Target motion management in breast cancer radiation therapy

A promising solution for intrafractional motion monitoring in the chest wall irradiation and breast cancer RT is optical surface imaging.26 Using three optical cameras and light projectors, the 3D map of a patient’s topography is generated and allows visualization of the patient in any position or gantry angle (Figure 1).27

Figure 1

Surface imaging provides mobile target monitoring in the case of breast irradiation. Surface imaging is characterized by easy utilization and high temporal frequency without further radiation dose to the patient.26 It can be matched with a variety of RT techniques (for example, breath-hold and respiratory gating) to reduce setup uncertainties during delivery, which can lead to a reduction in target margins and nearby sparing. Several studies have shown that surface guidance for intrafractional monitoring was mainly utilized for breast breath-hold RT.28,29 Additional benefits of surface imaging include (1) reducing interfractional setup error, (2) monitoring intrafractional motion, and (3) using less invasive patient fixation than other immobilization techniques, and more comfortability of patient as well.30 However, surface guidance comes with some limitations. The visibility of the patient’s skin surface for surface imaging is essential. Therefore, there is a compromise between surface imaging ability and the degree of immobilization. Also, any obstacle on the skin can lead to impossible reflectivity and restricting the function of surface imaging. An important limitation of surface imaging relevance to target localization is insufficient adaption between the external and internal surfaces. However, in breast cancer RT in which the external surface is the target surface, this problem becomes less important.26 Nonetheless, surface-guided RT (SGRT) technology enables adaptive radiation therapy (ART) in which a motion history related to the patient is applied to perform narrower margins in the next following treatment fractions. Current applications of real-time surface imaging rely on breath-hold, respiratory gating, and tumor tracking deliveries.31

Accessibility of the breast (compared to deep organs such as liver or prostate) and typically shallow-seated targets, facilitate the application of internal markers.32 Additionally, breast motion is well characterized by external markers.33 Internal/external markers result in superior performance compared to the surgical clips in terms of both accuracy and detectability on kilovoltage (kV) images.34 Organ displacement and real-time localization during beam delivery can be directly evaluated by employing external surrogate and/or internal radio-opaque fiducial markers. The fiducial marker tracking technique was first introduced for conventional RT and later for particle therapy.32 Target motion tracking using internal markers is usually combined with more than two fluoroscopic imaging examinations. The fiducial markers are implanted near to or inside of the target. Markers (or surgical clips) are usually made from high-Z material such as gold, platinum, carbon-coated zirconium oxide to be visible in X-ray images.35

Using markers for motion monitoring in breast cancer, Kinoshita et al. showed the median range of respiratory motion is 1.0 ± 0.6 mm, 1.3 ± 0.5 mm, and 2.6 ± 1.4 mm for the RL, CC, and AP directions, respectively. The range of motion was the largest in the AP direction in all cases.23,36,37,38 In a work by Korreman et al., it was reported that variability in motion patterns for target and surrogate using an internally placed gold marker and a reflective marker implanted on the chest wall can be considerable.39,40 However, the difference between the surrogate marker position and the real tumor position in breast cancer is not a shortcoming as of other organs, mainly due to a good correlation between tumor displacement and that of the markers.

While fiducial markers find a wide range of applications in breast cancer due to the existing well signal correlation between tumor site and marker location, their usage is hampered by (1) the invasive nature of marker implantation, (2) possible displacement of the markers even more than few millimeters for tumor volumes far from the skin, (3) lack of volumetric information about anatomy deformations close to organ-at-risks (OARs), and (4) ionizing radiation imaging needed to localize them. Marker displacement from the implanted place, tumor deformation, and tumefaction of surrounding tissues are common reasons leading to such positional error41,42 Artifacts in computed tomography (CT) images caused by high-Z fiducial markers are also problematic.43 Electromagnetic

transducers/transponders (ET) are alternatives to high-Z internal markers providing continuous real-time 3D localization of the target without radiation imaging.26 The Calypso system detects the fiducial marker location in real-time without X-ray imaging.44 Commonly, three transponders with a variety of resonance frequencies (300-500 kHz) are placed in or close to the tumor. While the implementation techniques for ET are feasible and safe, they cannot be standalone. Several works indicate that interfractional variations of transponder location are significant and therefore hybrid real-time monitoring, for example, real-time tumor tracking is recommended.45,46

