Investigation of the radiation dose from cone‐beam CT for image‐guided radiotherapy: A comparison of methodologies

Abstract Four methodologies were evaluated for quantifying kilovoltage cone‐beam computed tomography (CBCT) dose: the Cone‐Beam Dose Index (CBDI), IAEA Report 5 recommended methodology (IAEA), the AAPM Task Group 111 methodology (TG111), and the current dose metric; the Computed Tomography Dose Index (CTDI) on two commercial Varian cone‐beam CT imaging systems; the Clinac iX On‐Board Imager (OBI); and the TrueBeam X‐ray Imaging system (XI). The TG111 methodology measured the highest overall dose (21.199 ± 0.035 mGy OBI and 22.420 ± 0.002 XI for pelvis imaging) due to the full scatter of the TG111 phantom and was within 5% of CTDI measurements taken using a full scatter TG111 phantom and 30‐cm film strips. CBDI measured the second highest overall dose, within 10% of the TG111, with IAEA measuring the third highest dose. For head CBCT protocols, CBDI measured the highest dose, followed by IAEA. The CTDI method measured lowest across all scan modes highlighting its limitations for CBCT dosimetry. The XI imaging system delivered lower doses for head and thorax scan modes and similar doses to the OBI system for pelvis scan modes due to additional beam hardening filtration in the XI system. The TG111 method measured the highest dose in the center of a CBCT scan during image guidance procedures; however, CBDI provided a good approximation to TG111 with existing CTDI equipment and may be more applicable clinically.

Image guidance for patient positioning was originally performed using the megavoltage (MV) treatment beam and an electronic portal imaging device (EPID) or film placed behind the patient. However, at these MV energies, the inherent Compton scatter results in poor soft-tissue contrast, limiting reference points within the body to higher Z tissue such as bone, or internal fiducial markers. To resolve this, many linear accelerators (LINACs) now have built-in kilovoltage (kV) imaging systems that can produce images with improved softtissue contrast to correct for internal organ motion and patient setup errors. Examples include the On-Board Imager (OBI) and the True-Beam X-ray imaging system (XI) of Varian Medical Systems (Palo Alto, CA, USA). These devices consist of a kV X-ray source and an amorphous silicon detector mounted to the LINAC gantry on extendable robotic arms orthogonal to the treatment beam. These devices can acquire 3D cone-beam CT (CBCT) images of the patient in a single rotation of the gantry allowing registration with the radiotherapy planning CT to check for positional errors and make corrections as necessary with a high degree of accuracy. 1,2 Currently, imaging dose is often omitted from treatment plans since, being typically less than 1 Gy for an entire treatment schedule or 1-10 cGy for a single scan, it is two orders of magnitude smaller than the therapeutic doses. 1,[3][4][5][6][7][8][9] However, during an imaging procedure, large portions of the body are irradiated, including radiosensitive structures such as lung, breast, thyroid, and reproductive organs. Bone structures also receive higher doses than other tissue at kV energies due to increased photoelectric absorption.
Simulated doses in the femoral heads as high as 1.5-2.5 Gy have been reported due to daily pelvis CBCT imaging during a course of radiotherapy. 10 Each clinic has its own protocols for frequency of CBCT imaging depending on tumor site and experience in day-to-day setup variations. While some radiotherapy clinics use daily CBCT imaging, often a typical CBCT schedule might be daily for the first week then once per week for the remainder of the treatment course. During a treatment of 30-40 fractions, the imaging dose has been shown to be significant, with reported effective doses of between 8 mSv and 46 mSv per CBCT scan leading to an increased risk of a patient developing a secondary primary malignancy. 1,11,12 Therefore, a method for quantifying the imaging dose is necessary to evaluate any increased risk to the patient and aid in making informed decisions on the appropriate use of CBCT imaging during the course of treatment.
The traditional methods for quantifying fan-bean CT dose, the Computed Tomography Dose Index (CTDI), underestimate CBCT dose due to an insufficient detector length to capture the full dose profile, and inadequate phantom length to achieve scatter equilibrium in the center of the detector. 13 The underestimation worsens with increased beam width, which can be up to 40 cm for CBCT scans. 14  Group 111 Report. 16 While each protocol attempts to account for the limitations of CTDI in determining CBCT dose, their approach is somewhat different in terms of both equipment and measurement conditions. A comparison of all four methodologies for measuring CBCT dose forms the scope of this paper.

