An investigation of image guidance dose for breast radiotherapy

Cone‐beam computed tomography (CBCT) is used for external‐beam radiation therapy setup and target localization. As with all medical applications of ionizing radiation, radiation exposure should be managed safely and optimized to achieve the necessary image quality using the lowest possible dose. The present study investigates doses from standard kilovoltage kV radiographic and CBCT imaging protocol, and proposes two novel reduced dose CBCT protocols for the setup of breast cancer patients undergoing external beam radiotherapy. The standard thorax kV and low‐dose thorax CBCT protocols available on Varian's On‐Board Imaging system was chosen as the reference technique for breast imaging. Two new CBCT protocols were created by modifying the low‐dose thorax protocol, one with a reduced gantry rotation range (“Under breast” protocol) and the other with a reduced tube current‐time product setting (“Low dose thorax 10ms” protocol). The absorbed doses to lungs, heart, breasts, and skin were measured using XRQA2 radiochromic film in an anthropomorphic female phantom. The absorbed doses to lungs, heart, and breasts were also calculated using the PCXMC Monte Carlo simulation software. The effective dose was calculated using the measured doses to the included organs and the ICRP 103 tissue weighting factors. The deviation between measured and simulated organ doses was between 3% and 24%. Reducing the protocol exposure time to half of its original value resulted in a reduction in the absorbed doses of the organs of 50%, while the reduced rotation range resulted in a dose reduction of at least 60%. Absorbed doses obtained from “Low dose thorax 10ms” protocol were higher than the doses from our departments orthogonal kV‐kV imaging protocol. Doses acquired from “Under breast” protocol were comparable to the doses measured from the orthogonal kV‐kV imaging protocol. The effective dose per fraction using the CBCT for standard low‐dose thorax protocol was 5.00±0.30 mSv; for the “Low dose thorax 10ms” protocol it was 2.44±0.21 mSv; and for the “Under breast” protocol it was 1.23±0.25 mSv when the image isocenter was positioned at the phantom center and 1.17±0.30 mSv when the image isocenter was positioned in the middle of right breast. The effective dose per fraction using the orthogonal kV‐kV protocol was 1.14±0.16 mSv. The reduction of the scan exposure time or beam rotation range of the CBCT imaging significantly reduced the dose to the organs investigated. The doses from the “Under breast” protocol and orthogonal kV‐kV imaging protocol were comparable. Simulated organ doses correlated well with measured doses. Effective doses from imaging techniques should be considered with the increase use of kV imaging protocols in order to support the use of IGRT. PACS numbers: 87.55.Qr, 87.55.ne, 87.53Bn, 87.55.kh


A. Absorbed dose determination
Following the AAPM TG 61 Protocol and the low-energy IPEMB Code of Practice, (23)(24)(25) the absorbed dose to water at the surface of a water-equivalent phantom for the OBI kV beam (Varian Medical Systems, Palo Alto, CA) was determined. A NE2571 Farmer-type ionization chamber (QADOS, Sandhurst, UK) was arranged with the center of its sensitive air cavity placed at the measurement point, the isocenter, using fixed geometry with a field size of 10 × 10 cm 2 . A single exposure is acquired with beam voltage and current-time product of 110 kVp and 20 mAs, respectively. Each irradiation was repeated four times to assess kV source output repeatability. The uncertainty in the determined absorbed dose to water at the phantom surface was estimated following ISO guidelines. (26) B. Radiochromic film measurements