4D CT provides a high spatial and temporal resolution image of the thorax region during the planning phase to construct the breathing modeling used for managing respiration-induced motion. In other words, 4D CT enables 4D treatment planning. In 4D CT, the respiration cycle is first monitored by an external indicator such as real-time position management (RPM) system followed by dividing the cycle into several gates. Richter et al. showed motion amplitude of the chest in the 4D CT scanning is about 1.8±0.9 mm and target coverage was decreased by < 5%, caused by breathing motion.47 4D CT imaging/respiratory-correlated CT procedure is a promising solution for obtaining a time-resolved CT image at the cost of a substantial increase in radiation dose.48,49,50,51,52,53

Chan et al. showed a better estimation of the real amount of heart in the radiation field is possible using 4D CT imaging of the patient with breast cancer.54 Qi et al. assessed respiration-induced heart motion by proposing two indices, the maximum heart depth (MHD) and the depth of the left ascending aorta (DLAD) extracted from the 4D CT dataset. They showed the dosimetric variation of the heart is highly correlated with these two metrics in gated RT for the left-sided breast cancer. Larger respiration-induced heart displacements (nearly 1 cm) are observed based on 4D CT scans. Also, a mean maximal dose to the left ventricle reduced from 49.14 (3D conformal RT (CRT)) to 33.97 Gy (intensity-modulated RT (IMRT)) when 4D CT imaging is used. The findings illustrated the potential use of 4D CT-based planning for cardiac sparing.21 In a similar work, Yue et al. showed the changes (the difference between 4D and conventional plans) in D95, D90, V100, V95, and V90 of the target volume were -5.4%, -3.1%, -13.4%, -5.1%, and -3.2%, respectively.12 In addition, V100 decreases from 81.8% in the conventional plan to 74.9% in 4D CT-based planning.12 For evaluating cardiac sparing in tangential breast IMRT, Mahmoudzadeh et al. modeled the breathing-induced motion with deformable registration using 4D CT imaging in RT simulation in order to calculate accumulated heart dose for robust optimized and clinical plans.55 Compared to the regular CT, the main drawback of 4D CT imaging for RT is the added radiation dose to the patient. The extra dose from the 4D CT imaging can be compensated by a substantial reduction of the RT dose to the OARs.55

Recently, 4D magnetic resonance imaging (MRI) has been used to estimate respiratory motion variations and as a procedure to complement and support 4D CT enabling 4D RT planning and simulation.56 Owing to superior soft-tissue contrast and radiation-free imaging features, MRI allows frequent multiple data acquisitions than CT. Due to limited time resolution associated with true 4D MRI, 2D cine-MRI is suggested.57 Individualization of planning target volume (PTV) margin based on cine MRI data in the simulation seems to be a promising solution for the intrafractional motion problem.58 Respiratory-correlated 4D MRI has attained more interest as an alternative to 4D CT for the measurement of respiratory motion.59 Cai et al. presented the feasibility of 4D MRI using an image-based respiratory surrogate in the planning phase.60 They investigated the accuracy of 4D MRI for motion measurement using 4D phantoms, for example, XCAT in terms of stability. Moreover, motion tracks can be estimated based on 4D MRI and 2D cine-MRI with an acceptable difference in motion amplitude up to -0.3 ± 0.5 mm.60 4D MRI provides an estimation of the respiratory motion for the two human subjects as much as 0.88 and 1.32 cm.60 Also, Hu et al. showed a respiratory amplitude-based system to guide 4D MRI image acquisition is more robust to control irregular breathing compared to phase-based ones.61

Oar et al. performed a comparison between 4D CT and 4D MRI data quality based on the amplitude of motion in abdominal RT planning.52 Motion uncertainty due to respiratory was estimated to be less than 0.2 mm in both the 4D CT and the ground truth; the median amplitude of motion was 11.2 mm and 10.1 mm for 4D CT and 4D MRI, respectively.62 Paganelli et al. showed that the 4D MRI sequence enables describing organ motion and reduction of safety margins in RT planning.63 Hurst et al. developed and optimized 4D MRI based on respiratory triggering using an external surrogate for abdominal tumors.64 They concluded that any irregularity in patient breathing significantly affects 4D MRI performance. In addition, irregular and slow breathing rates deteriorate 4D MRI efficiency. A limitation of 4D MRI is, however, being sensitive to the change of breathing pattern between the preparation and acquisition periods. In addition, low temporal resolution is another limiting factor resulting in frequent scanner halts when breathing is irregular.61 Long scan time is also uncomfortable for the patients. However, a reduction in acquisition time in a high field 4D MRI scanner is expected.64

