Real-time intra-fraction motion management in breast cancer radiotherapy: analysis of 2028 treatment sessions

Over the last decade, modern radiotherapy (RT) techniques including intensity modulated radiation therapy (IMRT), volumetric-modulated arc therapy (VMAT) or hypofractionated radiotherapy regimens have been introduced in breast cancer irradiation. These technological and technical improvements in RT require an accurate and reliable patient positioning. [1] Inter-fractional variabilities in patient positioning are typically handled by X-ray images acquired with electronic portal imaging devices (EPID) before a treatment session additional to laser-assisted positioning. As a consequence patients are exposed to additional ionizing radiation. [2‐4]

Patient movements during breast cancer radiotherapy and especially during dose delivery have been limited in scope in clinical context, although it might have an impact on adequate planning target volume (PTV) setup margins.[5, 6] Nowadays, optical surface imaging offers the possibility to monitor patient movements in real-time using a non-invasive approach without any additional radiation exposure. The system offers a concrete option for safe patient positioning and has been analyzed by several study groups. [7‐9] Crop et al. showed that the optical system Catalyst™ was superior to laser-assisted positioning. In addition, Stieler et al. reported a reduction in cone beam computed tomography (CBCT) scans needed in patients with a fixed tumour to the surface relationship.[10, 11] In regard to intra-fractional motion, Ricotti et al. recently reported results from breast cancer RT using an infrared light system with surface skin markers. The mean baseline drift was 0 ± 0.7 mm in right-left, − 0.5 ± 1.7 mm in inferior-superior and − 1.4 ± 1.8 mm in posterior-anterior direction. [12] Furthermore, Gaisberger et al. used 3D body surface imaging and noticed intrafractional shifts of 1.2 ± 0.7 mm during the first 2 min of observation. [13]

The aim of the present study was to quantify the uncertainties of possible intra-fractional chest motion during breast radiotherapy by using the real-time optical surface imaging system Catalyst™.

Between October, 2016 and June, 2017, 104 women diagnosed with breast cancer, who received postoperative RT at the Department of Radiation Oncology, University Hospital, LMU Munich were consecutively recruited in a prospective study. RT-related information, tumour and patient characteristics of each patient were retrieved from medical records. The conventional fractionated RT scheme to the breast or chest wall was 2Gy to a total dose of 50Gy in 25 fractions, a sequential boost to the tumour bed was applied with a dose of 5× 2Gy or 8× 2Gy. Hypofractionated treatments were applied with a dose of 2.67Gy per fraction to a total dose of 40Gy. Patients who received RT in deep inspiration breath hold (DIBH) were excluded from the present analysis.[14, 15] The study was approved by the local ethics committee of the University Hospital, LMU Munich (No. 352–16 ex 09/2016) and registered at German Clinical Trials Register (DRKS-ID: DRKS00011407). Written informed consent was obtained for all patients.

For each patient a planning CT scan using a Toshiba Aquillion LB CT Scanner (Toshiba Medical Systems Corporation, Japan) was carried out and target delineation and treatment planning were performed. The clinical target volume (CTV_{breast/chestwall}) encompassed the chest wall or the glandular breast parenchyma. The planning target volume (PTV_{breast/chestwall}) was obtained by adding a 5 mm margin to the CTV_{breast/chestwall}. In cases where a boost was applied, CTV_boost included the tumor bed, visible surgical clips and anatomical distortion. The PTV_boost was generated using a 5 mm isotropic expansion on CTV_boost. Patient set-up was in a supine position on a positioning device (WingSTEP™, IT-V, Austria) with both arms elevated above the head. 3D conformal radiation therapy was used. Treatment planning was performed using the Oncentra 4.5.2 software (Elekta AB, Stockholm, Sweden). All plans consisted of two opposing tangential beams for the breast/chest wall with the addition of some subfields to increase dose homogeneity, as well as anterior/posterior fields for regional nodal irradiation (RNI). RNI included lymph node levels IV, III, Rotter lymph nodes and some parts of lymph node level II according to the ESTRO-guidelines. [16]

Before the first treatment fraction, patients were positioned using the skin marks and a laser-alignment-system followed by the acquisition of iView™ portal images (Elekta AB, Sweden) to verify proper positioning. Furthermore, a surface reference image of the region of interest was acquired using the Catalyst system after the patient had been finally positioned.

The present study faces the issue of quality assurance during breast irradiation: for this aim, intra-fractional monitoring of the patients’ surface was evaluated using an optical non-invasive method. Moreover, to our knowledge, the present experience reports and discusses the data of the largest patient population (104 patients and over 2000 treatment sessions) in the setting of intra-fractional motion management by using an optical surface imaging system for breast cancer RT. Since the maximum single deviation during treatment delivery may be interpreted as a “worst case” deviation during a single treatment session and could influence adequate coverage within the applied setup margins (5 mm CTV to PTV), we have analyzed them separately from all the other samples recorded during every treatment session.

During beam-on time only a maximum deviation (magnitude of vector calculated as shown in Fig. 4) of mean 1.93 ± 1.14 mm (95%-CI: 0.48–4.65 mm) was found. Along the lateral axis the deviations were smaller (95%-CI: -1.88 – 2.1 mm) than on the longitudinal (95%-CI: -2.89 – 3.11 mm) or vertical (95%-CI: -2.9 – 3.55 mm) axis. Therefore, we were able to show in the present study that even this “worst case” momentary deviation was within the limits of applied setup margins, while most of the time, deviations were abundantly within the PTV margins. In fact, all samples during beam-on time, reported a lateral deviation of mean 0.08 ± 0.65 mm, longitudinal mean 0.09 ± 0.81 mm and vertical mean 0.39 ± 0.98 mm. The cumulative distribution functions for the spatial axes and the deviation vector (Fig. 4) gives an overview of the deviation distribution.

