Delivered Dose Verification for Lung Cancer Stereotactic Body Radiotherapy Using Cone-Beam CT

Purpose: Verified the delivered dose distribution of lung cancer Stereotactic Body Radiotherapy (SBRT) using the cone-beam CT images. Methods: Five lung cancer patients underwent SBRT with 25 CBCT images were enrolled in this study. Delivered dose distributions were recalculated on CBCT images with deformed and nondeformed method, respectively. The planned and delivered dose distributions were compared using the dose-volume histograms. Results: The delivered target coverage (V100) per patient inside target volume deviated on average were 0.83% ± 0.86% and 1.38% ±1.40% for Pct vs. Pcbct and Pct vs. Pdcbct, respectively. The Conformity Index (CI) and Gradient Index (GI) showed a good agreement among the plans. For the critical organs, only minor differences were observed between the planned dose and delivered dose.


Introduction
Currently, Stereotactic Body Radiation Therapy (SBRT) has been widely used in the treatment of the stage I/II medical inoperable and surgically unresectable non-small-cell and metastatic lung cancer in recent years [1][2][3][4][5][6]. The treatment that allows delivery of very high dose of radiation in a few fractions to small regions (hypo-fractionated). Therefore, an accurate dose delivery is crucial for a successful SBRT. Due to the effect of breathing motion on small targets volumes is more severe, steep dose gradients and inhomogeneous densities in the thoracic region, accurate treatment delivery is not always guaranteed [7,8]. Hence, dose delivery verification is a prerequisite to assure correct treatment planning and delivery for SBRT treatment.
Over the past years, the kilovoltage Cone-Beam-Compute Tomography (CBCT) system attached to a linear accelerator has become commercially available for Image-Guide Radiotherapy (IGRT) [9]. The CBCT images are useful for verifying not only the position of a treated tumor, but also the regression or progression of the tumor. Optimal radiation therapy may be accomplished with regard to the field margin related to change in the target size and position, as well as the location of critical organs on every treatment day [10,11]. That is, a dose distribution that takes account of any modification of the targeted region is calculated at each radiation therapy session [12,13].
At the time of treatment, the patient anatomy is imaged with the current in-room capabilities. We have used the kv CBCT imaging currently available on our treatment machine. Recently, the Philips Pinnacle treatment planning system provides a dynamic planning module. It has two implementation methods; one is only copying all the structures from the original image to the second image; the other is copying the structures to the second image according the deformable registration. The delivered dose in the patient anatomy is then re-optimization in CBCT images using the Pinnacle dynamic planning module. This procedure provides 3D dose verification in the patient anatomy during a treatment fraction.
The aim of this study was to verify the delivered dose distribution of lung cancer Stereotactic Body Radiotherapy (SBRT) using the cone-beam CT images.

Materials and Methods
For this study, we retrospectively selected five non-small cell lung cancer patients treated in our clinic with SBRT, which every fraction the CBCT image can be used dose calculation. Patients were immobilized using Body Fix system (Elekta, Crawley, UK) to improve positioning reproducibility and to reduce the target motion with arms placed on their forehead. For these patients, a 4D CT image was scanned with a Philips Brilliant spiral CT (Philips Brilliant, Cleveland, OH) according to standard procedures with 1 mm slice spacing. The CBCT image was acquired before every fraction treatment using the Elekta Synergy which equipped the kv cone-beam CT.

Treatment planning
All 4D CT and CBCT datasets were transferred into a commercial treatment planning system (Pinnacle 9.8, Philips). The Gross Target Volume (GTV) was contoured by a senior radiotherapy oncologist using standardized lung window level setting. The amplitude of tumor motion was acquired from the 4D CT scan and incorporated as a patient-specific margin in the Internal Target Volume (ITV). The ITV to Planning Target Volume (PTV) margin was fixed in 3 mm.
Fifty-five full arc or partial arc Volumetric Modulated Arc Therapy (VMAT) plans were generated using the Pinnacle 9.8 Treatment Planning System (TPS) (Philips). It consisted five delivered plans on CT (Pct), 25 recalculated plans with deformable mapping (Pdcbct) and 25 plans non-deformable mapping on CBCT (Pcbct), respectively. For the Pct plans, the double full arc or partial arc plans were generated according the tumor position. For the sake of dosimetirc comparison, prescription was normalized to 50 Gy at 5 fractions at 6 MV for the all plans. The plans were optimized to reach clinically acceptable PTV coverage and Organ at Risk (OAR) sparing and normalized so that 95% of the PTV was covered by 100% of the prescription dose.

Dosimetric comparison
Several metrics were used for plan comparison. The target coverage (V100%) was defined the PTV volume receiving of Prescribed Dose (PD). The Conformity Index (CI) was defined as the ratio of the volume of the 100% isodose line (V100) to the target volume (PTV): CI=V100/PTV. The Gradient Index (GI) was defined as the ratio of the 50% isodose line (V50) volume to the target volume (PTV): GI=V50/PTV. For the health tissue, the V10 Gy, V15 Gy, V20 Gy of lungs, D2 of the cord, V20 Gy and V30 Gy of chest wall were evaluated compared among the Pct, Pdcbct and Pcbct plan.

