Clinical utility of a new immobilization method in image-guided intensity-modulated radiotherapy for breast cancer patients after radical mastectomy

1 Introduction

The latest global cancer data from the World Health Organization in 2020 reported that the number of newly diagnosed breast cancer patients had reached 2.26 million, making it the number one cancer in this globe [1]. Radiotherapy takes an important part in comprehensive treatment for breast cancer, and intensity-modulated radiotherapy (IMRT) after radical mastectomy can further reduce cancer recurrence and prolong patient survival [2, 3]. IMRT needs a higher reproducibility of patient immobilization. However, the use of a poor or inappropriate immobilization device may compromise the reproducibility of patient immobilization, bringing an excessive positional error, inducing a displacement of prescription dose and decreasing the IMRT efficacy [4].

The current immobilization device in breast cancer IMRT include breast bracket, vacuum bag, styrofoam, thermoplastic mask, and a combination of any of the above devices. Use of each device mentioned above has its pros and cons [5–8]. Among them, the thermoplastic mask has been frequently used because of its excellent reproducibility in patient immobilization. However, it is noteworthy that the thermoplastic masks do have a certain thickness, which moves forward the dose buildup and thus increases the unwanted dose and potential risks to the patient skin. Mean dose was found to increase by 0.5 Gy and 0.44 Gy for skin dose of left and right breast cancer patients after breast mastectomy, respectively [9]. Recently, a study reported an application of a new cervicothoracic thermoplastic mask in breast cancer IMRT, where a support board was also used, and the affected side of the thermoplastic mask was hollowed out [10]. This type of mask can provide good consistency between setup errors and the position of the upper and lower target volumes by immobilizing patient torso, easily movable neck and ipsilateral shoulder joint. Furthermore, this mask also exposes the irradiation breast surface (make it easy to check SSD of each radiation field on patients’ surface and so on) and reduces radiation-induced toxic on patient skin. However, the reproducibility of patient immobilization with this hollowed-out cervicothoracic thermoplastic mask has rarely been investigated Therefore, we explored the clinical utility of the hollowed-out cervicothoracic thermoplastic mask and compared its use with no use of it in breast cancer IMRT treatment after radical mastectomy for the first time. Comparison between two groups of patients (one group with the mask combined with a support board and the other with only the board) was conducted to verify whether there is any difference in setup errors between the two groups by daily image guidance. The main motivation of this study was to prospectively investigate the effect of the new cervicothoracic thermoplastic mask in breast cancer IMRT through daily image30-Mar-22 guidance. This paper explored the following purposes in the two immobilization methods:

to compare the differences and analyze the distribution of initial setup errors for the two groups in three translational directions

2 Materials and methods

2.2 Immobilization and CT simulation

The patients were randomly and equally divided into two groups. In one group (Group A), the patients firstly lay in supine position on a carbon-fiber board (Klarity R612, Guangzhou, China) with their arms raised over the head and positioned on an arm support. In addition, a headrest was selected (from a standard set of three) for each patient which offered the best comfort and stability. Then, a cervicothoracic thermoplastic material (Klarity R322, Guangzhou, China) which had part of it cut off according to the side of each patient’s affected breast was chosen and heated for 5–8 minutes in a constant temperature oven (70°C) (Klarity KT-820A, Guangzhou, China). The material was then lightly stretched horizontally before it was molded to the patient’s neck and chest for 10 min and eventually shaped into a hollowed-out cervicothoracic mask. The thermoplastic masks were not reused. Each mask was labeled with the patient’s name and personal identification number. The other group (Group B) used the board only. The different immobilization methods in the two groups can be seen in Fig. 1. Figure 2 shows structure and function of the board. Moreover, patients were immobilized in a supine position with spontaneous respiration and directed to breath calmly at a rate of around 14–18 breaths/min. Markers were drawn on the masks and patient skin to show the projection of the laser beams before all patients underwent large-aperture CT (Brilliance CT Big Bore, Philips Medical Systems Inc., Cleveland, OH, USA) at a 2.5-mm slice spacing from mandible to the level of 15 cm under the inferior border of patient breast.

