Dosimetric study of a respiratory gating technique based on four-dimensional computed tomography in non-small-cell lung cancer

Abstract

This study sought to compare the differences in target volumes and dose distributions to the targets and organs at risk (OARs) between a four-dimensional computed tomography (4DCT)-based respiratory-gated intensity-modulated radiation therapy (IMRT) plan (Plan_EOE) and a three-dimensional CT (3DCT)-based IMRT plan (Plan_3D) in patients with non-small-cell lung cancer (NSCLC). For 17 patients with Stages I–III NSCLC, both 4DCT data and conventional 3DCT data were obtained. The Plan_3D and Plan_EOE were designed based on 3DCT data and 4DCT data, respectively. The displacements of the gross tumor volume (GTV) centroid were 0.13 ± 0.09 cm, 0.15 ± 0.1 cm, and 0.27 ± 0.27 cm in the right–left, anterior–posterior, and superior–inferior directions, respectively. The volume of the GTV_EOE was 3.05 ± 5.17 cm³ larger than that of the GTV_3D. The volume of the PTV_3D was 72.82 ± 48.65 cm³ larger than that of the PTV_EOE. There was no significant difference between the PTV_3D and PTV_EOE for V55.8, V60, V66 and the homogeneity index. The PTV_3D had a lower target conformity index than the PTV_EOE (P = 0.036). Plan_EOE had a significantly lower lung V10, V20, V30, V40 and mean lung dose (MLD) than Plan_3D. For the heart, Plan_EOE had a significantly lower V₃₀ and mean dose. In conclusion, 4DCT is an appropriate method for assessing the displacement of the GTV centroid in three dimensions. Plan_EOE has smaller PTVs and a decreased dose and volume for the normal lung and heart, as compared with Plan_3D.

INTRODUCTION

One of the challenges frequently encountered by radiation oncologists and medical physicists during a radiation treatment planning session is how to design an appropriate plan for a moving target. Traditionally, the margin created from the clinical target volume (CTV) to form the planning target volume (PTV) is determined by the physician's experience and literature report when there is no precise measurement available for evaluating the exact extent of the tumor motion. This margin includes various geometrical errors caused by tumor motion and daily patient set-up. However, the PTV margin is often too large or too small, since the extent of tumor motion in three dimensions is generally anisotropic. An increased risk of unnecessary irradiation to surrounding normal tissues may occur if the PTV margin is too large, which may result in severe side-effects associated with radiation therapy (RT). If the PTV margin is too small, however, a decreased local and/or regional control rate may occur as a consequence of the increased chance of missing targets.

Several researchers have investigated tumor motion and the impact of respiration-induced target motion on target volume for patients with non-small-cell lung cancer (NSCLC) by using four-dimensional computed tomography (4DCT) and found that the gross tumor volume (GTV) moved anisotropically with breathing in each spatial dimension. Treatment planning based on 4DCT created smaller target volumes, compared with treatment planning based on three-dimensional computed tomography (3DCT) [1–3]. In addition, the plan based on 4DCT resulted in a lower radiation dose to the surrounding normal tissues, particularly the lungs [2–4].

In this study, we aimed to evaluate the impact of respiratory motion on displacement of the GTV centroid and to compare the differences in the size of GTVs and PTVs, and dose distribution to the PTVs and critical structures between an intensity-modulated radiation therapy (IMRT) plan based on 3DCT and a respiratory-gating plan based on 4DCT for 17 inoperable patients with NSCLC.

MATERIALS AND METHODS

Patient characteristics

From March 2010 to August 2012, 17 patients with newly diagnosed NSCLC were included in this study. Each patient underwent both a 3DCT helical scan and a 4DCT helical scan using a 24-slice CT scanner (Somatom Sensation Open, Siemens Medical Solutions, Erlangen, Germany) and was considered inoperable due to medical or other reasons (Table 1). Written informed consent was obtained for all patients. The study was approved by the institutional review board (IRB).

Table 1.