B.1 Film calibration
Radiochromic XRQA2 film (GAF Chemicals Corporation, Wayne, NJ) was used for absorbed dose measurements. Film readout was done using an Epson 10000X flatbed scanner (Epson America, Inc. Long Beach, CA) and the "Epson Scan" software. The following settings were used: reflective mode, positive film with area guide, 48-bit color (RGB), 50 dpi and full pattern. The scanner was allowed to warm up for 10 minutes before scanning to provide a more stable light source and more consistent readings. (27) The film was cut into pieces that were always oriented in landscape mode and aligned in identified locations on the scanner bed. (28) Unexposed film was scanned to provide a baseline for the background noise of the dosimeter and reduce inhomogeneity effects of the film, as well as scanner nonuniformity. The background flatbed scans were repeated five times to account for scanner noise. (29) The preirradiation and postirradiation readings of the XRQA2 films were analyzed using an in-house film analysis program written in MATLAB (The MathWorks, Inc., Natwick, USA). The program extracts the red color channel from the scanned images and calculates optical density (OD) on a pixel-by-pixel basis using a pixel size of 1 × 1 mm 2 . The output is a text file containing the mean OD and its standard deviation from the central 100 pixels of each film, which can be manipulated using Excel (Microsoft, Washington, DC).
XRQA2 films were calibrated by relating their OD to the tube current-time product of the incident beam. The correlation between absorbed dose and OD of the films was established as described above. To obtain the calibration curve, 15 films of 2 × 2 cm 2 were irradiated one at a time, using the tube current-time products of 0, 0.5, 1, 2, 3.2, 5, 10, 20, 40, 80, 126, and 262 mAs, which corresponded to the calculated doses of 0, 0.008, 0.017, 0.032, 0.052, 0.081, 0.161, 0.643, 1.285, 2.52, and 5.24 cGy. Irradiations were performed in static delivery mode, with the films positioned on the central axis of the kV beam at the surface of 10 cm of solid water and SSD 100 cm. The beam energy was 110 kVp and the field size was 10 × 10 cm 2 . Knowing the doses corresponding to tube current-time products, it is possible to determine absorbed dose as a function of the OD of the films.
The calibration films were stored in a dark envelope to develop for 24 hours before postirradiation readout. After postirradiation readout, the mean OD for each film was calculated from the central region of interest.