OPEN IN VIEWER

Fig. 1

Immobilization objects for unilateral breast cancer patients after radical mastectomy: (A) A hollowed-out cervicothoracic thermoplastic material; (B) A carbon-fiber board with an arm support and a headrest; (C) A breast cancer patient immobilized with both the board and the hollowed-out cervicothoracic thermoplastic mask.

OPEN IN VIEWER

Fig. 2

Components of patient support board: (A) Adapter of headrest; (B) Holes for fixing grip bars; (C) A component used for relocating arm supports on the LONG direction; (D) A protractor used for adjusting the angle of arm supports; (E) A component used for relocating arm supports on LAT direction with A, B, C, D and E as indicators; (F) Adjustable holder with numbers from 1 to 6 as height indicators of each arm support.

2.4 Image guidance

An iSCOUT image-guided system was used not only to acquire patient images and initial setup errors but also to correct the errors before each treatment fraction. The iSCOUT system which applies stereoscopic non-orthogonal kilovoltage (kV) imaging was equipped to 23EX linac. The system consists of two amorphous silicon panel detectors fixed on the ceiling and two kV X-ray sources mounted on both sides of the treatment couch. The X-ray generators supplying high-voltage power operate at 50 kW at peak power output and can deliver X-rays with technique factors of 40–150 kVp, 10–500 mA, and 40–500 ms. The effective imaging area of each panel is 434 mm×434 mm with a resolution of 1024×1024 pixels. X-ray planar images were obtained in 45° and 135° directions with an imaging time of approximately ten seconds and an imaging dose of 0.2–1.0 mGy. Figure 3(A and B) shows the structure and workflow of the iSCOUT system.

OPEN IN VIEWER

Fig. 3

(A) A schematic diagram of iSCOUT system and translational directions; (B) A brief workflow of iSCOUT image-guided system; (C) The registration interface where DRR images and acquired real-time images can be aligned to acquired setup errors. The red frame indicates DRR images, the orange frame indicates acquired real-time images, and the green frame includes registration results in three translational directions (LAT, LONG, VERT).

As for the clinical implementation of the iSCOUT system, this system was used for image guidance at each treatment fraction. Before each treatment, the real-time patient images were acquired and aligned with the two digitally reconstructed radiograph (DRR) images derived from the planning CT in the iSCOUT system software. Image alignment was performed to obtain initial setup errors in three translational directions (including the left-right (LAT), superior-inferior (LONG), anterior-posterior (VERT) directions) using bony registration algorithm followed by careful visual inspection from two radiation therapists and online confirmation from one experienced oncologist. Manual registration would be performed if there was a misalignment. Bony regions of interest included sternum, ribs, thoracic vertebrae and clavicles. The registration interface is shown in Fig. 3(C). If any initial error in any of the three translational directions was greater than a tolerance threshold of 3 mm, the therapists would correct the treatment couch position in the iSCOUT system using the online remote control unit. This alignment and correction process shall be repeatedly executed until the three translational errors are within 3 mm. These final errors were recorded as residual errors.

2.5 Data comparison and statistical analysis

For all the patients, the initial and residual setup errors in LAT, LONG, VERT directions were recorded for each treatment fraction. Overall setup error was calculated as the root-mean-square (RMS) of the setup errors in all three translational directions. Data collection and process were as follows: (1) Compare the initial setup errors and their distribution in the three translational directions between the two groups. Setup errors were divided using two threshold values (3 mm and 5 mm) into three ranges (≤3 mm, 3–5 mm, and > 5 mm). (2) Compare the daily and weekly changes of the initial setup errors from the two groups to find any potential changing pattern. (3) Calculate PTV margins with initial and residual setup errors according to the PTV margin formula M = 2.5∑ + 0.7δ proposed by Van Herk [11], where ∑ is the standard deviation (SD) of the systematic error and δ is the random error calculated as the RMS of patient-specific SDs. Figure 4 gives an example on how to calculate PTV margin in left-right direction for patients from Group A.