Various techniques for intra-fractional motion analysis in breast irradiation are reported and discussed in literature. X-ray imaging as described by Hirata et al. represents one of the options. [19] The authors used orthogonal X-ray imaging on 23 patients before and after a treatment session to observe internal target motions referring to surgical clips displacements in accelerated partial breast irradiation. The magnitude of intra-fractional motion was depending on the direction, e.g. it showed a systemic baseline drift of 1.5 mm in the posterior direction but also depending on patients’ characteristics and tumour cavity location. Yue et al. pursued a similar approach by comparing marker- and bone-based matching resulting in an average intra-fractional motion magnitude of 4.2 mm vs. 2.6 mm. [20] Later, Yue et al. described tumour side dependent directional motion along the lateral axis between left- and right-sided breast cancer patients. [21]

Other authors applied continuous portal imaging with a sample frequency of 7.5 Hz during the delivery of two tangential fields in free-breathing modality on breast cancer patients. The amplitude of the intra-fractional chest wall motion had a mean value of 2.0 ± 0.7 mm and a maximum value of 8 mm. [22] The main disadvantages of X-ray imaging in motion analysis are represented by the additional radiation exposure and the restriction to bone- or marker-based matching strategies. Especially in the case of bone-matching, results may differ from the actual tissue movement. Nevertheless, the obtained results remained within 1 cm in any direction and therefore comparable to the results of the experience reported in the present study.

Another strategy was applied by the group of Kinoshita et al., who positioned a gold marker near the nipple on the skin of breast cancer patients and used a fluoroscopic real-time tracking system for monitoring. Baseline shift and range of motion stayed within a few millimeters. [23] An optical approach method to measure intra-fractional motions was recently reported by Ricotti et al. [12], who used an optical system with infrared reflective surface markers for motion tracking. Based on their high sample frequency, the investigators were able to divide the motion into a baseline drift and respiratory-caused changes. Measurements were quite time-consuming because a calibration before every session was required and patients needed several surface markers positioned on their chest wall. In contrast, in our here reported experience, no additional markers were required. However, the findings reported by Ricotti and colleagues seem to be comparable to our analyses; the authors calculated a baseline drift of 0 ± 1.1 mm in the lateral direction, − 0.5 ± 1.7 mm in the longitudinal and − 1.4 ± 1.8 mm in the vertical direction. [12]

Regarding the influence of duration of the RT session, Wiant et al. analyzed data of 33 patients during 831 monitoring sessions. Most patients stayed within a 5 mm drift range. [24] However, the authors reported a strong correlation between patient drift and duration of treatment session. It is noteworthy that in contrast to the present study, the mean observation time was nearly twice as long (mean 342 s vs. 154 s). This underlines the fact, that treatment duration might strongly influence intra-fractional baseline drifts. Similarly, Jensen et al. used a self-developed laser-system for measuring displacements along the anterior-posterior (AP)-direction for intrafraction motion in breast cancer RT, showing a baseline drift in the posterior direction of − 1.3 mm at the end of the treatment session. Approximately 4% of treatments had a larger baseline drift than 5 mm at 5 min. [25] In our series, the mean treatment session time was very short with 154 s (95%CI: 79–268 s), thus resulting in a very weak correlation between duration of RT session and the maximum deviation of displacement.

As a more modern approach, Acharya et al. analyzed intra-fractional motion using magnetic resonance imaging (MR) guided-IMRT and calculated an intra-fractional tumour bed cavity displacement of about 0.6 ± 0.4 mm along the longitudinal and vertical axis. According to their results, the mean difference between planned and delivered dose was within 1 %. [26] Similarly, another group used MRI with two time scales (2D and 3D MRI) showing a median deviation of about 2 vs. 2.2 mm. [27] MR-imaging during RT represents an elegant and promising approach to track target volumes in real time, especially in soft tissues including breast. Nevertheless, compared to other here described optical tracking systems, these image guided systems still remain very expensive and currently not widely available in clinical practice.

One of the limitations of the present study is represented by the increased deviation vectors during gantry rotation. At certain angles, the gantry can mask one or two of the three cameras of the Catalyst system, causing misleading measurement outputs. This situation can be observed in higher deviation values during the whole treatment session time as compared to the results of the beam-on time only (see Table 2). Nevertheless, considering that only patients receiving 3D conformal RT, with a fixed gantry position during beam-on time, were recruited in the present study, this technical issue can be estimated of minor relevance for the present analysis.

In summary, intra-fraction motion during 2028 sessions of breast cancer RT observed in the present study was maintained within 5 mm in any direction as the confidence intervals show. Although specific advices on the size of PTV margins are not provided by current guidelines (since they should be based on actual measurements of set-up performance), the here reported deviations are within the clinically most common standard PTV-setup margins of 5 mm. [16] One further aspect of implications of the present findings for treatment planning could be the skin flashing of treatment fields in breast IMRT. Intensity extension with the help of auto-flash tools usually aims to account for intra-fractional patient motion or set-up errors. Based on the present results, typically used margins of up to 2.5 cm could be substantially reduced. [28]

Due to the large number of data analyzed (2028 sessions in 104 patients) and the detailed reported results of specific deviations (< 5 mm), the optical real-time surface imaging system utilized in the present study showed to be an important tool for intra-fraction motion management in breast cancer RT. Not less important, the here discussed surface imaging system with visible light represents the peculiarity to be absolutely safe for patients, as it is a non-invasive approach to monitor patients position in real time without any additional radiation exposure.