Patient characteristics

Characteristics	Value (%)
Number of patients	17
Male/Female	14/3 (82.3%/17.7%)
Histology
Squamous cell carcinoma	6 (35.3%)
Adenocarcinoma	11 (64.7%)
Staging (AJCC 2002)
I	2 (11.75%)
II	2 (11.75%)
III	13 (76.5%)
Tumor location
Upper, middle lobe	10 (58.8%)
Lower lobe	7 (41.2%)

Immobilization and CT scans

Verbal coaching was required for all patients prior to, and during the 3DCT and 4DCT scans to achieve a regular and stable breathing pattern. Patients were positioned head first and supine in an individualized vacuum bag and were instructed to breath normally during the scans. The Respiratory Gating System AZ-733V (Anzai Medical, Tokyo, Japan) was used to monitor and coach the patient's breathing.

Non-contrast 3DCT imaging (effective mAs, 400 mAs; kV, 120 kV; rotation time, 0.5s; slice width, 4.0 mm; pitch, 1.2; reconstruction increment, 4.0 mm) was performed with a 4-mm slice thickness during free breathing from the cricoid cartilage to the lower edge of the liver. After that, intravenous contrast-enhanced 4DCT data with a 4-mm slice thickness were acquired (effective mAs, 400 mAs; kV, 120 kV; slice width, 4.0 mm; pitch, 0.1; reconstruction increment, 4.0 mm). For the 4DCT data, the acquired images were sorted by respiratory phases determined from the AZ-733V system to generate 8-phase 4DCT image datasets, namely 25% inhalation phase (25%_in), 50% inhalation phase (50%_in), 75% inhalation phase (75%_in), 100% inhalation phase (100%_in), 25% exhalation phase (25%_ex), 50% exhalation phase (50%_ex), 75% exhalation phase (75%_ex), and 100% exhalation phase (100%_ex). All 8-phase 4DCT datasets and the 3DCT datasets were imported to MiM Maestro V.5.6.1 (MIM Software Inc., Cleveland, OH) for contouring target volumes and organs at risk (OARs).

Target delineation and plan design

An appropriate window width and window level were selected for the lungs (window width, 1600 HU; window level, − 600 HU) and the mediastinum (window width, 400 HU; window level, 20 HU) to better view and determine the targets. The target delineation was in accordance with the International Commission on Radiation Units and Measurements Reports 50 and 62, as well as an institutional treatment protocol, and reviewed by a radiation oncologist (H. Lu).

The GTV was contoured on the 25%_in images and was then propagated automatically to the other 7-phase 4DCT image datasets through the VoxAlign Deformation Engine provided by the MiM Maestro software. Modification of the GTV was made whenever needed at the discretion of the treating physician. All GTVs created from the 8-phase images were used to evaluate the displacement of the GTV centroid in the three dimensions.

The GTV_3D was defined as the primary GTV contoured on the 3DCT datasets. It was then expanded with a 6–8 mm margin (squamous cell carcinoma, 6 mm; adenocarcinoma, 8 mm) to generate the CTV_3D. The PTV_3D was created by adding certain margins based on the tumor location to the CTV_3D (10 mm in the right–left and anterior–posterior directions; 10 mm in the superior–inferior direction for a tumor located in the upper and middle lobes; 15 mm in the superior–inferior direction for a tumor located in the lower lobe). A regional metastatic lymph node (GTV_n) was defined as any lymph node larger than 1 cm in the short axis in CT imaging study. It was expanded with a 5-mm margin to generate the CTV_n. The PTV for the lymph nodes was formed by adding a 10-mm margin to the CTV_n. The delineation of the lungs was performed automatically and didn't include the trachea, hilum or any GTVs. Other OARs including the heart, esophagus and spinal cord were contoured manually. All contours of the target volume and OARs were sent to the CMS-XiO planning system (CMS, St Louis, MO) to design the IMRT plan (Plan_3D).

