Between February 2022 and October 2023, 17 patients with left-sided breast cancer treated with surface-guided voluntary DIBH and 20 patients with left-sided or right-sided breast cancer treated by surface guided FB at the Department of Radiation Oncology, Sichuan Cancer Hospital, were recruited for the retrospective study. This retrospective study was approved by the Ethics Committee of our hospital (Approval Number No. SCCHEC-02-2021-026). The mean age and median age were 57.6 ± 12.2 yrs and 55.2 yrs, respectively. Patient characteristics and radiotherapy prescription are summarized in Table 1.

The patients were immobilized with a WingStep (IT-V, Innsbruck, Austria) breast board in the supine position with their arms above their head. CT scans with a 3 mm slice thickness were acquired with a 16-slice Brilliance Big Bore CT (Philips Medical Systems, Cleveland, OH, USA). The CT scans were performed in free breathing (FB) and breath hold (BH) positions or with FB alone for their use in DIBH or FB treatment, respectively. After the CT scan, imaging datasets were imported to MIM Version 7.0.5 (MIM Software Inc.) for contouring. The clinical target volume (CTV) and organ at risks (OARs) were delineated on each DIBH and FB scan by experienced radiation oncologists of the breast department. For the DIBH group, the clinical target volume (CTVDIBH) encompassed the whole breast, excluding chest wall muscles, ribs, and pectoralis muscles, while for the FB group, CTVFB encompassed the whole breast and supraclavicular fossa region, and gross tumor volume (GTV) included the tumor bed, visible surgical clips, and anatomical distortion. The planning target volume (PTVFB and PTVDIBH) was generated as an isotropic expansion of the CTVFB and CTVDIBH with a 3 mm margin in all directions, while PGTV was generated as an isotropic expansion of the GTV with a 5 mm margin in all directions. The OARs of this study were contoured on the CT image, which included the lung, heart, spinal cord, breast, liver, thyroid, esophagus, and trachea. Patients were treated either with hypo-fractionated therapy with simultaneously integrated boost (2.67 Gy for PTVFB and 3.2 Gy for PGTV in 15 fraction) for FB or hypo-fractionated therapy for DIBH (2.67 Gy for PTVDIBH in 15 fractions).

The AlignRT system (Vision RT, London, UK) employs a combination of light projectors, and the position of the patient is monitored with three cameras that generate a 3D map of the patient’s topography. Moreover, the system consists of software and a computer workstation, does not require the use of body film, and produces no irradiation during the imaging process. DIBH and FB patients were set up and monitored throughout treatment using AlignRT in real-time mode. In real-time mode, AlignRT displays three axis linear translations (vertical, lateral and longitudinal), the root mean square of the linear translations (RMS), and three axis rotations (yaw, pitch, and roll) (Fig. 1). The tolerance of linear translations and rotations is set to 3 mm and 3˚ based on manufacture recommend, respectively.

An example of the AlignRT monitoring screen during treatment

For both DIBH and FB treatments, the SGRT workflow consists of initial setup in the AlignRT system, and preparation before DIBH and FB treatment and daily treatment (Fig. 2). The workflow of DIBH is the same as that published by our group in other, previous studies [15, 16]. First, for the DIBH and FB treatments, import the DIBH and FB body contour into the Align RT workstation, delineate the surface-monitoring region for the initial setup position. Second, the AlignRT and CBCT are used for daily patient setup and to assess the agreement of AlignRT with CBCT. Based on the system prompt for setup errors, manually adjust the rotational direction by ≤ 3°, and then perform linear translation by ≤ 3 mm. Third, acquire the CBCT images and record the deviations from the XVI workstation. Then, shift the couch based on CBCT registration and capture the present surface image as a reference image. Finally, turn on the gating switch in the Align RT workstation and activate the Elekta Response controller to monitor patient respiratory motion during beam delivery.

