PET-CT system
Three Discovery MI (GE Healthcare, Milwaukee, WI, USA) PET-CT systems were used for image acquisition. The systems were configured with four rings of detector blocks with lutetium yttrium oxyorthosilicate crystals (crystal size 4.0 × 5.3 × 25 mm3) coupled to an array of SiPM. The PET detector has a transaxial field of view of 70 cm, an axial field of view of 20 cm, and an overlap of 24% between multi-bed positions (per vendor recommendation). The sensitivity, according to NEMA standards, was 13 cps/kBq. The PET-CTs were cross-calibrated to the dose calibrator, and the calibration is validated monthly using an SUV control with phantoms. The PET system was combined with a 128-slice CT.
Patients and imaging
This study included 25 patients aged 18 years or more referred for clinical 18F-FDG PET-CT due to known or suspected malignancies at Skåne University Hospital between 30 April and 20 June 2018 (lung cancer n = 11, colorectal cancer n = 3, esophageal cancer n = 2, cholangiocarcinoma n = 2, breast cancer n = 1, malignant melanoma n = 1, testicular cancer n = 1, sarcoma n = 1, vulvar cancer n = 1, tonsil cancer n = 1, and giant cell carcinoma n = 1). In total, 40% (n = 10) of the patients were women. The mean (± SD) weight was 70 ± 11 kg (range 44–92 kg), the mean BMI was 23.6 ± 3.5 (range 16.5–29.7), the mean age was 59 ± 14 years (range 24–81 years), and the mean glucose level was 102.6 ± 21.6 mg/dL (range 75.6–172.8).
Imaging was performed 60 min after administration, and the patients were scanned from the inguinal region to the base of the skull. The mean administrated 18F-FDG was 4.0 ± 0.1 MBq/kg (range 3.8–4.3 MBq/kg), and the mean accumulation time was 63 ± 4 min (range 55–74 min). The PET images were reconstructed using the commercially available BSREM algorithm Q. Clear (GE Healthcare, Milwaukee, WI, USA) including the time-of-flight and point spread functions with a 256 × 256 matrix (pixel size 2.7 × 2.7 mm2, slice thickness 2.8 mm). The patients were examined with an acquisition time of 4 min/bed in list mode. Sinograms with acquisition times of 1, 1.5, 2, 3, and 4 min/bed were created from the list files, and these were reconstructed with seven different β values, 100, 200, 300, 400, 500, 600, and 700, thus yielding 35 image series per patient.
Due to unexpected uptakes visible on 1 min/bed images but not on 4 min/bed images (see the “Results” section) in a couple of the patients, the images from these patients were also reconstructed with a OSEM algorithm (4 iterations, 16 subsets, 2 mm post-filter) with time of flight (TOF), one with and one without point spread function (PSF). We also reconstructed the 4-min/bed position acquisition into four different 1-min series (0–1 min, 1–2 min, 2–3 min, and 3–4 min), using BSREM with β 500.
The activity and acquisition times are to a close approximation interchangeable, as long as the count rate is within the linear part of the noise-equivalent count rate curve, which is generally the case in clinical 18F-FDG studies [2]: 8 MBq/kg with an acquisition time of 1 min/bed is equivalent to 4 MBq/kg and 2 min/bed, assuming the same accumulation time between administration and scan time (1 h in this study). Therefore, going forward, we will use AT to refer to the product of the administered activity per unit body weight and the acquisition time (MBq/kg × min/bed), to emphasize that the results do not depend on the acquisition time alone but the combination of time and activity.
CT images were acquired for attenuation correction and anatomic correlation of the PET images. A diagnostic CT with intravenous and oral contrast (9 patients) or a low-dose CT without contrast (16 patients) was performed. In our clinical routine, a low-dose CT is performed if a previous diagnostic CT has been performed within 4 weeks. For diagnostic CTs, tube current modulation was applied by adjusting the tube current for each individual with a noise index of 42.25 and a tube voltage of 100 kV. The CT used for attenuation correction was acquired in a late venous phase. For low-dose CT, the tube voltage was 120 kV with a noise index of 45. The adaptive statistical iterative reconstruction technique (ASiR-V) was applied for all CT reconstructions.
This study was approved by the Regional Ethical Review Board (#2016/417) and was performed in accordance with the Declaration of Helsinki. All patients provided written informed consent.
Image analysis
Quantitative analysis
The noise level was calculated from regions of interest (ROIs) in the liver drawn on transaxial images using Hermes 2.0.0 (Hermes Medical Solutions, Stockholm, Sweden). Three 5-cm2 ROIs were drawn in subsequent transaxial slices with one image in-between, and these measurements were averaged. None of the ROIs were placed where liver metastases or large vessels were seen. The ROIs were drawn in one image series and copied to the other image series. The noise level was calculated as the ratio between the SUVstandard deviation (SD) and the SUVmean in the liver.
Two lesions per patient were selected and marked by an experienced nuclear medicine physician (see details below). SUVmax and SUVpeak (SUVmean in a 1-cm3 volume sphere) were calculated for all lesions.
Qualitative analysis
Ten image series were regarded as clinically interesting/relevant with reasonable acquisition times and reasonable image quality (1.5 and 1 min/bed position with β values of 300, 400, 500, 600, and 700) and were chosen for further evaluation. Acquisition times of ≥ 2.0 min/bed position were not evaluated since they were regarded as too long for most nuclear medicine departments. β values of 100–200 were not evaluated due to noisy images (see the “Results” section below).
Two lesions per patient were selected and marked by an experienced nuclear medicine physician in the image series with 4 min/bed position and a β value of 300, which based on visual analysis was regarded as the reference. The lesions chosen were approximately 1 cm in diameter and had slight to moderate hypermetabolism. Preferably, malignant lesions were selected, but if no malignant lesions fitting the criteria were found, reactive or physiological uptakes were chosen. Four nuclear medicine physicians reviewed the reference as well as the ten clinically relevant image series simultaneously and assessed the detectability of the lesions. The detectability was graded on a scale of 1–3, where 1 = lesion is visible, 2 = uncertain if lesion is visible, and 3 = lesion is not visible (compared with local background). The ten image series of clinical relevance were also evaluated for image quality. Four reviewers assessed which image series had acceptable overall image quality based on noise level, artifacts, contrast, and sharpness.
Statistical analysis
Continuous patient parameters are presented as mean ± SD and range and categorical variables as a percent (%). Noise level, SUVmax, and SUVpeak were tested for normality using the Shapiro-Wilks test. Since the variables were not normally distributed, all quantitative PET data are shown as a median ± interquartile range (IQR). Statistical significance was considered for p values less than 0.05, and all statistical tests were performed using IBM SPSS version 25 (IBM, Armonk, NY, USA). Noise level, SUVmax, and SUVpeak were all normalized to the reference (4 min/bed, β of 300) in the figures. Lesion detectability was calculated for all of the lesions and all of the observers (two lesions per patient, 25 patients, four observers = 200 assessed lesions in total) and expressed as a percent. The total number of images with acceptable image quality was calculated (one score per patient, 25 patients, four observers = 100 assessments of image quality in total) and expressed as a percent.