Gantry-mounted X-ray imaging refers to those X-ray imaging modalities mounted on the treatment gantry allowing monoscopic and stereoscopic X-ray imaging. Portal imaging using electronic portal imaging devices (EPID) is a popular example of gantry-mounted imaging. Beam’s eye view (BEV) portal imaging also enables real-time target motion tracking. Portal imaging is acquired with the help of the therapeutic megavoltage (MV) beam. Recently, gantry-mounted kV X-ray radiographic/fluoroscopic imaging is also available by either kV X-ray tubes or reduction of linac beam energy from MV to kV ranges.47 The Vero, ExcaTrac, and CyberKnife systems offer stereoscopic imaging using two kV sources coupled with two flat-panel detectors.26

The acquisition of portal imaging is proved to be fast as well as easy to use in order to measure patient movement during breast cancer RT.65 Richer et al. presented that tracking breast motion in EPID results in patient-specific maximum motion amplitude of from 0.8 to 2.2 mm, 1.5 mm on average.25 In another work, respiratory motion during daily treatment on the CLD was investigated by EPID. The results of their work showed that intrafractional variation in each patient during treatment day was minimal. The daily maximum range for any patient was 0.25 cm.17 For evaluating intrafractional and interfractional motion in breast cancer RT using EPID, Kron et al. concluded that the largest variation is in the CC direction with 1.3 ± 0.4 mm and 2.6 ± 1.3 mm for intrafractional and interfractional motions, respectively.65 In a recent study based upon stereoscopic imaging enabled by the Cyberknife machine, Hoekstra et al. evaluated the effect of baseline and breathing motion on PTV margins for accelerated PBI (APBI) irradiation. They showed that the PTV margin depends on the treatment time.66 However, poor image quality because of dominant Compton scattering in MV beams remains a major problem in portal imaging. Furthermore, according to the AAPM Task Group 75 report, a significant disadvantage of kV imaging-based motion monitoring is the extra dose to the patient, particularly at the skin surface.67 Depending on the imaging technique, a typical dose of 1–3 mGy per image is delivered in any kV imaging.26

Rapid imaging along with no ionizing radiation makes ultrasound (US) imaging suitable for estimating intrafractional motion during the planning and simulation phases. The real-time US is also of interest in breast imaging mainly due to the lack of bony structures and also easy accessibility of the organ.26,68 4D US provides almost real-time 3D rendered image data and is considered as a basis of multidimensional imaging of the breast.68 In addition, 3D/4D US of the breast provides diagnostic information of the coronal plane.68

US imaging typically provides good soft-tissue contrast and therefore allows contouring breast tumors. Furthermore, imaging artifact limits the application of real-time US imaging.68,69,70 Because of its manual operation, the image quality is also user-dependent as well.68 Despite well-established applications of US in diagnostics, target delineation, and pre-treatment localization, the use of real-time US imaging for intrafractional motion estimation and mitigation for breast cancer is limited and there is no commercially available system. The only commercial US system is Clarity Autoscan (Elekta) for monitoring intrafractional motion26 that is approved specifically for prostate and/or prostate bed RT. However, Wong et al. applied the Clarity system to breast imaging to evaluate the error between the Clarity and pre-treatment CT images and observed that the errors are clinically insignificant.71 However, in the era of surface imaging, the US methods cannot hold great advantages over ultrasound techniques for estimating breast intrafractional motion.72

In the previous section, the main motion monitoring techniques of breast target were presented. The next step in the RT workflow is to assist the irradiation of mobile targets with motion monitoring data. Common irradiation approaches addressing the respiration-induced intrafractional motion in breast cancer treatment include breath-hold, respiratory gating, and real-time tumor tracking techniques. The influence of intrafractional target motion is of particular concern in APBI due to high doses per fraction, particularly for target volumes close to inhomogeneities (i.e., skin or chest wall).73,33 Therefore, motion mitigation techniques have to be perused in such treatment options.

