A preliminary investigation of re‐evaluating the irradiation dose in hepatocellular carcinoma radiotherapy applying 4D CT and deformable registration

Abstract Purpose To investigate the effect of breathing motion on dose distribution for hepatocellular carcinoma (HCC) patients using four‐dimensional (4D) CT and deformable registration. Methods Fifty HCC patients who were going to receive radiotherapy were enrolled in this study. All patients had been treated with transarterial chemoembolization beforehand. Three‐dimensional (3D) and 4D CT scans in free breathing were acquired sequentially. Volumetric modulated arc therapy (VMAT) was planned on the 3D CT images and maximum intensity projection (MIP) images. Thus, the 3D dose (Dose‐3D) and MIP dose (Dose‐MIP) were obtained, respectively. Then, the Dose‐3D and Dose‐MIP were recalculated on 10 phases of 4D CT images, respectively, in which the end‐inhale and end‐exhale phase doses were defined as Dose‐3D‐EI, Dose‐3D‐EE, Dose‐MIP‐EI, and Dose‐MIP‐EE. The 4D dose (Dose‐4D‐3D and Dose‐4D‐MIP) were obtained by deforming 10 phase doses to the end‐exhale CT to accumulate. The dosimetric difference in Dose‐3D, Dose‐EI3D, Dose‐EE3D, Dose‐4D‐3D, Dose‐MIP, Dose‐EIMIP, Dose‐EEMIP, and Dose‐4D‐MIP were compared to evaluate the motion effect on dose delivery to the planning target volume (PTV) and normal liver. Results Compared with Dose‐3D, PTV D99 in Dose‐EI3D, Dose‐EE3D and Dose‐4D‐3D decreased by an average of 6.02%, 1.32%, 2.43%, respectively (P < 0.05); while PTV D95 decreased by an average of 3.34%, 1.51%, 1.93%, respectively (P < 0.05). However, CI and HI of the PTV in Dose‐3D was superior to the other three distributions (P < 0.05). There was no significant differences for the PTV between Dose‐EI and Dose‐EE, and between the two extreme phase doses and Dose‐4D (P> 0.05). Negligible difference was observed for normal liver in all dose distributions (P> 0.05). Conclusions Four‐dimensional dose calculations potentially ensure target volume coverage when breathing motion may affect the dose distribution. Dose escalation can be considered to improve the local control of HCC on the basis of accurately predicting the probability of radiation‐induced liver disease.


| INTRODUCTION
For patients with unresectable hepatocellular carcinoma (HCC), transarterial chemoembolization (TACE) followed by volumetric modulated arc therapy (VMAT) is a safe and effective treatment which can achieve better outcomes than either of these therapies alone. 1 However, radiation-induced liver disease (RILD) remains the limiting factor of dose escalation for the target volume. Abdominal organs can move and deform significantly during breathing. In order to estimate the RILD, it is important to evaluate the effect of breathing motion on dose distribution.
The majority of patients are often treated in free breathing, which may bring the geometric uncertainties to either the target volume or normal liver. 2 Further, the interplay between the incident beam and tumor motion derived from breathing motion could potentially introduce dosimetric error, leading to difference between the dose distributions from 3D static plan and the radiation dose actually delivered. 3 The 3D plan in free breathing could not reflect the dose to the target volume and normal liver during whole breathing cycle. This would make the prediction to the probability of RILD inaccurate.
Recently, four-dimensional (4D) CT has revealed the real motion information of regions of interest (ROIs) during breathing. 4 4D CT has been used in precision radiotherapy of lung cancer, HCC and pancreatic carcinoma, which were affected by breathing motion significantly. 5 It had been proved that the application of 4D CT in HCC radiotherapy could improve the target localization, and the potential clinical benefits have been summarized. 6 To estimate real dose delivered to the patients, 4D accumulated dose using 4D CT and deformable registration (DR) has been studied to explicitly account for the effects of breathing motion on dose distribution. 7 DR can model the tissue deformation by solving the voxel-to-voxel transformations between two images. 8  In this study, we investigated the feasibility of 4D dose accumulation for the target volume and normal liver in radiotherapy of HCC using Morfeus, and analyzed the dosimetric differences between 3D and 4D dose distributions. In addition, end-inhale and end-exhale phase doses were calculated to evaluate the dosimetric effects of rigid motion on the target and normal liver.

2.A | Patient selection
Fifty HCC patients were consecutively included in this study. All patients were treated on a Research Ethics Board of our institution approved protocol, with written informed consent obtained from them. The inclusion criteria included the patients with unsuitable or unwilling resection, Child-Pugh liver score A, KPS more than 80. All patients had received TACE then re-planned by VMAT, each with complete lipiodol retention. The patients included 39 men and 11 women, with a median age of 63.5 years (range, 41-77 years). Of the 50 tumors, 22 located in the left lobe of the liver, 28 in the right lobe.