2.A | Phantom design and materials
A standard 32-cm-diameter poly methyl methacrylate (PMMA) CTDI body phantom and 16-cm-diameter head phantom were used for CTDI, CBDI, and IAEA measurements as shown in Fig. 1. Both phantoms have insert spaces for a 100-mm pencil ionization chamber at the center and at the four peripheral locations. The phantom was placed on the couch with the center positioned at the isocenter using the room alignment lasers.
CTDI and CBDI measurements were taken using the UNFORS Xi detector system from RaySafe TM . The system includes a base unit which connects to several detectors including a CT detector and HVL measurement tool for kV energies. The pencil ionization chamber for CT measurement has a sensitive length of 100 mm. The Xi device is a self-contained detector, including the ionization chamber, electronics, and automatic temperature and pressure adjustments. The detector has a dose range of 10 lGy to 9999 Gy with an uncertainty of AE5%. Its energy dependence is <5% with an axial and radial uniformity of AE2% and AE3%, respectively. The chamber used for the TG111 measurements was a 0.6-cc NE 2571 Farmer ionization chamber. The sensitive air volume has a length of 24 mm and radius 3.2 mm. It is enclosed by a graphite thimble of thickness 0.065 g/cm À2 . The chamber operates at a bias voltage of 300 V between the stem and chamber wall. The chamber was calibrated at the Australian Radiation Protection and Nuclear Safety Agency (ARPANSA) and is traceable to the Australian primary standard.

2.B | OBI and XI CBCT imaging systems
Measurement of CBCT dose was performed on both a Varian 21iX on-board imaging (OBI) and Varian TrueBeam X-ray imaging (XI) systems. Each system can acquire 2D kV and 3D CBCT images as the source and detector are rotated around the patient. The clinical CBCT protocol settings used for the measurements in this study are given in Table 1.
The beam width is modulated with independently adjustable Xand Y-lead blade collimators. The field size at isocenter can be varied from 2.0 9 2.0 mm to 50.0 9 50.0 cm on both the XI and OBI systems. The X1 and Y1 collimators have a range of À25.0 to +3.5 cm and the X2, Y2 from À3.5 cm to +25 cm. On the XI system, a titanium beam hardening foil filter further hardens the X-ray spectrum to reduce low-energy photons. The axial plane is further modulated with an aluminum bow tie filter varying in thickness from 2 to 28 mm.

2.C | Determining the CTDI and CBDI
where D Measured represents the measured dose collected in scanning length L c = 100 mm. To obtain the CTDI, the DLI is divided by the superior-inferior (S-I) collimation width (coll): The CBDI is calculated by dividing the DLI by the 100 mm sensitive length of the chamber L c : Weighted CTDI w and CBDI w were then calculated from the CTDI c measured in the center of the CTDI phantom and average of CTDI measurements in the peripheral positions CTDI p : The normalized n CTDI w and n CBDI w values, which represent CTDI w per 100 mAs, were determined from the weighted CTDI w values and corresponding mAs for the given scan: The CBDI methodology proposed by Amer et al. stipulates additional scatter material be placed superior and inferior to the CTDI phantom to achieve scatter equilibrium in the center of the phantom. In this study, no additional scatter material was used for CBDI measurements. It should be noted that this will result in a reduction in measured dose, as reported by Amer et al. 14

