Dose coverage and breath-hold analysis of breast cancer patients treated with surface-guided radiotherapy

Patient characteristics

91 patients who received local or locoregional breast radiation therapy with voluntary moderate deep inspiration breath-hold were included in this study from September 2020 to August 2021. Table 1 summarizes the patient characteristics. 6 patients were excluded from participation before the treatment started, due to several reasons (treatment did not meet the inclusion criteria (4 patients), technical failure of the SGRT system during the CT-scan (1 patient), and 1 patient withdrew her consent). One patient was excluded after the first treatment fraction, since she was not able to hold her breath long enough and it was decided to apply radiation during free breathing. Hence, in total data of 84 treated patients were used for the analysis. Informed consent was given by the participants and the study was approved by the medical ethics committee METC Brabant (CCMO register NL69214.028.19).

Table 1 Radiotherapy parameters and characteristics of the patients (n = 84)Treatment technique

Patients were treated on a Varian TrueBeam™, Clinac iX™, or Trilogy™ machine using 6 or 10 MV photon beams. The treatment plans were generated using the external beam planning module in ARIA 15.6 (Varian Medical Systems, Palo Alto, CA, USA) with the Acuros dose calculation algorithm [24, 25]. For most local breast cancer patients the treatment technique consisted of two tangential dynamic IMRT beams. For a few patients, a 180° VMAT beam with a weight of 10% was added. For local breast cancer patients receiving a simultaneously integrated boost, the technique consisted of two tangential IMRT beams, combined with two boost IMRT beams or one 180° VMAT beam. A hybrid technique consisting of two tangential open beams (80% of breast dose) and three 60˚ VMAT arcs (20% breast dose and 100% nodal dose) was used for the locoregional patients. We required a minimum value of 95% of the PTV volume that received at least 95% of the prescribed dose. We required a mean heart dose of maximum 3 Gy and aimed to ≤ 1 Gy for right breast irradiation and ≤ 2 Gy for left breast irradiation. When the dose to the heart exceeded 3.2 Gy for one of these plans, a plan with only VMAT arcs was applied to reduce the heart dose [9]. Delineation of the target volumes was performed according to the ESTRO guidelines [26]. The CTV to planning target volume (PTV) margin was 5 mm in all directions, for all CTV’s. Moreover, the CTV and PTV are cropped at 3 mm under the skin, and thus aligned at the skin side.

Optical surface scanning system

The IDENTIFY™ system (Varian Medical System, Palo Alto, CA, USA) was applied for patient set-up and intra-fractional patient motion monitoring. For patient set-up, first a time-of-flight camera was used to generate the whole-body set-up. An infra-red light signal was send and the time needed to detect the reflected light was measured to determine the distance to the object [22]. This technique was used to make a 3D reconstruction of the entire body of a patient and was especially useful to position the arms of the patient. The whole-body reference image was acquired after finishing the CT-scan, while the couch was in the lowest position. Secondly, for precise positioning stereo vision based cameras were used. Two cameras captured the structured light pattern that was projected on the object. Hence, the location of every unique point was determined. Real-time surface images were compared to the reference image, obtained from the CT-scan. A rigid registration algorithm calculated the deviations of and around the isocenter to bring the two surfaces into alignment. The SGRT system used in this study consists of three sets of stereo vision cameras. A daily QA procedure was performed to verify the isocenter alignment. Submillimeter accuracy was achieved for the three translational degrees of freedom and for each rotation [22].

Patient set-up using SGRT

All patients were positioned on a breast board with an inclination of 7.5°, with the arms lifted above the head in an arm support. Patient set-up was performed during breath-hold. Since a CT-scan takes about 45s to obtain, the procedure was split and the scan was acquired during 3 breath-holds. The patient body outline on the CT-scan was used to generate a surface image that served as reference image during set-up. The CT-scan was also used to calculate the absolute couch position at the treatment machine, where the patient is positioned to match the surface of the patient with the reference surface [4]. Since the position of the couch is known, a CT scan during breath-hold is enough to be able to reproduce the breath-hold level and a free-breathing scan is not needed.

During patient treatment, optical surface scanning was used for set-up, instead of the conventional method of using tattoo points on the skin and laser lines. The patient was positioned at the couch and a whole-body surface scan was used for the initial patient positioning, especially concerning the arms, to make the final alignment of the patient easier. Subsequently, SGRT cameras were used for precise positioning, while the patient held her breath. A threshold value of 5 mm for vertical, lateral and longitudinal direction, as well as for vector deviation was used, and a threshold value of 2° for pitch, roll or yaw rotation was used during set-up. The online match procedure was performed using orthogonal kV-images. The focus was on the implanted radio opaque fiducial clips, inserted in the breast during surgery before radiotherapy, followed by a check of the patient anatomy. Once a week, an extended CBCT scan was acquired.

After the online match procedure, a couch displacement was performed to correct for setup errors. To make sure the breath-hold level did not change during the couch displacement, an MV-image in the medio-lateral direction was acquired as additional check whether the breath-hold was within 5 mm of the reference value, using the position of the chest wall. Repositioning was performed when the residual error was more than 5 mm. After the online match procedure and couch displacement, a new surface scanning reference image, also during breath-hold, was captured for patient monitoring during the treatment fraction of that day.