B.2 Phantom measurements
The standard kV CBCT low-dose thorax protocol (OBI v1.5) was used as the reference CBCT protocol for this study as there is no specific protocol for breast imaging in the current software (Table 1). It has a field of view (FOV) of 45 cm diameter, a field size of 30.3 × 20.6 cm 2 , and uses a full rotation around the patient, which gives a high-resolution volumetric image that, when centered in the patient, will include the breast, heart, lungs, and sternum. (11,30) In many radiotherapy centers, orthogonal kV-kV thorax images are used for evaluation of breast patient setup and thus this protocol was also evaluated.
In order to reduce the absorbed dose from the standard kV CBCT imaging protocol, two new scan modes were proposed ( Table 1). Because of mechanical limitations, the imager can only be rotated 180° above the patient, below the patient, or a full 360° rotation. With the rationale of delivering as little dose as possible to the contralateral breast, the "Under breast" protocol was created with a kV source rotation occurring below the patient (i.e., from 90° to 290° from the vertical axis). The "Under breast" protocol has a FOV of 25 cm and 360 projections ( Table 1). The "Low dose thorax 10ms" protocol was created by taking the standard low-dose thorax CBCT protocol as a baseline and reducing its current time product to half of the original value (Table 1).
An anthropomorphic RANDO thorax phantom representative of a 163 cm tall and 54 kg female (The Phantom Laboratory, New York, NY), sectioned in the transverse plane in 2.5 cm thick slices, was used. The phantom was placed supine on the treatment couch. Absorbed dose was determined for four different organs: breast, heart, lungs, and skin. Dose measurements were performed with each CBCT and the kV-kV image protocol. The female RANDO phantom has a flat chest contours onto which breast sections of different cup sizes can be mounted. Only C-cup breasts were used in this study. The phantom was held together with one strap running in the sagittal plane, with several strips of tape running across the breast and thorax to eliminate air between slices and keep the phantom rigid.
Four batches of 61 2 × 2 cm 2 films were prescanned, irradiated, and scanned again, as described above. Point doses and their corresponding standard deviation were estimated from the dose map in the central 100 pixels region of the film. Films were placed between the phantom slices using a method based on the work described by Brady et al. and Stovall et al. (27,31) Twenty films were located in the lungs, five films in the heart, 20 films inside the breasts, 13 films were distributed on selected points on the skin, and three films were not irradiated and were used as background. The heart location inside the phantom was manually selected based on CT images of real patients and thoracic viscera in supine position images. (32) Within the breasts, more films were positioned in the superolateral quadrant or upper outer quadrant of the breast because this is the most common site for primary and secondary breast cancer. (31,33) On the skin, two films were placed on each side of the armpit to determine an estimate of auxiliary dose, two films were attached at both sides of the manubrium of the sternum, two films were placed in the submammary region, two films were placed on the surface of both outer upper quadrant breasts, two films were placed on the nipples, two films were attached on the back at T7 vertebrae level, and one film was positioned in the intermammary cleft. All film positions are shown in Fig. 1. The anthropomorphic phantom with films in situ was carefully aligned to the isocenter of the Varian 21iX linear accelerator (Varian Medical Systems) using room Axial views ((a) to (f)) of XRQA2 films attached to the phantom slices for measuring dose in lungs, heart, and breasts; film positions in skin ((g) and (h)). Films S1 and S2 were placed in each side of coronal plane, film S3 and S4 were attached at both sides of the manubrium of the sternum, films S5 and S6 were placed on the submammary region, films S7 and S8 were placed on outer quadrant breasts surface, films S9 and S10 were placed on the nipples, films S11 and S12 were attached on the back at T7 vertebrae level, and film S13 was positioned in the intermammary cleft.
lasers. The isocenter was positioned at the phantom center, to center the FOV. In the anteriorposterior direction, the FSD at the midpoint laterally between the breasts and at the midpoint in the superior-inferior direction was 92.5 cm.
The average reading from all films placed in each organ was adopted as the average organ dose. The uncertainties associated with the organ absorbed dose determination were calculated following the guidelines published by the International Organization of Standardization (ISO). (26) The uncertainty in the determined absorbed dose to water at the phantom surface was estimated using the combined standard uncertainty. (26) The total uncertainty for the measured organ absorbed dose would be the combination of: (1) the absorbed dose uncertainty associated with the ion chamber measurements, which comprises the standard deviation of ion chamber readings and electrometer current leakage, uncertainty in air kerma calibration factor, and uncertainty in correction factors ΔD ∕ W,Z , and (2) the absorbed dose uncertainty associated with the film measurements ΔD ∕ ∕ W,Z which are film readout uncertainty and film exposure uncertainty. (1)

C. Monte Carlo simulations
Absorbed doses were calculated using the commercial kV Monte Carlo program PCXMC v2.0. PCXMC is a program developed by the Radiation and Nuclear Safety Authority of Finland (STUK, Helsinki, Finland) that calculates the mean value of absorbed doses averaged over the organ volume and effective doses from medical X-ray examinations, based on the incident air kerma. The dose calculation method in PCXMC is the Monte Carlo method, where 100,000 independent photons are generated to estimate the mean energy deposition values in the organ of the phantom. PCXMC simulates an X-ray beam irradiating a mathematical hermaphrodite phantom that represents a human. The height and mass parameters of the phantom can be varied and matched to data of individual patients. Adjustments of these parameters result in a change of the scaling factors where s z is the scaling factor in the direction of the z-axis or phantom height, and s xy is the scaling factor in the directions of x-and y-axes or phantom width and thickness, respectively. M 0 and h 0 are the weight and height of the unscaled phantom. These scaling factors are then multiplied to all phantom dimensions, and the organ masses are changed accordingly. An adult phantom was selected for simulated dose calculations performed in this project, using the height and weight of the subject represented by the RANDO female phantom.
For the dose calculations, the Monte Carlo program requires input of the anode angle, filtration, tube voltage, tube current, exposure time, imager gantry angle, focus skin distance (FSD), field size, and number of photons for each projection. The anode angle and total filtration were taken as 14° and 2.8 mm Al, respectively. The X-ray tube voltage was 110 kV for all CBCT acquisitions, while for the kV thorax AP and kV thorax lateral left images it was 75 kV and 95 kV, respectively. The distances were extracted from Eclipse (Varian Medical Systems) using a CT scan of the phantom. The beam field sizes were extracted from the CBCT software. The number of photon simulations was set to 105 photons for each setup, with maximum photon energy of 150 keV. The central axis of the X-ray beam was directed to the center of the phantom at nipple height, which was the position of the isocenter in the longitudinal direction used for the film measurements. Additionally, the central axis of the X-ray beam was directed to the middle of right breast, as is usually done for conventional breast radiotherapy treatments. The organs included in the PCXMC phantom were skeleton, heart, lungs, and skin, simulating the RANDO phantom tissue composition. The uncertainty of dose is based on statistical fluctuation, depending on number of simulated interactions in the organs considered.
In order to simulate the CBCT dose from the scans, the following procedure was executed. Low-dose thorax and "Low dose thorax 10ms" protocols with a gantry rotation range of 360° and 655 projections, as well as "Under breast" protocol with a gantry rotation range of 200° and 360 projections, were sampled as discrete beams every 5°. The "Under breast" protocol was calculated at two isocenter positions: with isocenter at image center and with isocenter at center of right breast. Anterior-posterior kV and left lateral kV images were simulated once.

