Eighteen patients (nine men, nine women; mean age, 61.4 years; age range, 24–82 years) with known pulmonary nodules underwent conventional whole-body PET/CT followed by an additional gated PET acquisition over the chest. Scanning was initiated approximately 60 minutes after IV injection of ¹⁸F-FDG (mean injected activity, 747 ± 56 [SD] MBq). PET/CT showed one well-circumscribed lung lesion in 14 patients, two lesions in three patients, and three lesions in one patient. All readings and diagnoses were performed using PET and CT data according to the standard protocols at our institution. Eleven patients had regional lymph node metastasis or metastases that involved from one to 16 lymph nodes (average number of lymph nodes involved ± standard error [SE], 4.27 ± 1.53). In six patients, pleural carcinomatosis was diagnosed, whereas only one patient had distant (nonregional) metastasis at a porta hepatis lymph node. All patients were scanned in the arms-up position to enhance image quality.

All patient data were collected on a PET/CT scanner (Discovery LS, GE Healthcare). For whole-body static PET, six or seven bed positions that were chosen to cover from the mid thigh to the eyes, which is sometimes also referred to as “torso PET,” were collected, each with an acquisition time of 5 minutes and an axial field of view of 15 cm. CT settings were 140 kVp, 120 mAs in the helical mode with a 10-mm collimation (4 × 2.5 mm), and a pitch of 1.5; no IV CT contrast agent was used. Subsequently, a single thoracic bed position was acquired in the gated mode, which we refer to as the “gated data,” in the area of the largest lesion with a PET acquisition time of 10 minutes to compensate for the lower number of counts per image frame.

Respiratory gating was performed using a respiratory-gating system (Varian Real-time Position Management, Varian Medical Systems). The system uses an infrared camera that is mounted on the PET/CT table and an infrared-reflective marker block that is fixed to each patient's upper abdomen so it is visible to the camera at all times and guarantees continuous respiratory motion tracking (Fig. 1). While activated, the system issues an end-expiratory trigger signal to start the gated data acquisition and to mark the beginning and end of the respiratory cycle, which was divided into eight equal periods of time for all patients except one (10 periods). The system works by distributing PET data into each of these bins according to the corresponding temporal phase; for a patient exhibiting rhythmic breathing, the temporal phase corresponds to the position of the thorax in the respiratory cycle. CT data were collected in the helical mode with free-breathing patients before PET acquisition and were used for intrinsic attenuation correction. None of the lesions was located in the vicinity of the diaphragm, which is an area of considerable attenuation-correction artifacts caused by breathing motion [11].

In addition to the extra 10-minute scanning duration for a gated PET acquisition, we estimate that an additional 10 minutes was required to set up the respiratory-gating equipment. Setup was simple; straightforward; and, except for the occasional need to move an obscuring piece of clothing or the need to reposition the reflector block, relatively trouble-free. The amount of time needed to inform the patient and obtain his or her consent varied but could be easily incorporated into the clinical routine without introducing additional delays.

Lesion size on CT was determined using software (Volume Viewer Plus, version 5.8.0, GE Healthcare) on a workstation (Advantage, GE Healthcare), and PET studies were viewed on a functional imaging workstation (Xeleris, GE Healthcare). If a patient had more than one lung nodule, the size of the largest lesion was determined and that lesion was used for all subsequent analyses, thus reducing the influence of partial volume effects [12]. For visual comparison and display (Figs. 2C and 2D), we created overlay images of all bins acquired from one patient using the Image Overlay function in Photo-Paint (version 11, Corel) with an opacity of 20% per layer for fusion images and the Image Addition function with an opacity of 100% per layer for maximum intensity projections. These images are mimicking the volume that the lesion moves during respiratory motion, but all information regarding signal intensity is lost in the creation process.