A region of interest including the breast, sternum, and caudal part of the contralateral breast was drawn and was used both for set-up and intra-fractional motion monitoring by the surface scanner. Patients were instructed by a spoken recorded command to hold their breath. The breath-hold was monitored by means of the chest wall surface excursion. If the isocenter deviation was outside the threshold values of 5 mm translation or 2° of rotation around any orthogonal axis, the radiation was manually interrupted by the radiation technologist. The patient was coached by the technologists to change the depth of inspiration and subsequently the radiotherapy session was continued.

Set-up accuracy of the arm

The accuracy of the arm positioning of the patient was assessed by determining the clavicle position. The anterior–posterior kV-image was used to match the clavicle to the CT-scan and the rotation error was determined. Success of the treatment was defined when in no more than five (for the 15 fraction scheme) or six (for the 20 fraction scheme) treatment fractions the difference in clavicle rotation was more than 3°, and in maximal two fractions it exceeded 5°. Based on a preliminary investigation in breast cancer patients treated without SGRT, the null hypothesis was that the probability of success equals 0.8. The study was designed to prove that in more than 80% of the patient treatments the set-up was successful. Statistical power was based on a one-sided exact binomial test using statistical package R (online R code compiler [27]).

In this study, the online acquired kV-images were compared to the digitally reconstructed radiographs of the reference CT-scan in an offline match procedure. The rotation around the dorso-ventral axis was used as outcome measure for the clavicle set-up.

Dose calculation

To investigate the effect of changes in anatomy and changes in patient set-up, the actual delivered dose was calculated. Once a week, during fraction 2, 7, 12, and 17 if applicable, an extended CBCT was acquired, a CBCT-CT registration was performed, and the delivered dose was calculated on the extended CBCT scan. In this way we simulate an online match procedure based on an online CBCT match which we introduced in our institute after the trial. To ensure that the dose calculation on the CBCT reflects real values, Hounsfield units of the lungs and breast tissue on the CT and CBCT were compared for three patients. The differences between the mean values of each region were less than 20 HU, resulting in an accurate dose calculation using the CBCT.

A deformable registration of the CT to the CBCT was applied to deform the structures of the CT onto the CBCT, using the Image Registration module in the Varian treatment planning system. The registration was checked manually, and if necessary, changes were applied to the acquired contours. Second, the dose to the CBCT was calculated using Varian Eclipse treatment planning system. A database with relevant dose-volume values was calculated by means of in-house written scripts using the Python dicompyler-core library and subsequent analysis was performed using MATLAB R2015b (The MathWorks, Inc., Natick, USA). The minimum dose in percentage of the prescribed dose that was received by 98% of the CTV volume (D98) was determined for the breast, chest wall, axillary lymph nodes, and internal mammary lymph nodes (IMN). The D98 was determined for the fractions during which an extended CBCT was made. To make an estimation of the dose coverage of the whole treatment for each patient, the dose coverage was averaged over all CBCTs.

To investigate whether there is a correlation between the clavicle position and the dose coverage, the D98 for the different CTVs was plotted against the clavicle rotation.

Breath-hold analysis

With SGRT, the position of the chest wall was monitored, and this was reported as the translations and rotations around the isocenter. The deviations in vertical, lateral, and longitudinal direction as function of the time were used to analyse the breath-hold. The outcome parameters used to assess the quality of the breath-hold (BH) were the residual set-up error (RE), the reproducibility (R), and the stability (S). The RE per breath-hold i is defined as the mean displacement from the start of the breath-hold (t = 1) till the end (t = T):

$$RE_ = \frac\mathop \sum \limits_^ BH_$$

The RE per treatment fraction f is defined as the average REi in that fraction:

$$RE_ = \frac\mathop \sum \limits_^ RE_$$

with N is the total number of breath-holds per treatment fraction.

Reproducibility of a treatment fraction was defined as the consistency of the depth of the breath-hold, given by the standard deviation of the mean vertical displacement of each breath-hold i during a treatment fraction:

$$R_ = \sqrt ^ \left( - RE_ } \right)^ }}}$$

with N is the total number of breath-holds per treatment fraction.

Stability of a single breath-hold was defined by the standard deviation of the breath-hold level, according to the definition used by Reitz et al. [28]:

$$S_ = \sqrt ^ \left( - RE_ } \right)^ }}}$$

Stability of the breath-hold of a treatment fraction was defined by the mean of the breath-hold levels in a treatment fraction, given by:

$$S_ = \frac\mathop \sum \limits_^ S_$$

with N is the total number of breath-holds per treatment fraction, BHj is the breath-hold level at timestamp j, and T is the number of measurements of a breath-hold.

The breath-hold parameters REi, REf, and Si are visually explained in Fig. 1.

Fig. 1figure 1

Plot of a breath-hold signal, visualizing the breath-hold parameters REi, REf, and Si

留言 (0)

沒有登入
gif