The gating window was defined as three consecutive exhalation phases, namely 50%_ex, 75%_ex, and 100%_ex [2–5]. The 100%_ex images were used to design a treatment plan for the respiratory gating technique. A maximal intensity projection (MIP) dataset was created by assigning the highest density value for each pixel throughout the above three phase datasets and was used to define the GTV for the gating treatment plan (GTV_EOE). The GTV_EOE was expanded with an 8-mm margin for adenocarcinoma and a 6-mm margin for squamous cell carcinoma to generate the internal target volume (ITV), which is defined as a volume encompassing the CTV, taking into account the possibilities of the CTV variation in position, shape and size [6]. A 7-mm margin was added to the ITV to form the PTV_EOE. The criteria for determining the GTV_n and the CTV_n were the same as the above (except for the PTV_n, to which a 7-mm margin was addedz0. The OARs were contoured on the treatment planning images. All contours of the target volume and OARs were sent to the CMS-XiO planning system to design the gating treatment plan (Plan_EOE).

The prescribed radiation dose was 60 Gy in 30 fractions at 2 Gy per fraction delivered to the PTVs. The prescription criteria for the PTVs were as follows: 95% of the PTVs were covered by the prescription dose. Less than 10% of the PTVs received a dose >66 Gy, and less than 3% of the PTVs received a dose <55.8 Gy. A dose >66 Gy was not allowed to occur in any area outside the PTVs. Vx denoted the percentage of a specific structure exposed to a dose exceeding X Gy. Dose constraints for the OARs were as following: bilateral lungs, V₂₀ < 35%; mean lung dose, <20 Gy; heart, V₃₀ < 45%; mean heart dose, <26 Gy; esophagus, V₅₀ < 40%; mean esophagus dose, <34 Gy; spinal cord, maximum dose < 45 Gy.

Statistical analysis

The displacement of the GTV centroid in the right–left (RL), anterior–posterior (AP), and superior–inferior (SI) directions was calculated by using the MiM Maestro software, according three spatial coordinates of the 8-phase images. The size of the GTVs and the PTVs, radiation dose and volume of the targets and OARs were compared between the Plan_3D and the Plan_EOE. The mean ± standard deviation was used for data with a normal distribution, whereas the interquartile range and median were used for data not normally distributed. Analysis of variance (ANOVA) was performed to determine whether there were any significant differences in the displacements of the GTV centroid among the three dimensions. A paired sample T-test or Wilcoxon rank-sum test was chosen based on the data types. All P values were two-sided and considered significant at <0.05. The Statistical Package for Social Sciences, version 16.0 software (SPSS, Chicago, IL) was used for statistical analysis.

RESULTS

Displacement of the GTV centroid

The displacements of the GTV centroid in the RL, AP, and SI directions were 0.13 ± 0.09 cm, 0.15 ± 0.10 cm, and 0.27 ± 0.27 cm, respectively. Significant differences in the displacement of the GTV centroid were found between the SI and RL directions and between the SI and AP directions, with P-values of 0.023 and 0.048, respectively (Table 2).

Table 2.

Displacement of GTV centroid in three dimensions

Dimension	Mean ± SD	P-value
Left–right	0.13 ± 0.09	0.023*a
Anterior–posterior	0.15 ± 0.10	0.756b
Superior–inferior	0.27 ± 0.27	0.048*c

^aDifferences in the superior–inferior direction vs the left–right direction as analyzed by ANOVA.

^bDifferences in the anterior–posterior direction vs the left–right direction as analyzed by ANOVA.

^cDifferences in the superior–inferior direction vs the anterior–posterior direction as analyzed by ANOVA. *Significant difference.

Size of the GTVs and PTVs

The size of the GTV_EOE was 54.41 ± 47.96 cm³, 3.05 ± 5.17 cm³ larger than that of the GTV_3D, which accounted for 4 ± 8% of the total volume of the GTV_EOE. The size of the PTV_EOE was 314.41 ± 188.76 cm³, 72.82 ± 48.65 cm³ smaller than that of the PTV_3D, which accounted for 20 ± 9% of the total volume of the PTV_3D (Table 3).