D. Effective dose calculations
Using the CBCT or orthogonal kV-kV protocols, only the tissues and organs covered by the imaging field (skin, breast, heart, and lungs) will be irradiated if it is assumed that radiation leakage outside the imaging field, as well as scatter, is negligible. Hence, the effective dose resulting from a CBCT acquisition will be the sum of all of the weighted equivalent doses in tissues and organs covered by the imaging field. Using the tissue weighting factors published in the ICRP 103 Report (34) (Table 2), the mean effective dose for each imaging technique was calculated as the sum of the weighted effective doses for lung, heart, breast, and skin. The standard deviation of the mean effective dose was derived as the sum of the tissue weighting factors multiplied by the uncertainty in the mean absorbed dose.

A. Radiochromic film measurements
The absorbed dose to the left lung, right lung, heart, left breast, and right breast obtained using film dosimetry for the four imaging protocols is shown in Fig. 2. Using the standard low-dose thorax protocol, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.77 ± 0.05 cGy, 0.78 ± 0.05 cGy, 1.04 ± 0.07 cGy, 0.76 ± 0.05 cGy, and 0.75 ± 0.05 cGy, respectively. Using the "Low dose thorax 10ms" protocol, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.37 ± 0.03 cGy, 0.37 ± 0.03 cGy, 0.52 ± 0.04 cGy, 0.38 ± 0.03 cGy, and 0.37 ± 0.03 cGy, respectively. Using the "Under breast" protocol, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.29 ± 0.03 cGy, 0.30 ± 0.03 cGy, 0.24 ± 0.03 cGy, 0.07 ± 0.02 cGy, and 0.11 ± 0.04 cGy, respectively. Using the orthogonal kV-kV image protocol, doses per fraction for left and right lung varied from 0.25 to 0.05 cGy, and for left and right breast varied from 0.40 to 0.07 cGy. The skin doses to the 13 locations investigated for the four imaging protocols assessed are shown in Fig. 3. Overall, the skin dose per fraction varied, depending on position, between 0.08 cGy and 1.2 cGy. Table 2. Nominal risk coefficients adjusted for detriment for stochastic effects (units 10 -2 Sv -1 ) after exposure to radiation at low-dose rate (values taken from ICRP Publication 103 (34) ).

Tissue w T Σ w T
Bone marrow (red), colon, lung, stomach, breast, remainder tissues a 0.12 0.72 Gonads 0.08 0.08 Bladder, esophagus, liver, thyroid 0.04 0.16 Bone surface, brain, salivary glands, skin 0.01 0.04 a Remainder tissues: adrenals, extrathoracic region, gallbladder, heart, kidneys, lymphatic nodes, muscle, oral mucosa, pancreas, prostate, small intestine, spleen, thymus, and uterus/cervix.   3. Skin dose measured with XRQA2 radiochromic film at several positions: films S1 and S2 were placed in each side of coronal plane to determine auxiliary dose, films S3 and S4 were attached at both sides of the manubrium of the sternum, films S5 and S6 were placed on the submammary region, films S7 and S8 were placed on the surface of both upper outer quadrant breasts, S9 and S10 were placed on the nipples, S11 and S12 were attached on the back at T7 vertebrae level, and film S13 was positioned in the intermammary cleft. The imaging protocols used were: (a) CBCT low-dose thorax, (b) CBCT "Low dose thorax 10ms", (c) CBCT "Under breast", and (d) orthogonal kV-kV.