Statistic
Results were described as mean ± Standard Deviation (SD). Comparisons among the plans were analyzed with one-way ANOVA method. All statistical analysis was conducted with R program software. Difference was considered statistically significant when p<0.05. Table 1 presents the characteristic of the enrolled 5 patients involved. There were 1 woman and 4-man patients with a median age of 71 (range from 67 to 79) years old. Total of 55 plans were generated for these patients. Figure 1 complies the mean volume of PTV for the 5 patients among the Pct, Pcbct, Pdcbct. Table 2 shows the detailed dosimetric comparison among the Pct, Pcbct, Pdcbct. The target coverage (V100) per patient inside target volume deviated on average were 0.83% ± 0.86% and 1.38% ± 1.40% for Pct vs. Pcbct and Pct vs. Pdcbct, respectively. One-way ANOVA analysis found only one patient have a significance difference among the three Pct, Pcbct, Pdcbct (p value <0.05). The Conformity Index (CI) and Gradient Index (GI) showed a good agreement between the planned and delivered dose distribution for patients without changes in anatomy. The one-way ANOVA results showed there have no significance difference; the p value was 0.73 ± 0.09 and 0.58 ± 0.22 for CI and GI, respectively. Table 3 lists the OARs protection comparison among the three planning modalities. In this table, we observed all the metrics have a significant difference in the patient 1. In others patients, only a few of the metrics have a significant difference.

Discussion
Treatment verification using CBCT images based on information acquired in the treatment room is feasible and provides an independent verification for lung cancer SBRT patients. In this study, we retrospectively investigated the delivery dose effect of without deformable and deformable planning on five lung SBRT patients. We have analyzed the five patient cases and fifty-five plans included 5 Pct, 25 Pdcbct, 25 Pcbct plans. We found the volume of PTV in the three modalities is essentially the same. This was expected, because for these patients only small changes in anatomy were observed compared to the planning CT image. For the PTV metrics, we just found the V100 have a significant difference in one patient, the CI and GI we didn't find any significant difference for the patients. It shows that the treatment plan in most patients can get an exact delivery.
There have several advantages for dose verification using the information acquired prior to a treatment session. First, at the first day of treatment the patient anatomy is scanned with an in-room image-guided radiotherapy technique: in this study KV cone-beam  CT imaging. The treatment planning CT (4D) image acquired several days or weeks prior to the first fraction. The patient anatomy may occur some changes in this period. If these possible changes really happen, it is visible in the CBCT image acquired for the IGRT procedure used for patient set-up. In the Table 1, we can find the changes are very small. In our institution, only a 3D imaging technique (3D CBCT) is available. The results may be improved with a 4D CBTC technique, not only for the anatomy changes, but also for the possible changes in the breathing motion [14].
Second, we generated the plans used the same beam parameter, optimization parameter on the fraction CBCT with the dynamic module of the pinnacle TPS. It has two ways to implementation, one is put the structures from the original CT images to the CBCT images without any deformable, the other put the structures from the original CT images to the CBCT images depend on the deformable registration, it also called adaptive radiotherapy. In these three modalities plan, the contours only delineated in the original CT and the beam and optimization parameters were the same, therefore the   difference of the manufactured could be decreased as far as possible.
Feasibility of CBCT to calculation dose has been previously investigated. Yoo et al. [13] studied the Hounsfield Unit (HU) value difference between CT and CBCT images for the Catphan phantoms (The Phantom Laboratory, NY). Yang et al. [12] evaluated the dose difference between CT plan and CBCT plan; it found the dose difference between the two was within 2%. However, due to the respiratory motion, the dose differences can be higher (3%) in the lungs. Both studies suggest that CBCT can be employed directly in dose calculation.
In this study, the target coverage (V100%) (p<0.01) and HI (p=0.01) had a significance difference in the patient 1, and not observed any significance difference in others patients. From the Table 1, we found the volume of PTV was higher than the other four patients. Qin et al. [15] studied the target size change has been shown an important metric determining the dosimetric effects of adaptive planning for lung SBRT, and the patients with small target size changes are less likely to profit from adaptive planning due to anticipated small dosimetric changes. The CI and GI were not found any significance difference for all the patients.
The 3D dose verification is to add up the dose distributions of the various fractions. For this purpose, the non-deformable and deformable mappings are need. From the results, we found the planning dose could be got a better delivery. For the OARs, the large volume of the PTV the delivered dose has a significant difference, while the smaller PTV there have no significant difference. Therefore, an adaptive lung SBRT should be better for the larger target volumes to reduce to delivery dose.

Funding
The study was partially supported by a grant from Young Innovative Talents Project (2018KQNCX128).