OPEN IN VIEWER

Fig. 4

A example on calculation of PTV margin. The mean initial setup errors in left-right direction on the totally 15 fractions for each patient from Group A are calculated. The standard deviation (SD) of these 20 mean initial setup errors is the standard deviation (SD) of the systematic error, i.e. ∑. The SD of δ initial setup errors for each patient from Group A are also calculated. δ is the random error calculated as the root-mean-square of patient-specific SDs. PTV margin formula M = 2.5∑ +0.7δ proposed by Van Herk was then used to calculate the PTV margin.

SPSS 26.0 was used for statistical analysis. Independent sample T-test and Chi-square test were performed for comparison. A P value < 0.05 was considered statistically significant.

All the 40 patients had successfully completed the entire IMRT treatment course with a total of 300 initial setup errors and 300 residual errors in each of three translational directions for each group. Patient demographics of the two groups were analyzed and their results are presented in Table 1. There was no statistical difference for all the characteristics (P > 0.05).

Table 2 presents the initial setup errors in three translational directions and their overall setup errors between the two groups. The initial setup errors of the Group A in LONG direction were significantly smaller than those of the Group B (P < 0.05). There was no significant difference between the initial setup errors in both the LAT and VERT directions. The distribution of the initial setup errors in three different ranges is shown in Table 3. With the performance of Chi-square test, it was found that there was no significant difference between the two groups on the distribution of initial setup errors in all three directions (all X² = 0.046, P > 0.05). Figure 5 shows the distribution of the initial setup errors in the two groups. PTV margins were calculated with the systematic and random errors of the initial setup errors for the patients in both groups (see Table 4). The PTV margins were 8.01, 5.44, 5.45 mm and 8.14, 10.89, 6.29 mm in three translational directions for the Group A and Group B, respectively.

OPEN IN VIEWER

Fig. 5

(A) Distribution of initial setup errors in three translational directions between Group A and Group B; (B) Distribution of initial setup errors with three different ranges in Group A and Group B; (C)Mean weekly initial setup errors in three translational directions for Group A and Group B; (D)Inter-fraction initial setup errors in LAT direction for Group A and Group B; (E) Inter-fraction initial setup errors in LONG direction for Group A and Group B; (F) Inter-fraction initial setup errors in VERT direction for Group A and Group B.

Residual errors in the three translational directions and their overall setup errors for the two groups are shown in Table 5. It showed that there were no significant statistical differences in the residual errors between the two groups (P > 0.05). It should also be noted that initial setup errors were greatly reduced into residual errors in the LAT, LONG and VERT directions for both the two groups. In Group A, the initial and residual setup errors in the LAT, LONG and VERT directions were 3.14±3.07 mm vs. 1.74±1.03 mm; 2.21±1.92 mm vs. 1.62±0.92 mm; 2.45±1.92 mm vs. 1.66±0.91 mm, respectively (all P < 0.001). In Group B, the initial and residual setup errors in the LAT, LONG and VERT directions were 3.14±2.97 mm vs. 1.70±0.97 mm; 2.94±3.35 mm vs. 1.68±1.18 mm; 2.80±2.47 mm vs. 1.58±0.98 mm, respectively (all P < 0.01). PTV margins calculated with the residual errors for the Group A and Group B in the LAT, LONG and VERT directions were 3.54, 2.99, 2.89 mm and 2.67, 3.64 2.74 mm, respectively (see Table 6). Figure 6 illustrates CTV (white), PTV with residual setup error (green), clinical PTV (yellow) and PTV with initial setup error (red) by size.

OPEN IN VIEWER

Fig. 6

Illustrates CTV (white), PTV with residual setup error (green), clinical PTV (yellow) and PTV with initial setup error (red) of one patient from Group A (A, B, C) and another from Group B(D, E, F).