B. Monte Carlo simulations
The results of the absorbed dose to the left lung, right lung, heart, left breast, and right breast for the imaging protocols obtained using the PCXMC Monte Carlo program are shown in Table 3. Using the standard low-dose thorax protocol, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.65 ± 0.06 cGy, 0.65 ± 0.06 cGy, 1.08 ± 0.07 cGy, 0.78 ± 0.07 cGy, and 0.78 ± 0.1 cGy, respectively. For the standard low-dose thorax protocol, the deviation between measured and simulated doses was between 3% and 15% (Table 3). Using the "Low dose thorax 10ms" protocol, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.32 ± 0.07 cGy, 0.32 ± 0.07 cGy, 0.54 ± 0.09 cGy, 0.39 ± 0.07 cGy, and 0.39 ± 0.07 cGy, respectively. For the "Low dose thorax 10ms" protocol, the deviation between measured and simulated doses was between 4% and 12% (Table 3). Using the "Under breast" protocol with the isocenter in the center of the phantom at nipple height, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.26 ± 0.08 cGy, 0.26 ± 0.09 cGy, 0.30 ± 0.07 cGy, 0.16 ± 0.08 cGy, and 0.20 ± 0.1 cGy, respectively. For the "Under breast" protocol, the deviation between measured and simulated doses was between 9% and 25% (Table 3). Using the "Under breast" protocol with the isocenter in the middle of right breast, the absorbed doses per fraction to left lung, right lung, heart, left breast, and right breast were 0.20 ± 0.09 cGy, 0.20 ± 0.09 cGy, 0.27 ± 0.09 cGy, 0.13 ± 0.07 cGy, and 0.16 ± 0.06 cGy, respectively. Using the orthogonal kV-kV image protocol, doses per fraction for left and right lung varied from 0 to 0.19 cGy, and for left and right breast varied from 0.05 to 0.42 cGy. For orthogonal kV-kV, the deviation between measured and simulated doses was between 5% and 24% (Table 3).

C. Effective doses
The effective dose per fraction using the CBCT for standard low-dose thorax protocol was 5.00 ± 0.30 mSv, for the "Low dose thorax 10ms" protocol it was 2.44 ± 0.21 mSv, and for the "Under breast" protocol it was 1.23 ± 0.25 mSv with the isocenter positioned at image center and 1.17 ± 0.30 mSv with the isocenter positioned at middle of right breast. The effective dose per fraction using the orthogonal kV-kV protocol was 1.14 ± 0.16 mSv.