Table 3.

Volumes of GTVs and PTVs

Volume	Mean ± SD	P-value
GTV3D (cm3)	51.36 ± 45.99	0.019*
GTVEOE (cm3)	54.41 ± 47.96
GTVEOE − GTV3D (cm3)	3.05 ± 5.17
(GTVEOE − GTV3D)/GTVEOE (%)	4.00 ± 8.00
PTV3D (cm3)	387.23 ± 227.92	<0.001*
PTVEOE (cm3)	314.41 ± 188.76
PTV3D − PTVEOE (cm3)	72.82 ± 48.65
(PTV3D − PTVEOE)/PTV3D (%)	20.00 ± 9.00

*Significant difference.

Radiation dose to the targets and OARs

No significant differences were found in homogeneity index (HI), V_55.8, V₆₀ or V₆₆ between the PTV_3D and the PTV_EOE, with P-values of 0.705, 0.093, 0.550 and 0.499, respectively. The PTV_3D had a better conformity index (CI), compared with the PTV_EOE (1.06 ± 0.08 vs 1.09 ± 0.06, P = 0.036). There were significantly lower V₁₀, V₂₀, V₃₀, V₄₀ and MLD for the bilateral lungs in the Plan_EOE, compared with the Plan_3D, with P-values of 0.047, 0.011, 0.013, 0.4006 and 0.013, respectively. However, no significant difference was found in V₅ for the lungs between the two plans (P = 0.178). For the heart, there was no significant difference in V₄₀ between the two plans (P = 0.161). However, Plan_EOE had a significantly lower V₃₀ and mean dose, compared with Plan_3D. The V₃₀ was 15.46 ± 12.15% in Plan_3D and 13.30 ± 10.34% in Plan_EOE (P = 0.025). The mean dose to Plan_3D was 10.39 ± 8.56 Gy, while in Plan_EOE this figure was dropped to 9.54 ± 7.94 Gy (P = 0.009). For the esophagus, no significant differences were found in V₄₅, V₅₀ or V₆₀ between the two plans. However, the mean dose to the esophagus was significantly lower in Plan_3D, compared with Plan_EOE [14.05 (10.78–29.43) vs 15.60 (11.06–28.02) (P = 0.025)]. There was no significant difference in the maximum dose to the spinal cord (Table 4).

Table 4.

Dose distributions to the PTV, lungs, heart, esophagus and spinal cord

Index	Plan3D	PlanEOE	P-value
PTV
V55.8 (%)	99.80 (99.37–99.92)	99.79 (99.32–99.97)	0.088
V60 (%)	95.41 (95.26–95.83)	95.57 (95.19–95.85)	0.550
V66 (%)	0.38 (0.07–1.32)	0.99 (0.06–2.55)	0.499
CI	1.06 ± 0.08	1.10 ± 0.05	0.036*
HI	1.09 ± 0.02	1.09 ± 0.02	0.705
Bilateral lungs
V5 (%)	50.61 ± 18.36	48.83 ± 17.61	0.178
V10 (%)	36.59 ± 14.25	34.89 ± 14.41	0.047*
V20 (%)	25.60 ± 12.67	24.13 ± 11.96	0.011*
V30 (%)	19.97 ± 10.36	18.48 ± 9.17	0.013*
V40 (%)	15.42 ± 7.69	13.93 ± 6.66	0.006*
MLD(Gy)	14.88 ± 5.88	14.03 ± 5.53	0.013*
Heart
V30 (%)	15.46 ± 12.15	13.30 ± 10.34	0.025*
V40 (%)	10.18 ± 7.45	8.84 ± 6.05	0.161
MD(Gy)	10.39 ± 8.56	9.54 ± 7.94	0.009*
Esophagus
V45 (%)	37.26 ± 29.33	35.01 ± 29.53	0.184
V50 (%)	34.16 ± 29.46	32.08 ± 29.31	0.193
V60 (%)	13.59 (3.93–44.62)	16.83 (12.87–31.82)	0.173
MD(Gy)	14.05 (10.78–29.43)	15.60 (11.06–28.02)	0.025*
Spinal cord
Maximum dose	39.59 (33.28–42.43)	38.29 (34.34–42.36)	0.523

CI = conformity index, HI = homogeneity index, MLD = mean lung dose, MD = mean dose. *Significant difference.

