A comparison of the different 3D CT scanning modes on the GTV delineation for the solitary pulmonary lesion

The aim of radiation treatment is to increase the radiotherapy effects on the tumor while avoiding excessive toxicity [1, 2]. Computed tomography (CT) simulation and accurately determining the target margin were prerequisite for the successful oncology radiotherapy. Three-dimensional CT (3DCT) and four-dimensional CT (4DCT) simulation techniques are both in used, and 4DCT simulation is currently the leading simulation technology. It has enabled CT data acquisition to be correlated to the respiratory cycle(s), and allows a series of 3DCT data sets to be acquired at a number of points in a patient’s breathing cycle offering visualization of tumor motion on a patient by basis [3‐5]. However, the 3DCT simulation is still widely used for 3D conformal radiotherapy (3D-CRT) at present. There were two scanning modes in 3DCT simulation, which were axial and spiral scanning mode. Because the speed of tube rotation in axial mode was slower than that in spiral mode, we often treat head and neck tumors that are not affected by respiratory motion with spiral scan [6]. For thoracic and abdominal tumor, respiratory motion and heartbeat are important factors to determine the gross tumor volume (GTV). The slow speed scanning mode could collect more respiratory motion information, so we usually adopted axial scanning mode for simulation during free-breathing [7]. But whether the axial scan and spiral scan that are usually adopted would influence the GTV of the lung tumor in motion during free-breathing was unclear. Basing on 4DCT technique, this study evaluates how the axial and spiral scanning modes affect the position, volume and spatial match relationships on the GTV delineation for the solitary pulmonary lesion (SPL) in 3DCT simulation.

The average ± standard deviation of GTV_A and GTV_S volumes were 9.94 ± 9.74 cm³ (range 1.72 ~ 36.56 cm³) and 10.14 ± 10.25 cm³ (range 1.45 ~ 35.41 cm³), respectively. There was no statistically significant difference between the volume of GTV_A and GTV_S (p = 0.16) (Table 1). The volume of ITV_MIP was 14.33 ± 12.47 cm³ (rang 2.93 ~ 41.16 cm³). On pair wise analysis using Wilcoxon rank test, ITV_MIP was significantly larger than GTV_A and GTV_S (p _s = 0.000) (Table 1), and the ratios of ITV_MIP to GTV_A and GTV_S were 1.57 ± 0.54 and 1.66 ± 0.61, respectively.

The center of tumor coordinates for GTV_A, GTV_S and ITV_MIP were listed in Table 2 and the averages ± standard deviations were showed in Table 3. No differences were determined in x, y and z directions (p_x = 0.17, p_y = 0.40, p _z = 0.48). The distances from the center of GTV_A and GTV_S to the origin of coordinate were 36.94 ± 24.64 cm and 36.88 ± 24.39 cm, respectively. There are no difference between the two distances (p = 0.51).

MI between GTV_A and GTV_S was 0.41 ± 0.24; the variation range was from 0 to 0.89. The correlation between the MI and GTV_S was illustrated in Figure 2. There was a positive correlation between MI and GTV_S. The correlation coefficients was (r = 0.64 p = 0.002).

SBRT is a safe and effective alternative therapy for patients with the inoperable lung cancer or metastatic tumor [12]. Physiologic motions such as respiratory motion and heart beat would lead to tumor displacement and change in shape, so delineation of the GTV was a critical component in the radiotherapy process for thoracic tumor [2]. Investigators have determined the target margin of thoracic tumor using the technology of active breathing control and 4DCT, which could reduce the interference on radiotherapy and increase the irradiation dose in target margin [13‐16].