Breath-hold techniques refer to the management of target motion from the patient side. The deep-inspiration breath-hold (DIBH) method is a practical and easy-to-use solution for breast cancer RT.6 During inhalation, the diaphragm moves the heart posteriorly and inferiorly away from the breast leading to a potential reduction of both heart and lung toxicities.16 As illustrated in Figures 2 and 3, the major role of DIBH in motion-addressed breast cancer RT is increasing the distance between tumor volume and the heart leading to less dose to the heart and therefore a lower rate of toxicity.74,75,76 DIBH is always linked to the beam gating to repeatedly on and off the irradiation beam based upon the patient respiratory cycle.

Figure 2

Figure 3

The DIBH for breast cancer RT is mostly employed in two manners: (1) moderate DIBH and

(2) voluntary DIBH (vDIBH).76,77 The former is also known as active breathing control (ABC) in the literature.79 ABC uses special devices to control airflow during the respiratory cycle77,78, while in vDIBH the patient is partially freely breathing. A decrease in the mean heart dose and the left artery dose to about 67% and 73%, respectively, is observed when using the ABC for breast cancer RT.76 In addition, the ABC devices allow a reduction in setup uncertainties to less than 2 mm.76 The vDIBH is sometimes used in conjunction with respiratory motion monitoring to capture breath function at certain points in the breathing period. As for the ABC, the vDIBH decreases the time for RT simulation and daily setup.76,79 In contrast to ABC, vDIBH offers more patient comfort while it is also inexpensive.75,79 Recently, the DIBH treatment using volumetric-modulated arc therapy (VMAT) is utilized for a patient with the left-sided breast cancer to irradiate both whole breast and regional node with superior target coverage and good cardiac sparing.80,81

Fassi et al. investigated target position reproducibility in the left-sided breast irradiation with DIBH using multiple optical control points. They compared the performance of optical surface imaging with that of the RPM-based methods and showed that the use of multiple surface fiducials leads to improved target and surface reproducibility.82 Betgen et al. reported a systematic interfractional translation up to 5 mm and intrafractional errors of about 1.4 mm during voluntary DIBH using 3D surface imaging in patients with the left-sided breast cancer.83 Borst et al. quantified the influence of breathing with DIBH in breast cancer RT. The percentage of the left ventricle (LV) irradiated volume was 28% and 71% for DIBH and free-breathing (FB), respectively.84

An efficient method of dealing with moving targets is to gate the radiation field. Respiratory gating refers to the management of target motion during treatment by rapid beam switching within the breathing cycle synchronized with an internal/ external tracking system. Respiratory gating is usually implemented in two fashions: phase-based and amplitude-based gating. The former is accomplished by defining a set of phases (gates) over a complete breathing cycle. The irradiation beam is on in only one or few gates. In contrast, the latter is performed by setting a threshold value on the amplitude of the respiratory signal. Once the respiration signal falls below the predefined threshold, the irradiation beam is on. In a small gating window, the phase-based gating method can result in missing the tumor caused by interfractional position variations.

In contrast to the DIBH, the patient freely breathes while being irradiated with the therapeutic beam in respiratory-gated RT. Therefore, more patient comfort is obtained with respiratory gating.85,86 Korreman et al. highlighted the dosimetric advantages of free-breathing gated breast cancer RT over vDIBH in terms of cardiopulmonary dose sparing.73 Giraud et al. conducted a multicenter prospective study to compare respiratory-gated RT with conventional CRT for patients with breast cancer. They observed a significant reduction in lungs and cardiac toxicities when using the respiratory gating method.87 Also, Qi et al. reported that the median heart volume receiving at least 50% of the maximum dose was decreased from 19.2% for free-breathing to 2.8% for end-inspiration gating. A substantial coronary artery volume sparing patients with the left-sided breast cancer was also observed. In addition, for both the right- and left-sided breast cancers, the median lung volume receiving 50% of the prescribed target dose reduced from 45.6% for free-breathing to 29.5% for inspiration gating.21