2.C | ROIs delineation and treatment planning
The GTVs and liver were delineated on all CT images. The GTV contours were determined by the radiation oncologists with consultations from the radiologists. The internal gross tumor volume (IGTV) was generated by merging 10 GTVs on all phases of 4D CT images.
PTV in 3D images were obtained from GTV plus a 1.2-cm margin in head-foot direction and 1.0-cm margin in other direction. 10 PTV in MIP images were obtained from GTV plus a 0.8-cm margin in each direction. Normal liver was defined to subtract PTV from the whole liver. The treatment planning goal was to achieve 98% of PTV coverage with 100% of the prescription dose, while 10% of the volume of

2.D | DR
Dose accumulation requires that tissues be accurately tracked between images. In this study, a biomechanical DR based on the Morfeus method was used. 12 This algorithm deforms the structures by solving a linear elasticity problem using the finite element model.
The problem is set up by controlling ROIs represented by meshes with vertex-to-vertex correspondence in the reference and target image sets. The meshes can be generated with model-based segmentation or with dedicated tools from contours. For controlling ROIs representing interior structures, including liver, PTV, GTV, the interface between the structure and the surrounding tissue can be modeled as either fixed or sliding. Thus the biomechanical DR is completely geometry-based and does not incorporate any grayscale information. The accuracy of this method for all deformed tissues is less than 0.2 cm. 12

2.F | Plan evaluation
For PTV, the D99, D95, and D1 were defined as the least doses received by 99%, 95%, and 1% of the target volume. The homogeneity index (HI) was described by using the ratio of D2 to D98, as follows: The conformal index (CI) was calculated as follows: where TV RI is the target volume covered by the prescription dose, TV is the target volume, and V RI is the volume of the prescription dose. The mean dose delivered to the normal liver (D mean ), V5, V10, V20, V30, and V40 were also evaluated, where V x represents the percentage of the volume of x Gy in the normal liver.

2.G | Statistical analysis
The data were analyzed using SPSS 17.0 software package (IBM, Chicago, IL, USA). The dosimetric parameters of Dose-3D , Dose-EI , Dose-EE and Dose-4D for PTV and normal liver were compared using Friedman test. The Wilcoxon test was used for the pairwise data.
Differences were considered significant at P < 0.05. The flow chart of this study was depicted in Fig. 1.

3.B | Registration evaluation
A quantitative assessment have been completed previously by Brock et al. 12 Further, visual inspection was performed to test the registration differences between target and reference images. Figure 2 displays the results of GTV showing the largest breathing amplitude.
Inspection of these images before and after registration showed the good performance of the algorithm, and differences were substantially reduced. Note the good agreement of the liver and tumor after registration. The corresponding deformable vector fields were shown in Fig. 3. The color and arrows corresponds to the magnitude and direction of the vector in each point.

3.C | Dosimetric indices for the PTV
As shown in Table 1, for PTV D99 and D95, Dose-3D was higher than those from Dose-EI-3D , Dose-EE-3D , and Dose-4D-3D ; while for PTV CI and HI, Dose-3D had the optimal values, compared with the other dose distributions (P < 0.05). No significant differences were observed for PTV D99, D95, CI, and HI among Dose-4D , Dose-EI , and Dose-EE (P> 0.05). Figure 4(a) displays the decrease of the PTV coverage in Dose-3D-EI , Dose-3D-EE , and Dose-3D-4D due to breathing motion compared with Dose-3D . Underdosing of these PTVs was also displays as the wider shoulder in the DVHs in Fig. 5(a).

3.D | Dosimetric indices for normal liver
Dose-4D-3D (P> 0.05). In Dose-EI-3D , a decrease of normal liver mean dose is statistical difference (P < 0.05). Table 4 shows the mean differences for D mean , V5, V10, V20, V30, and V40 in Dose-MIP , Dose-EI-MIP , Dose-EE-MIP , and Dose-4D-MIP , with none of these differences being significant (P> 0.05). In Fig. 5 This approach is especially beneficial for the clinic to effectively predict the probability of RILD.
In 3D plans, differences between Dose-3D and Dose-4D derived largely from geometric uncertainties induced by breathing motion.    18 Another limitation was that we did not evaluate the biological effect of the 4D dose. It was reported that significant increases (decreases) in liver NTCP occurred for tumors located toward the bottom (top) of the liver when patients scanned at exhalation. 21 As described previously, the NTCP of normal liver in this study was less than 5% when we set a criteria for ROIs. 11 This temporal variation in anatomy highlights the importance of breathing motion management techniques. Active breathing control or gating, if suitable, is an effective method to reduce the magnitude of tumor motion. 22,23 Through the use of these techniques for better control of breathing motion, the dose differences will be smaller.
However, the reproducibility of these techniques should not be neglected.

| CONCLUSION
Four-dimensional dose accumulation can be implemented with the availability of 4D CT and deformable registration in the presence of breathing motion. Compared with 3D static plan, the accumulated dose can reflect the more real dose to the target and the normal liver. And it is important for HCC patients to accurately predict the probability of RILD and facilitate the further safe dose escalation.

CONFLI CTS OF INTEREST
The authors have no other relevant conflicts of interest to disclose.