Reproducibility of patient positioning is an important factor for high-precision breast cancer IMRT. Song et al. [12] demonstrated in a study that using thermoplastic masks can significantly reduce inter-fraction setup errors in IMRT for breast cancer compared to the use of vacuum-lock bags. Similarly, a research by Ma et al. [13] reported that patient immobilization with both a breast bracket and a thermoplastic head mask could improve the reproducibility of treatment setup for breast cancer patients who received whole breast and supraclavicular nodal region irradiation. In that research, the translational displacement using the breast board without/with the head mask in the LAT, LONG and VERT directions were 0.212±0.174 cm vs. 0.272±0.242 cm, P = 0.07; 0.364±0.246 cm vs. 0.242±0.171 cm, P = 0.001; 0.423±0.302 cm vs. 0.364±0.269 cm, P = 0.204, respectively. Although the pleasant effectiveness of thermoplastic mask fixation has been demonstrated by some studies, there are still some issues. It was suggested in a paper that immobilization with thermoplastic masks, especially covering the whole irradiated area with a thermoplastic mask, might have an effect on increasing the irradiated skin dose at the irradiated site and thus aggravating the radiological skin reaction [9, 14]. Therefore, the design of thermoplastic mask should also be careful and customized according to the needs of different disease types, but there were few studies on the thermoplastic mask immobilization for breast cancer IMRT.

In order to further explore the clinical utility of the thermoplastic mask immobilization, our study selected cervicothoracic thermoplastic masks which were hollowed out on the side of patient’s affected breast. Our results showed that there was statistically significant difference on the initial setup errors only in the LONG direction between patients using hollowed-out cervicothoracic thermoplastic masks combined with a carbon-fiber board (Group A) and those using only the board (Group B). The initial errors in the LONG direction for Group A and B were (2.21±1.92) mm and (2.94±3.35) mm, respectively. But there were no significant differences for the initial setup errors in the LAT and VERT directions. These results agree with the results of the study from Ma et al. [14]). Such difference in the LONG direction led to a significant difference on overall initial setup errors between the two groups (Table 2). These masks significantly reduced mean overall initial setup error by 0.76 mm and lowered SD by 0.87 mm (5.15±3.32 mm vs 5.91±4.19 mm). On the other hand, a study presented by Li et al [15] showed no significant difference in CTV dosimetry for an error difference of 1 mm or less, and further they suggested that an error of more than 3 mm perpendicular to the mammary tangential field direction would cause a significant dose difference in CTV. Hence, our study compared the distribution of initial setup errors for the two groups in three different ranges (i.e., less than 3 mm, from 3 mm to 5 mm, and more than 5 mm). It was found that in the range of less than 3 mm, the proportion of group A in the LONG direction was 8.34% more than that of group B (83.67% vs 75.33%), which suggests that the use of these thermoplastic masks can serve a better immobilization in the LONG direction and thus reduce prescription dose displacement for the patients. Notably, no significant difference was shown in the weekly and daily initial setup errors. The reason for better pretreatment setup in the Group A might be that the cervicothoracic thermoplastic masks provided excellent reproducibility on the neck and axilla. Moreover, a thermoplastic mask, fixed with fasteners to the board, can prevent the patients from shifting in the LONG direction, which may serve as a good solution for patients with smooth skin. This is of clear benefit to patients who tend to sliding on the support board. It also suggests that patients wear rough pants during IMRT treatment course and that the support board be designed with higher skin friction to reduce potential risk of inter- and intra-fraction sliding.