Respiratory gating results in two clinical benefits: (1) acceptable levels of target dose conformity and (2) OARs/normal tissues sparing. There are, however, several challenges associated with respiratory gating mandating further researches. First, time latency at the gating process has a result in underdosage and overdosage of proximal tissue. Thus, a successful gating process needs to minimize time latency during the gating window. Another challenge is a long treatment time by respiratory gating. The longer treatment time is inconvenient for the patients and can result in respiratory pattern variation, such as shift motion.31 Another noticeable challenge for gated IMRT delivery is an increase in delivery time. The low efficiency of gated IMRT, as a product of the IMRT efficiency (20% to 30%) and the gating duty cycle (20% to 30%), results in a 10 to 25-fold increase in delivery time than conventional CRT treatments.88

To obtain benefits of the respiratory gating method, higher temporal resolution, higher soft-tissue contrast, and lower radiation exposure imaging techniques in the RT planning are mandated.67 In some cases, however, motion occurs within the gate window, called residual motion.88 Therefore, there is always a compromise between the amount of residual motion and the duty cycle to search for optimal gating parameters.89 As heart dose automatically leads to an increase in cardiac mortality90, a key question in respiratory gating is, therefore, the selection of optimal gating window parameters. Many studies have proved that the end of inspiration is optimal in terms of heart and lung tissue sparing in the left-sided breast cancer RT.74,21 While the absolute lung volume irradiated is largest in respiratory-gated breast RT, the relative lung volume is smallest in the inspiration phases. Thus, the inspiration phases are optimal for beam gating in breast cancer RT by providing the longest distance between the breast and heart and also minimizing the lung density.74 Although not implemented yet, respiratory gating based on the data from real-time cine MRI data would be a solution for online motion mitigation.

Real-time tumor tracking is generally performed by either robotic radiosurgery, dynamic multi-leaf collimators (DMLCs), or couch movement.91 Owing to the benefits of stereotactic body RT (SBRT), Cyberknife APBI can be considered as a real-time tumor tracking mitigating the intrafractional respiratory motion.92 Methods like kV/MV radiographic imaging with and without markers, US imaging, portal imaging through EPID, kV/MV imaging are real-time tumor tracking methods. A combination of imaging methods with DMLCs (called dynamic IMRT) results in a solution for real-time tumor tracking.93

In breast cancer RT, real-time tumor tracking results in a substantial reduction in the volume of the heart receiving a high radiation dose.93,94 Continues portal imaging during RT has shown promising results for estimating intrafractional chest wall motion of patients with breast cancer by providing time-resolved visualization of the internal organ from BEV.95 As an estimate, Hijal et al. showed the irradiated volume of the heart of 30 Gy (V30) is 0.03% and 1.14%, and the mean heart dose is 1.35 Gy and 2.22 Gy, for real-time 3D CRT and static 3D CRT, respectively.96 Leonardo et al. showed that real-time tumor tracking leads to significant heart sparing in a prone position in APBI and provides a daily precision treatment while reducing clinical target volume (CTV) to PTV margin.93 In addition, in patients with abnormal anatomies as the significant volume of the heart may be irradiated, real-time tumor tracking would be useful to avoid extreme doses.97

MLC tracking has been successfully performed for IMRT and VMAT deliveries to address intrafractional target motion.98,99,100 Dynamic IMRT enables dynamically reshaping the treatment field in the BEV based on the actually recorded target motion.101 Furthermore, real-time tumor tracking with IMRT delivery resulted in better cardiopulmonary sparing and improved target coverage for breast cancer treatment.102,103 While the dynamic IMRT provides a highly conformal dose distribution, it is usually challenged by the interplay effect that occurs in the time between leaf and the target motions. The interplay effect automatically leads to motion artifacts in dose distributions.104,105 Synchronization of real-time tumor tracking based on two sets of fluoroscopy and IMRT delivery is also feasible but at the expense of non-negligible skin surface dose.106 Real-time tumor tracking could also result in a percentage depth dose of 58% (at 5 cm) of the peak dose for long IMRT treatments.26 In SGRT-based tumor tracking, beam-on and beam-off delays might play a role and vary between the SGRT system and beam delivery.26 Smaller PTV margins are usually appropriate for patients with breast cancer who are actively monitored with surface imaging during RT.107 Hamming et al. showed that SGRT data correlated well with CBCT data in patients with breast cancer.108 In their study, the left-sided breast cancer was monitored continuously to maintain positional errors within 5 mm with SGRT.108 The combination of real-rime surface-guided DIBH is also successfully implemented in patients with breast cancer, resulting in a reliable and stable DIBH treatment.109