IV. DISCUSSION
We have reported organ doses for four different IGRT imaging techniques that can be used to set up patients undergoing radiotherapy for breast cancer. Comparing our results with the doses reported in previous publications (Table 4), we report lower doses. This is expected considering the changes that have been made between the two versions of OBI software used. In OBI version 1.3, used in the study by Kan et al., (20) the settings were 125 kV, 40 mA, and 10 ms per projection, while in the current study we used OBI version 1.5 with 110 kV, 20 mA, and 20 ms per projection ( Table 1). The lower doses can thus be attributed to the lower energy and lower exposure settings in OBI version 1.5. Moreover, Ding et al. (16) calculated organ doses using Monte Carlo for the CBCT low-dose thorax protocol (OBI version 1.4 which has same X-ray settings as version 1.5). The Monte Carlo simulations resulted in a dose to lung of ~ 0.8 cGy. (16) These values are comparable to the lung dose of 0.77 cGy measured in the current work. Reducing the exposure time of the standard low-dose thorax protocol from 20 ms to 10 ms per projection resulted in the expected 50% reduction in dose to all organs (Fig. 2). When using the "Under breast" protocol, assuming the patient is in supine position on the treatment table, the photon beam only rotates around the posterior side of the patient. This resulted in a considerable dose reduction to the breast and heart, with a dose reduction of more than 20% compared to the doses obtained with the standard low-dose thorax protocol (Fig. 2). The two isocenter positions for the "Under breast" protocol (centered in phantom and centered on right breast) had only a small effect on the effective dose. Isocenter position and FOV will affect effective dose and organ doses, so should be measured or inferred for each situation. The "Under breast" protocol, as presented, resulted in a lung dose reduction similar to the "Low dose thorax 10ms" protocol (Fig. 2).
For the orthogonal kV-kV image protocol, the absorbed dose in lungs and breasts was not equally distributed to left and right sides (Fig. 2). Remembering that kV-kV images consist of two orthogonal acquisitions where one is in the AP direction and the other in the lateral direction on the side of the treated breast (in this case the left lateral direction), it is as expected that Table 4. Organ doses obtained from the standard CBCT low-dose mode (OBI version 1.3) published by Kan et al. (20) and organ doses measured in this study using the same standard CBCT low-dose thorax mode (OBI version 1.5).