With the advent of image-guided equipment, positional setup errors can be more effectively corrected prior to patient treatment fraction [16]. The existing image guidance equipment includes electronic portal imaging device, cone beam CT and optical surface tracking system, and each of them has its own advantages [17–19]. The iSCOUT system used in this study can acquire X-ray images in two directions (i.e., 45° and 135°) simultaneously by its software controlling two sets of X-ray imaging units symmetrically positioned at the center point, and each image acquirement takes only ten seconds with only a little imaging dose of 0.2–1 mGy. Image guidance tremendously reduced initial setup errors to residual errors in all three translational directions for both Group A and B. In addition, the PTV margins from the residual errors calculated for the Group A and the Group B were 3.54, 2.99, 2.89 mm and 2.67, 3.64, 2.74 mm in the LAT, LONG and VERT directions, respectively. These residual PTV margins in all three translational directions were within 5 mm, which were much smaller than any of the initial PTV margins with a minimum of 5.44 mm and a maximum of 10.89 mm (see Table 4). Therefore, it is recommended to perform image guidance on the daily basis for patients’ sake.

Routine QA procedures for iSCOUT system are implemented in our institute. The accuracy (the difference between the real and expected setup errors) and precision (the repeatability) are parts of the acceptance testing procedures and the routine QA procedures. A pair of films and a cube phantom were used to identify the accuracy and precision (Fig. 7). Beam was delivered to the films fixed in the cube phantom from four gantry angles: 45°, 135°, 225°, and 315°, respectively. Each beam was delivered with 110 MU. The field size was set to 2.5 cm×2.5 cm. Planar kV images were acquired using the iSCOUT system to obtain the displacement A between the setup position and the iSCOUT system isocenter. After that, the pair of films was scanned and analyzed to acquire the displacement B between setup position and the linac isocenter. The difference between the A and B refers to the difference between the iSCOUT system isocenter and the linac isocenter. The procedures for the iSCOUT accuracy were repeated for another two times. Mean of the three difference values between the iSCOUT system isocenter and the linac isocenter was defined as accuracy. The deviation of the three difference values from the accuracy was considered as repeatability, namely precision. As we can observe from Table 7, the results show that the accuracy was approximately 0.6 mm (less than 1 mm) and the precision was within 0.06 mm (far less than even 0.1 mm), which means our iSCOUT system has a high accuracy and precision and thus our results could be trusted.

OPEN IN VIEWER

Fig. 7

A pair of films and a cube phantom were used to identify the accuracy and precision of the iSCOUT system.

In our study, Chinese BMI classification criteria was chosen because all breast patients enrolled were Chinese, where a BMI between 18.5 kg/m² and 23.9 kg/m² is considered normal weight, while overweight is defined as having a BMI no less than 24.0 kg/m² and obesity with a BMI no less than 27.9 kg/m². As for the patient demographics in our study, all detected BMIs indicated either normal or overweight, none of them were found obese. We then went deep into the correlation between pretreatment setup errors and fraction number for patients of normal weight (18.5 kg/m²≤BMI < 23.9 kg/m²) and overweight (24.0 kg/m²≤BMI < 27.9 kg/m²) in Group A and Group B, respectively. The mean initial setup errors of the two BMI groups in each fraction was compared for both the Group A and Group B. SPSS 26.0 was used for statistical analysis and Mann-Whitney U test was used for data result of inequal variance. Comparison of initial setup errors on the three translational directions between patients in the two BMI groups is illustrated in Fig. 8. It was found that pretreatment setup errors did not correlate with fraction number for the two BMI group in both Group A and Group B. Thus, the initial setup errors in different fractions can be averaged for the following comparisons. For patients in Group A (with masks), the mean translational displacement for patients of normal weight vs. overweight in the LAT, LONG and VERT directions were 2.86 mm vs. 3.65 mm, P = 0.187; 2.00 mm vs. 2.60 mm, P = 0.061; 2.37 mm vs. 2.60 mm, P = 0.539, respectively. For patients in Group B (without masks), the mean translational displacement for patients of normal weight vs. overweight in the LAT, LONG and VERT directions were 2.95 mm vs. 3.69 mm, P = 0.098; 3.03 mm vs. 2.69 mm, P = 0.161; 2.58 mm vs. 3.46 mm, P = 0.002, respectively. This result showed that setup errors were slightly greater for overweight patients compared to normal BMI patients, either in patients with masks or patients without masks. Only the mean translational displacement for patients of normal weight from Group B (without masks) was witnessed significantly smaller than that for overweight patients in the VERT direction. Notably, a recent French publication indicated that BMI (17–37 kg/m²) did not have a significant impact on repositioning accuracy for 24 breast cancer patients who received postmastectomy IMRT (P = 0.414) [20]. Therefore, BMI impact on setup errors still needs further investigation.

