concentration in the VOIs defined in each cold region (water and air, 4-mm diameter, encompassing the central 7.5 mm in length in the axial direction) divided by the mean activity concentration of the uniform area. Data were processed with AMIDE [21]. For acquisitions involving the homemade phantoms filled with either an 18F-FDG solution or a mixture of 18F-FDG plus iohexol, cylindrical VOI (7.5-mm diameter, 10-mm length) was drawn, and the ratios between true activity and measured activity were recorded.
Animal studies
In the first experiment, to evaluate the accuracy of quantitative values extracted from CEPET/CT, 3D VOIs were drawn over the tumors, thanks to an isocontour with a threshold set so that the VOI matched the apparent tumor volume on the CT component of PET/CT images. Even when a discordance occurred between PET metabolic volume and CT volume, the VOI was drawn according to CT images so that PET/CT images could be compared to ex vivo counting for which the entire tumor was harvested, irrespective of the presence of non-viable areas.
In the second experiment, the impact of the use of contrast media on tumor detection and accuracy of quantitative data was assessed. The CEPET/CT and UEPET/CT data were randomly interpreted 4 weeks apart by a senior researcher with 4 years of experience in SA-PET/CT and 7 years of experience in SA-PET and a junior researcher. Observers, who were blinded to the animals' status (i.e., presence or absence of tumors within the abdomen), rated each abdominal focus using a 5-point scale (1, definitely benign; 2, probably benign; 3, indeterminate; 4, probably malignant; 5, definitely malignant). Each rating was compared to the results of the necropsy. Diagnostic performance of CEPET/CT and UEPET/CT was compared by means of receiver operating characteristic analysis [22]. Then, PET quantitative values for UEPET/CT and CEPET/CT were compared. For that purpose, the two sets of images were displayed side by side, and 3D VOI was determined by means of an isocontour, which was set so that (a) the VOI matched the apparent tumor volume on PET images and (b) the tumor volume was equal on both sets of PET data.
Statistical analysis
The relationship between the radioactivity in animal tumors as determined by SA-PET/CT and by gamma counter was estimated using linear regression analysis. In addition to regression analysis, Bland-Altman plots were produced [23]. The same type of analysis was used to compare quantitative values extracted from UEPET/CT and CEPET/CT. Areas under the receiver operating characteristic curve for CEPET/CT versus UEPET/CT were compared according to the methodology of DeLong et al. [22].