DISCUSSION

In lung cancer, respiration is considered as a major cause of target motion. It is also one of the biggest uncertainties during the whole treatment course [7]. An expanded target volume with a certain margin is often required such that adequate dose coverage can be achieved. However, this may increase the risk of treatment complications to the surrounding normal tissues.

The displacement of the GTV centroid was found to be different from one direction to another. Yu et al. [1] studied the tumor motions in 191 patients with NSCLC using 4DCT and found that the displacement of the primary disease was also associated with tumor stages. The largest motion range for early stage disease occurred in the RL direction, while the largest motion range for advanced stage disease occurred in the SI direction. Similarly, Britton et al. [8] found in locally advanced NSCLC the motion along the SI direction was significantly greater (P < 0.001), with mean ± SD values of 0.86 ± 0.19 cm, as compared with 0.39 ± 0.08 cm and 0.19 ± 0.05 cm in the AP and RL directions, respectively. These results were consistent with ours. In our study, the largest tumor motion was found in the SI direction, since 76.5% (13/17) of the patients had Stage III disease. However, patients in our study had a relatively smaller tumor motion in the SI direction, as compared with those in the other studies mentioned above. This may largely be attributed to the fact that fewer patients in our study had coexisting disease, like chronic obstructive pulmonary disease (COPD), thus their regular breathing patterns were less likely to be interrupted by compromised pulmonary functions. Nevertheless, images obtained from 4DCT are more likely to represent the real status of patients as compared with 3DCT, since the respiratory signals are incorporated with the spatial 3DCT imaging during the image acquisition and reconstruction process. Therefore, 4DCT images could be used to evaluate the magnitude of the GTV variations in the three dimensions in the design of a treatment plan.

Cover et al. [5] found that among CT datasets corresponding to different breathing phases, three consecutive phases (one end-exhalation and two adjacent phases) resulted in a reduced tumor motion. Residual tumor mobility in the three phases better correlated with residual mobility of the marker block than that of the diaphragm. This finding suggests that these three phases are suitable as a window for respiratory gating. In our study, 50%_ex, 75%_ex, and 100%_ex phases were chosen as the gating window. The MIP dataset was created by assigning the highest density value for each pixel throughout the above three phase datasets and was used to define the GTV_EOE. We found that the size of the GTV_EOE was significantly larger than that of the GTV_3D (P = 0.019). However, the size of the PTV_EOE was significantly smaller than that of the PTV_3D (P < 0.01). The disadvantage of the 3DCT-based IMRT plan was that the PTVs were too large. The size of the PTV_3D was increased by 72.82 ± 48.65 cm³, as compared with the PTV_EOE, which accounted for 20 ± 9% of the size of the PTV_3D. Underberg et al. [2] studied 31 consecutive patients with Stage I NSCLC undergoing a 4DCT scan, in which 3DCT datasets for 10-phase bins of the respiratory cycle were acquired during free breathing. Three PTVs were generated based on different criteria: PTV_10bins derived from an internal target volume (ITV) that incorporated all observed mobility, with a 3-mm isotropic setup margin; PTV_gating derived from an ITV generated from mobility observed in three consecutive phases during tidal-expiration plus a 3-mm isotropic margin; and PTV_10mm derived from the addition of a 10-mm isotropic margin to the most central gross tumor volume in the three bins selected for gating. They found that the PTV_10mm was the largest in all patients. Use of an individualized PTV_10bins resulted in an average volume of 48.2 ± 14.3% of the PTV_10mm. The PTV_gating was even smaller, with an average value of 33.3 ± 9.6% of the PTV_10mm. The mean PTV_gating was 70.5 ± 15.7% of the PTV_10bins. In another study conducted by Vlachaki et al. [3], 10 patients with thoracic malignancies who underwent gated and ungated CT scans were analyzed. It was found that the average PTV was 292.68 cm³ for the gated plans, and 575.17 cm³ for the non-gated plans. The average PTV generated from the non-gated plans was ∼1-fold bigger than that generated from the gated plans. These findings were consistent with our results. Plans based on respiratory gating have the potential to reduce the target volume and, as a result, to avoid excessive normal tissues being involved in the PTV when compared with plans based on non-gated techniques.