OPEN IN VIEWER

Fig. 8

Initial setup errors in the LAT, LONG and VERT directions between patients of normal weight (18.5 kg/m²≤BMI < 24.0 kg/m²) and overweight (BMI≥24.0 kg/m²) from Group A and Group B.

It has also been mentioned in previous studies that the initial setup errors can be greater in patients with excessive breast volume and weight [21]. A more precise image-guided approach should be used for these patients. In addition, several measures could be taken to improve the accuracy in IMRT, such as taking a prone position immobilization to keep patient’s breasts in a natural drooping state, or having the patient consciously exercised to lost weight before IMRT treatment, or increasing the number of image-guided exposures or choosing a more appropriate image guidance modality if possible. With the increasing use of artificial intelligence (AI) in radiotherapy, we expect the application of AI into setup errors prediction [22, 23]. Therefore, with the help of mathematical models or computerized deep learning, the setup errors in breast cancer can be decreased in the future and the calculation of body indicators such as weight and BMI can also be quantified, which will be a future research topic.

There are some limitations in this study. First, all breast cancer patients enrolled in this study received radical mastectomy, but to date there have been no reports on patients after breast-conserving surgery using our hollowed-out cervicothoracic thermoplastic mask. By the way, we are now conducting a research in this area. Second, our study focused on inter-fraction setup errors rather than intra-fraction errors. Whether the thermoplastic masks may have a positive advantage on reducing intra-fraction errors remains unknown. A newly published research revealed that intra-fraction breast CTV displacement measured on 4DCT was much smaller (0.5±0.5 mm vertically, 0.5±1.0 mm longitudinally, and 0.3±0.3 mm laterally, respectively) than the inter-fraction CTV displacement measured by CBCT (2.6±2.2 mm vertically, 2.8±2.3 mm longitudinally, and 1.7±1.2 mm laterally, respectively) [24]. Thus, our research on inter-fraction setup errors was enough and this reduced the additional intra-fraction radiation exposures to the patients. Third, our hospital does not have a six-degree freedom couch, only has a four-degree freedom couch (three translational direction and yaw rotation). Without a six-degree freedom couch to correct initial rotational setup errors on pitch and roll rotations, the residual setup errors could not be sufficiently corrected and thus increased. Finally, dose justification for the daily image guidance used in this study has not been reported yet. Nevertheless, the iSCOUT dose is only at a mGy level compared to other image guidance modalities with more exposure dose. The additional dose from iSCOUT system on patients and secondary carcinogenic effects needs to be confirmed by more studies.

In conclusion, the use of hollowed-out cervicothoracic thermoplastic masks combined with a carbon-fiber support board showed better inter-fraction immobilization than the single use of the board in reducing longitudinal setup errors for breast cancer patients after radical mastectomy during IMRT treatment course. In addition, the application of an image-guided system on the daily basis can reduce setup errors and improve the pretreatment immobilization accuracy for breast cancer IMRT after radical mastectomy.

Author contributions

FD and XW drafted the manuscript and worked on the conception, design, and interpretation of data. XD and YY helped with data processing and manuscript drafted. BX and XL reviewed the data analysis. All authors contributed to the article and approved the submitted version.

Funding

XW is funded by Fujian Medical University Sailing Fund General Project (2020QH1078); XL is funded by Industry-University-Research Project of Fujian Science and Technology Department (2020Y4010). This work was funded by the above grants.