Organ Absorbed Doses for CBCT Low-dose Thorax
(OBI version 1.3, Kan et al. (20) ) the dose to the left lung and left breast is higher. The absorbed dose to the heart is the largest value for all imaging protocols. The organ doses from the "Under breast" protocol are very similar to the doses obtained from the orthogonal kV-kV image protocol (Fig. 2). This implies that if a radiation oncology department changes from IGRT using kV-kV images to using a CBCT protocol as "Under breast", the extra imaging dose may not be viewed as a clinical concern, provided such low-dose protocols are used. However, the doses need to be evaluated and discussed when setting up such new approaches. A significant dose reduction on the patient skin for the new CBCT protocols was accomplished. The highest skin doses, 1.24 cGy and 0.66 cGy, were observed in the intermammary cleft for the standard low-dose thorax protocol and "Low dose thorax 10ms" protocol, respectively (Fig. 3).
The uncertainty in measured absorbed doses for breasts, heart, and lungs ranged between 0.02 and 0.07 cGy, or 6% and 40%. The uncertainty in the measured dose to the skin ranged from 5% to 40%. The larger uncertainties are seen in low doses, particularly those close to the XRQA2 film lower threshold (~ 0.1 cGy). A source of uncertainty was the flatbed scanner nonuniformity. The scanner reflective mode can add 5% uncertainty to the OD values measured. (28,35) The effect is caused by the light scatter and is predominant at the edges of the scanner bed. Film scanning should be performed in the central region of the scanner to minimize uncertainties; however, this could be a very long process when a large number of film pieces are manipulated.
The PCXMC Monte Carlo program overestimated the absorbed dose for the heart and breast and underestimated the absorbed dose for the lungs, compared to the measured values using XRQA2 film. This trend was observed for all imaging protocols. One explanation for this phenomenon is most likely the anatomical differences between the RANDO and the PCXMC mathematical phantom. While RANDO is an anthropomorphic phantom constructed with a natural human skeleton and fitted soft tissue-simulating material, the phantom used in PCXMC is a computational hermaphrodite model. The position of the lungs in the PCXMC phantom is higher than the position of these organs in a standard-size person. Other reasons for observed uncertainties between simulated and measured organ doses could be uncertainty in the definition of the incident air kerma in terms of the tube current-time product (mAs) of the beam delivered, and no distinction between left and right breast and left and right lung in the PCXMC code. The program can specify the air kerma from the X-ray tube voltage and tube current-time product of the imaging system considered, but uncertainty of the input quantity is about 30%, assuming that X-ray source settings are correctly calibrated. Moreover, the program calculates the mean dose to organs, whereas the point doses measured using film inside the phantom are at specific positions. Considering the uncertainties in the dose simulation process, the simulated values agree well with measured organ doses.
For clinical purposes, there were no obvious losses in image quality or information for the "Low dose thorax 10ms" protocol compared to the standard low-dose thorax protocol. Visual inspection of the images by radiation oncologists, radiation therapists and physicists indicated that the images are good enough for patient setup. It is interesting to note that Sykes et al. (17) showed that accurate patient positioning could be achieved using images obtained with doses ten times lower than the standard practice using Elekta CBCT. For the "Under breast" protocol, with the isocenter placed centrally in the phantom, there was a significant loss of information because the FOV is 25 cm, while the other two CBCT protocols have a FOV of 48 cm. Patient contours cannot be seen if the isocenter is selected in the intermammary cleft of the patient. However, bony landmarks can be seen and if bone structures are used to set up the patient, these images are useful. If the isocenter is selected in the middle of breast, as is usually done for conformal treatments, patient contours can be seen (as illustrated in Fig. 4). The change in simulated organ doses is approximately the same independent of isocenter chosen in "Under breast" protocol (Table 3). Changes in absorbed doses due to reduced or increased beam path length that would be relevant using IMRT or VMAT treatment techniques can be estimated using "Under breast" protocol with isocenter in middle of breast. Moreover, the contrast of the bony anatomy is adequate compared with the other CBCT protocols. Further study would be required to evaluate whether image quality is adequate for monitoring seroma reduction during breast radiotherapy, soft tissue segmentation, detection of masses, and calcifications and localization of surgical clips which are taken as a reference for breast patient setup. (36,37,38) Images acquired for each CBCT acquisition mode are shown in Fig. 4. The isocenter for all protocols is chosen in the center of the thorax (Figs. 4(a)-4(c)) and in the patient breast ( Fig. 4(d)) for the "Underbreast" protocol to encompass soft tissue borders. Currently, for standard breast radiotherapy setup verification, use of tangential portal imaging is the standard of care; however, this method does not provide path length or FSD information. Imaging doses to breasts and other organs at risk are not widely reported in the literature. Use of CBCT setup provides full volumetric information including path length, FSD, bony anatomy, and any internal markers or clips, which can markedly improve beam targeting. Based on the current findings, a CBCT protocol with full rotation around the patient but with reduced current-time product can be used to provide adequate images for breast patient setup, but with significantly lower dose compared to the standard protocol.
The effective dose obtained for the standard low-dose thorax protocol of 5.00 ± 0.30 mSv shows 95% agreement with the effective dose reported by Kan et al. (20) for the CBCT chest scan using low-dose mode of 5.23 ± 0.12 mSv. Comparing the effective doses acquired in this investigation, it can be seen that "Low dose thorax 10ms" protocol presents a reduction of 51% in effective dose compared to the standard low-dose thorax protocol. The "Under breast" protocol goes further, with a reduction of 75% in effective dose compared to the standard low-dose thorax protocol. As expected, the effective dose from the "Under breast" protocol is very similar to the one from the orthogonal kV-kV protocol. To derive the effective doses in this study, we considered the weighting factors defined by the ICRP. In its publication, the factors are only given as sex averaged numbers (i.e., there are no specific female weight factors). Therefore, the reported effective doses are not intended for use in risk estimation, but rather for comparing different imaging protocols and equipment, as well as to give an idea of the patients' exposure due to imaging performed during breast radiotherapy.

V. CONCLUSIONS
Absorbed dose to breasts, heart, lungs, and skin from kV CBCT imaging during external-beam radiotherapy for breast cancer was quantified using radiochromic film dosimetry in an anthropomorphic phantom. The absorbed doses to breasts, heart, and lung were also calculated with PCXMC Monte Carlo simulation software. Good agreement between the measured and simulated organ doses was obtained. The effective dose to breast, heart, lung, and skin was calculated and two simple dose reduction methods demonstrated. The newly proposed CBCT protocols may be considered as a first step in the process to define an accurate imaging protocol with reduced patient dose and, thus, support the use of IGRT for breast cancer patients.