However, some concerns associated with real-time tumor tracking are the resource-intensive nature of delivery and also imposing the amount of additional radiation dose.110 According to the Report of AAPM Task Group 75 67, a typical in-room kV cone-beam CT of the chest (commonly used in the case of breast cancer RT) leads to a maximum skin dose of 85.4 mGy. Real-time CBCT breast imaging results in a dose of 2 mGy and 12 mGy per scan for the right- and left-sided breast cancers, respectively.110 Liu et al. showed that using 4D CBCT, PTV margin would be substantially reduced compared to kV CBCT treatments.111 Real-time imaging during treatment increases RT irradiation time while the patient lies on the couch.67 Real-time tumor tracking increases the complexity of the radiotherapy planning and delivery process, mandating rigid quality assurance at every level for precision and safe treatment.101 Furthermore, the time delay between the real tumor position and the beam positioning system is a major challenge in real-time tumor tracking.16 Besides, cycle-to-cycle fluctuations in the breathing cycle of the patient add complexity to the problem to some extent.112 However, adaptive filter algorithms are proposed to predict tumor position in advance.113

Patient positioning (i.e., supine or prone positions) plays a considerable role in motion mitigation techniques in patients with breast cancer.114 Prone position refers to hanging the breast tissue under its weight through a hole at the bottom of the couch. Prone position improves separation between tumor and OARs as heart and lung for some patients. In addition, the prone position results in fewer respiration movements when compared to the supine position. Furthermore, some prone boards allow regional node irradiation, as well. However, the prone positioning is dependent on the position of the original tumor. In addition, patient setup variations can be significantly larger in prone positioning resulting in an increased interfractional variation.115 In contrast, supine positioning is more common for staff and ease of setup. It can match nodal field to chest wall fields if this requires. Nonetheless, there is a lack of skin-sparing in women with large or pendulous breasts. Therefore, breast support by other devices is sometimes required to anteriorly position the breast away from the heart, lung, and abdomen. Referring to Figure 4, it is proven that the prone setup is more optimal for sparing lung volume compared to the supine position.115,116

Figure 4

Because of a significant decrease in irradiated lung volume and even irradiated heart volume in 87% of all patients with the left-sided breast cancer, the prone position outperforms the supine setup by exhibiting improved dose homogeneity and fewer toxicities. Morrow et al. showed that the respiratory motion of the chest wall substantially decreases from 2.3 ± 0.9 mm to -0.1 ± 0.4 mm in supine and prone positions, respectively. They also showed that the prone positioning of patients for breast irradiation reduces the error introduced by intrafractional respiratory motion.116 Veldeman et al. reported the 2-year better cosmetic outcome of prone positioning in comparison with supine positioning in large-breast patients.117 To summarize, while supine positioning is the ease of setup, it is suboptimal in terms of lung and heart doses in some cases.117

MRI guidance is considered the future of image-guided RT (IGRT).129 Real-time MR imaging is also safe in terms of radiation doses.130 The state-of-the-art MR-linac integration in SBRT can provide tracking of the respiratory motion during the treatment fraction. A present limitation of an integrated MR-RT gantry is the high installation cost that limits its use in clinical practices. Acharya et al. determined intrafractional motion and evaluated delivered dose versus planned dose.131 They demonstrated the mean difference of less than 1% between the planned and delivered dose using MR guidance for APBI delivery (Figure 5). They showed that a reduction in the PTV margin leads to a significant reduction in V50 and V100 for ipsilateral breast cancer MR-guided RT. When no additional PTV margin is applied, the mean cavity displacement in the AP and SI directions reaches 0.6 mm.131