The workflow of DIBH and FB with SGRT

Figure 3 shows the result of a typical breath-hold session with DIBH treatment and a free breath session with FB treatment, as tracked in the AlignRT system in a vertical direction and as printed from the system’s session reports, respectively. The shaded areas of Fig. 3 (A) indicate automatically gated beam hold when predetermined tolerance limits (± 3 mm) are exceeded.

The result of typical session as tracked in the AlignRT system in vertical direction: A DIBH treatment; B FB treatment

All clinical treatment plans were generated using Pinnacle TPS (version 9.10, Philips Radiation Oncology Systems, Fitchburg, WI, USA). Intensity modulation was performed using the direct machine parameter optimization (DMPO) algorithm. The collapsed cone (CC) algorithm was applied for final dose calculations, with a grid size of 3.0 mm. For the DIBH group, all plans used the tangential field-in-field (TFiF) technique, and treatments were performed with an Elekta Infinity linear accelerator (Elekta, Stockholm, Sweden) using 6 MV photons. Moreover, to be eligible for DIBH treatment, patients must be able to hold their breath for at least 25 s and demonstrate a stable breath-hold position. The Infinity linear accelerator is equipped with a multileaf collimator, which has 40 leaf pairs of 0.5 cm thickness. The TFiF treatment plan consists of two opposing tangential fields with gantry angles between 300˚ and 315˚ for the medial beam and 120˚ and 135˚ for the lateral beam, with two or three sub-segments included. For the FB group, all plans had two arcs, with an angle ranging from 181˚ to 30˚ for the right-side breast cancer patients and from 330˚ to 179˚ for the left-side breast cancer patients, respectively. The mean deviation and standard deviation in three directions of DIBH and FB were acquired from the AlignRT system during beam-on time. Then, eight setup variations with respect to the ± 95% confidence interval of deviation distribution (the mean deviation of three directions ± (1.96*standard deviation)) were introduced for each reference DIBH and FB plan to generate eight new plans, shifting the isocenter from its reference position in three directions. If the original isocenter is (x, y, z), the mean and standard deviations are a ± a1, b ± b1 and c ± c1 in three directions, respectively, the new eight plans as detailed in Table 2. A total of 296 perturbed plans were recalculated with these new isocenters and without changing any optimized parameters compared with the original plan (planorg), four groups of plans, DIBHmin, DIBHmax, FBmin and FBmax, were selected according to the maximum and minimum deviations of dosimetric parameters.

The dose constraints for the PTV were 1) D95 ≥ 100% of the prescribed dose, and 2) D2 ≤ 110% of the prescribed dose. For the OAR, both the DIBH and FB plans met the dose volume limits, as detailed in Table 3 [17, 18].

The dosimetric quality of the treatments was measured using a dose-volume histogram (DVH). For CTV, the target coverage (D95, D98, D99, Dmean, D50, and D2) and the conformity index (CI) were reported [19,20,21].

CI was defined as

$$ } = \left( }_}}} } \right)^ /\left( } \times }} \right) $$

(1)

where PV is the volume covered by the prescription isodose. The CI values range between 0 and 1, and a CI close to 1 represents better conformity. Furthermore, dosimetric parameters were evaluated for the lung, heart, spinal cord, breast, liver, thyroid, esophagus, and trachea. The dose administered to the ipsilateral lung was evaluated using V5, V10, V20, and the Dmean, and for the contralateral lung using V5, V10, and the Dmean; the Dmax of the spinal cord was also recorded. For the heart, the V5, V10, V20, and the Dmean were scored; Dmax and Dmean for the esophagus; Dmean for the thyroid; Dmean for the trachea; and V5 and Dmean for the liver. Dx represented the dose (in Gy) received by x% of the volume, Vy the volume (in percentage) receiving y Gy, Dmax the maximum dose, and Dmean the mean dose.

Datasets were statistically analyzed using SPSS 19.0 software (IBM, New York, USA). The dosimetric parameters of the PTV and OARs were compared using the Wilcoxon Cox test. A p-value < 0.05 indicates statistically significant differences.