4DCT allowed for the clear delineation of the anatomical structures during free- breathing. This technique is widely used in evaluating the displacement of tumor and organ for thoracic and abdominal cancers. The ten sequence images reconstructed by 4D volume data could demonstrate movements of the tumor dynamically, and the ITV_MIP represented the volume where tumor was present at any time within the respiratory cycle, which contained motion information of the tumor [9, 17‐19]. The axial scanning mode was scanning alternates with the movement of scanner couch, that was, “scanning-in couch-scanning” pattern. As a result, the simulation time was long. Taking the ordinary 16 layers CT for example, the X -ray tube needs 1 s to rotate 360 degree and the scan cycle was 2.8 s by the axial mode, the scanning range was 24 mm (with 16 × 1.5 mm detector). The spiral mode was continuous scanning with CT scanner couch moving at a speed of 30 mm/s. There was no interval during the scanning process. It takes only 0.75 s to complete the scan rang of 24 mm in superior-inferior direction. In this study, we recorded and compared the scanning time for the same range by axial and spiral mode, respectively. The result revealed that axial mode took a longer time to complete the same scan than spiral mode, T_A = 28.93 ± 4.63 s, T_S = 8.81 ± 1.41 s (p = 0.000). Furthermore, the spiral mode could reduce motion artifacts, and increase the accuracy of tumor shape and position. Though the high speed of spiral mode, the scanning time for region of interest (ROI) was short for the small lesions and GTV contains little amount of respiratory motion information. The respiratory rate of the patients enrolled in the study was 16.00 ± 2.81breaths/min, and the respiratory cycle was 3.87 ± 0.73 s. The axial and spiral scans often complete ROI within 1 s. The key point was the ratio of the scanning time for ROI to the respiratory cycle. The higher the ratio was, the more motion information was included in the images. Among our subjects, there were 14 patients whose GTV_A was larger than GTV_S, while 7 smaller. There was no statistically significant difference in volume between the two scanning modes. We analyzed the tumor motion information included in GTV_A and GTV_S by comparing the volumes of ITV_MIP with GTV_A and GTV_S, respectively. The ratio between ITV_MIP and GTV_A was 1.57 ± 0.54 (p = 0.000). It suggested that the tumor motion information included in GTV_A was limited. The result was similar to the reports of Nakamura M [20, 21]. There was statistically significant difference in volume between ITV_MIP and GTV_S (p = 0.000). The ratio between ITV_MIP and GTV_S was 1.66 ± 0.61, which suggested that the motion information included in GTV_S by spiral mode was limited too. The result of this study revealed that there was no difference in the volume of GTV by axial or spiral mode in 3DCT simulation. The tumor was moving when the axial and spiral scanning was performed during free-breathing. The GTV_3D (GTV_A and GTV_S) created from the axial and spiral images included little respiration-induced motion information. It only represented the instant shape and volume that comes randomly by axial or spiral mode during the respiratory cycle.

ITV_MIP was significantly larger than GTV_A and GTV_S. The differences in volume between GTV_3D and ITV_MIP were caused by the tumor motion in three dimensions. The motion information was included in ITV_MIP. The motion amplitude in superior-inferior direction was larger than those in lateral and anterior-posterior directions. The superior-inferior motion of GTV centroid in lower lobe was larger than the in upper lobe [22, 23]. So the internal target volume (ITV) was generated from GTV_3D by adding different margins in x, y and z directions. The compensated volume was still larger than ITV_MIP created by 4D images.

The axial and spiral scan were all performed during free-breathing. The centroid coordinates of GTV_A and GTV_S were instant location during the respiratory cycle. There was no statistically significant difference in 3D coordinates of GTV_A and GTV_S, nor the distances from the centroidal positions to origin of coordinates. What’s more, there was no difference in the centroidal position between GTV_3D and ITV_MIP (p_x = 0.17, p_y = 0.40, p _z = 0.48). This suggests that there was no obvious impact on for the tumors’ central position by the three scanning modes during free-breathing.

Although there was no difference in volume and the centroidal coordinates between the axial and spiral modes during free-breathing, there was still respiration-induced change in shape of tumor [24, 25]. We conducted target registration on GTV_A and GTV_S, MI range from 0 to 0.89. If the two target regions coincide completely, MI was 1, otherwise, that was 0. There was one case with no coincidence in our data, which suggested an obvious tumor displacement, and/or the tumor shapes varied greatly. Furthermore, there was a linear correlation between MI and GTV_S (r = 0.64, p = 0.002). The small volume of GTV was one of the reasons for MI declination. The variation of MI was large in individuals. The low MI didn’t mean increase the margins, and different MI didn’t mean asymmetric margin either, because MI wasn’t determined by the displacement and the change in shape, it was affected by the tumor volume also.

In summary, our study found no impact on the volume and centroidal position of GTV by the axial or spiral scan in 3DCT simulation. There were no difference in the centroidal positions between 3DCT and 4DCT as well. The tumor motion information included in GTV_A and GTV_S was both limited. A linear correlation was found between MI and GTV_S. Relative to axial scanning mode, spiral scan was more timesaving, reduced some motion artifacts and increased the accuracy of GTV due to its high scanning speed. So the spiral mode was a reasonable option to replace axial scan in 3DCT simulation for SPL.