Figure 5

Nachbar et al. in 2019, studied first-in-human APBI performed at a 1.5 T MR-linac for breast cancer using 7-beam IMRT delivery. Additionally, they have also investigated the influence of interactions of the secondary electrons with magnetic field on out-of-field dose.132 Individualization of PTV margin based on cine MRI data from the simulation is also a possible motion mitigation method.133 Although not yet implemented, real-time cine MRI-based beam gating seems also to be a promising solution.133 Despite several advantages of MRI guidance, an open question, however, is a dose uncertainty observed in air-tissue interfaces where secondary electrons slightly contribute to total proton dose delivery.133 Electron return and electron stream effects are two main concerns in treatment planning for a hybrid MR-linac delivery.133 Although some existing challenges such as the selection of suitable coils and the above issues for breast cancer remain, the first breast cancer was successfully treated with a hybrid MR-linac machine using an APBI technique.133 Additionally, the magnetic field has a little negative impact on skin dose in APBI relative to WBI due to the use of smaller fields.134

Artificial intelligence (AI) offers a set of key applications in RT workflow, including image segmentation (target and OAR delineation), image registration, radiomics, treatment response assessment/prediction, and tumor tracking. An interesting study showed that using single radiography, a whole 4D data is feasible to predict tumor movement during the treatment fraction using a deep convolutional neural network (DCNN).135,136 Another role of AI in 4D RT is to create synthetic 4D CT from the 4D MRI dataset in MR-only treatment planning.135 Chen et al. pointed out the usefulness of a deep U-net-based approach that synthesizes on-treatment CT-like images with accurate numbers from both planning CT and on-treatment CBCT. Based on their results, the proposed U-net can increase the accuracy of the CT number of CBCT, which makes possible further quantitative tools of CBCT, such as dose calculation and adaptive treatment planning.137 The uses of AI in dynamic/4D breast imaging, image registration, and automatic cancer diagnosis are attracting a lot of attention.138,139,140

The rescanning (repainting) approach is proved to be effective in managing motion-induced dose uncertainty in actively scanned particle therapy to address the interplay effect.141 However, some repainting methods mandate monitoring patient breathing to provide respiration parameters like period and rate.142 For large target movements (> 5 mm), a combination of the repainting techniques with, for example, respiratory gating and breath-hold techniques lead to a superior outcome in terms of target dose uniformity. It should be mentioned that repainting techniques do not eliminate the use of safety margins entirely covering the target along with its movement extent. A potential pitfall of the repainting approach is a significant increase in total irradiation time.142,143,144 Figure 6 shows the respiratory-correlated layered repainting method.32 An iso-energy layer is irradiated in the gating window. The gating window is then divided into three portions, and therefore the number of rescanning is three.32 While this method is proposed to be applied for lung cancer, its usefulness and applications in APBI are sparse and mandate extra researches.

Figure 6

The term “robust treatment planning” refers to the incorporation of CTV-to-PTV margins into the optimization function during inverse treatment planning in IMRT techniques. The concept of robust treatment planning for breast cancer IMRT is utilized via RayStation TPS, as the sole TPS supporting robust optimization for IMRT.54,145,146,147 Though, studies are shown that internal margin (IM) cannot be entirely eliminated in robust treatment planning.53 Due to some uncertainties in particle therapy, for example, range uncertainty, the definition of simple PTV in particle therapy is suboptimal. Therefore, the role of robust optimization is to effectively address the tumor motion and uncertainties in RT, particularly in particle therapy.145 Robust planning using VMAT delivery for a moving target in the breast generated clinically acceptable plans and was confirmed by real patient CBCT data.147 Not directly applied for intrafractional motion management, the robust optimization for intensity-modulated proton therapy was used to address residual setup errors.148

FLASH-RT refers to ultra high dose RT with treatment time shorter than 0.1 s enabling excellent intrafractional motion management.149 While maintaining local tumor control, FLASH-RT reduces normal tissue toxicity. Despite few clinical devices with the capability to deliver ultra-high dose rates, a lot of preclinical studies confirm the effectiveness of this paradigm-shifting technique.150 In 2019, the first patient with T-cell lymphoma was successfully treated using FLASH-RT with the superior outcome on normal skin and the tumor.151 Despite some technical challenges ahead, the combination of proton therapy (superior conformity) and FLASH-RT (shorter treatment time) can be a viable option for the treatment of breast cancer considering the intrafractional movements.