In our study, all plans were evaluated and were considered suitable for treating patients, although no patients were acturally treated with either Plan_EOE or Plan_3D. For the PTVs, the volume receiving <93% of the prescription dose was within 3%; the volume receiving 100% of the prescription dose was >95%; and the volume receiving more than 110% of the prescription dose was within 10%. The PTV_3D was superior to the PTV_EOE in terms of the CI. However, no significant difference was found in the HI between the two plans. Plan_EOE provided more advantages in organ preservation than Plan_3D. Significant reductions were found in Plan_EOE in the indices of V₁₀, V₂₀, V₃₀, V₄₀ and MLD for the combined lungs. Similar results were found in the heart, where lower V₃₀ and mean doses were observed in Plan_EOE. In Vlachaki's study [3], V₁₀ and V₂₀ were 26.26% and 30.96% for gated plans and 34.82% and 40.16% for non-gated plans, respectively (P < 0.0001). Gated plans resulted in lower mean lung, esophageal, and heart doses compared with non-gated plans (14.27, 17.28 and 10.86 Gy vs 19.5, 21.85 and 16.45 Gy, respectively; P ≤ 0.003). Underberg et al. [2] found that the mean volume of normal tissue that was encompassed by the 80% isodose line in the respiration-gated radiotherapy was only 39.1 ± 11.5% of that in the conventional radiotherapy, suggesting treatment plans based on gating technique have the potential for decreasing irradiated volumes of the normal tissues.

This study was initially designed to compare Plan_EOE and Plan_3D in a number of aspects. However, one may experience some dilemmas when doing this type of research, as there might be some inherent differences between the 3D and 4D techniques. For example, the scanning conditions may not be the same. Thus, the image quality and visibility of the tumor and normal structures between the two scans may be different, which may subsequently result in discrepancies in the target volumes and the size of the OARs between the two plans. In addition, the OARs for Plan_3D were outlined on the 3DCT images, which were actually the mixed-phase images, with different slices often being acquired from different respiratory phases. However, the OARs for Plan_EOE were outlined on the images in 100% exhalation phase. This may cause some differences in size of the two OARs. Although the contouring based on the 3DCT and 4DCT images in the present study is a widely accepted strategy, more research is needed to find an optimal contouring approach for treatment planning sessions.

Theoretically, a reduction in the size of PTVs could result in a decreased risk of radiation-induced pneumonitis. Reports from the literature have concluded that the incidence of radiation-induced pneumonitis is 11–14% in patients with NSCLC treated with IMRT, whereas this figure was usually <10% in those treated with stereotactic body radiotherapy (SBRT) [9–13]. This could be largely explained by the fact that SBRT has smaller PTVs, compared with the IMRT technique. In the present study, neither Plan_EOE nor Plan_3D was used to treat patients. The incidence differences of complications associated with RT (including radiation-induced pneumonitis) between the two techniques were not available. This was another limitation of our study.

CONCLUSION

In conclusion, 4DCT is an appropriate method for assessing the displacement of the GTV centroid in three dimensions. The respiratory-gated IMRT plans based on 4DCT have smaller PTVs and decreased dose and volume for the normal lung and heart, compared with the IMRT plans based on 3DCT. The dosimetric advantage was mainly attributed to the 4DCT technique.

FUNDING

This work was supported by a grant from the Sci-Tech Office of the Guangxi Zhuang Autonomous Region, China (2010GXNSFC013021).