To determine lesion volume as seen by PET, the adaptive thresholding segmentation scheme previously described by Erdi et al. [13] was applied for each slice and each bin. The maximum standardized uptake value (SUV_max) was determined for the lung lesion; then, the SUV_max was set as the upper display level and 42% of SUV_max as the lower threshold. In this manner, only voxels with SUV intensities within this range were displayed and could easily be converted into a region of interest (ROI). For each slice individually, ROIs were created, and the size, SUV_max, minimum SUV (SUV_min), and average SUV (SUV_avg) of the ROIs were assessed. To test the effect of increased noise in the gated images, the ratio of SUV_max to SUV_avg in lung tissue just outside the actual lesion on three slices on which the lesion was present was averaged and then compared. We also examined noise levels by placing identical ROIs in the lung parenchyma on three slices each of ungated and gated images. Each lesion's axial dimension was calculated from the acquired 2D axial PET/CT data sets by multiplying the number of slices covering the lesion with the slice thickness of 4.25 mm.

Data are presented as means ± SE unless otherwise noted. Statistical significance was assessed by the two-tailed paired Student's t test and confirmed with the robust nonparametric Wilcoxon's signed rank test (two-tailed asymptotic): Differences with p values of less than 0.05 were considered significant and those with p values equal to or less than 0.01 were considered highly significant. Additionally, we calculated correlation values and linear regression of paired samples. All calculations were performed using statistics software (SPSS version 15.0, SPSS).

The respiratory-gating method used in our investigations made it possible to observe the movement of lung lesions during PET/CT and to visually evaluate the extent of respiratory motion (Fig. 2A, 2B, 2C, 2D, 2E, 2F, 2G, 2H). Decreased lesion volume in gated studies is not always visible and varies substantially among patients (Figs. 2G and 2H). However, when lesion size is assessed quantitatively, the use of respiratory gating leads to a statistically significant decrease in lesion size on PET studies. Both the average area covered by the lesion on one slice and the extent of the lesion in slices—hence, a lesion's dimension along the z-axis (from head to foot)—decreased highly significantly (p = 0.014 and p = 0.007, respectively) (Figs. 3A and 3B and Table 1). As a direct consequence, the total lesion volume as measured on PET was also significantly lowered, by 44.5%, when respiratory gating was applied (p = 0.025) (Fig. 3C and Table 1).

Our data revealed that the difference between lesion size assessments on PET and CT was significantly lower for gated PET (mean lesion volume ± SE, 16.78 ± 9.44 cm³) than for ungated PET (38.03 ± 15.93 cm³; p = 0.025) (Table 1). Measurements of the average lesion volume were highest on ungated PET studies (69.03 ± 22.36 cm³), significantly lower (p = 0.025) on gated PET studies (47.78 ± 14.53 cm³), and lowest on CT (31.00 ± 10.35 cm³). The difference between measurements based on gated PET and those based on CT was insignificant (Fig. 3C and Table 1). In three patients, the PET volume measurements were slightly lower than the CT volume measurements and showed a decrease in volume with respiratory gating.

With gated PET, our data showed highly significant increases in SUV_max and SUV_avg (both, p < 0.01) relative to ungated PET, and no significant difference in SUV_min (p = 0.484) (Figs. 4A and 4B and Table 1). Comparisons of SUVs from gated and ungated PET data showed a strong positive correlation (p < 0.001): r = 0.972 for SUV_max, r = 0.965 for SUV_avg, and r = 0.936 for SUV_min (Figs. 4C, 4D, 4E). Correlation between lesion volume assessed by CT and the difference in SUV assessed by gated and nongated PET showed no statistical significance (SUV_max, r = 0.034; SUV_avg, r = 0.192; SUV_min, r = 0.401).

The ratio of SUV_max to SUV_avg was similar, approximately 1.6, in both gated and ungated PET studies (gated, 1.599 ± 0.020; ungated, 1.610 ± 0.025; p = 0.45). When comparing SUV_max values outside the lesion, there was a slight but insignificant decrease in SUV_max in gated studies from 0.673 ± 0.045 (ungated) to 0.637 ± 0.029 (gated) (p = 0.51).