2.D | IAEA methodology
The weighted IAEA w dose was determined for the clinical protocols for pelvis, thorax, and head CBCT. As per the IAEA protocol for beam widths greater than 60 mm, a reference CTDI ref is first determined with a S-I collimation of 2 cm. The CTDI ref is an in-phantom CTDI measurement with sufficiently narrow S-I collimation to facilitate the capture of the full dose profile by the 100-mm pencil chamber within the CTDI phantom. The CTDI ref is scaled by the ratio of CTDI, measured in free-air, with S-I collimations of 2 cm and that used in the clinical protocol, to give IAEA w as outlined in eq. (6): The free-air measurements were taken with the pencil chamber suspended away from the couch as shown in Fig. 3. The kV, mAs, and axial collimation settings specified by the clinical protocol were applied for all measurements. The DLI for each chamber position was summed and divided by the S-I collimation to yield the protocol width CTDI in-air :

2.E | AAPM TG111 methodology
The TG111 methodology for calculating CBCT dose is based on measuring dose in a phantom that provides close to full scattering conditions for broad cone-beam imaging systems. 16 As such, the custom-built phantom described in Section A was used to determine a weighted TG111 w dose for the pelvis and thorax clinical CBCT protocols on both the OBI and XI systems. A separate head phantom was not constructed for our study, and hence, CBCT dose measurement using the TG111 approach was limited to the full for beams of a known HVL ( Table 2). The HVL of the OBI and XI kV F I G . 3. Experimental setup for the measurement of CTDI in-air . The 100-mm chamber was stepped in 100-mm increments to achieve the necessary integration length to capture the full dose profile. The chamber was held in place using a retort stand and rod. The chamber was extended from the couch a distance equal to half the total integration length to minimize scatter from the couch as specified by the IAEA. 15 beams had previously been measured with the UNFORS detector.
The accuracy of the UNFORS to measure HVL was verified on an orthovoltage unit for several beam qualities with well-known HVL values. The charge collected (q) in the ionization chamber was corrected for ambient temperature and pressure and converted to dose using eq. (8): The distance between successive measurements was increased for wider beams as scatter equilibrium was approached, and further increases in dose were minimal for the wider collimation widths. The film was calibrated against air KERMA measured with a 0.6cc Farmer chamber with calibration traceable to a primary standard.
To avoid changes in film response due to beam rotation, dose calibration for the film was performed in kV fluoroscopy mode with a stationary X-ray tube. 19,22 The kVp, mAs, and collimator widths were set to be identical to the pelvis CBCT protocol for the calibration.
Furthermore, the bow tie filter was inserted in place to ensure that the calibration was performed in the same beam quality as the CBCT beam.
The film was analyzed in ImageJ (National Institute of Health, Bethesda, MD, USA) and average pixel intensity across a 1 9 2 cm region of interest (ROI) for each piece used to determine net reflectance (net DR) using a method previously described by Tomic et al. 23 The data were imported into MATLAB â (MathWorks Inc., Natick, MA, USA) and a curve fit to the data. The applied fitting function was of the form y ¼ ax bÀx where x and y represent net DR and air KERMA, respectively, with corresponding fitting parameters a and b.
A fitting function of this form has the benefit of being monotonically increasing and returns a zero value for zero dose.
For CTDI film measurements, 30 cm by 3 cm film strips were cut to measure the full length of the beam profile. A custom-built PMMA cylindrical rod designed to fit in the holes of the TG111 phantom was created to house the film strips. The rod was cut into two hemispheres allowing the film to be placed in between the hemispheres before inserting the rod into the phantom. Individual strips were then exposed in each of the five positions within the TG111 phantom on OBI and XI systems. Two CBCT scans were acquired for each strip to deliver a higher dose to the film. The film processing and scanning procedure described earlier for the film calibration were maintained for the film strips. Line profiles were taken across the film strips and converted to air KERMA using the respective calibration curves. The converted air KERMA values were then halved to obtain the dose profile for a single CBCT scan. Due to the high sensitivity of XR-QA2 film, two strips were irradiated and an average profile calculated. The center and peripheral DLIs were determined from the film profiles and divided by S-I collimation to yield the CTDI film which was compared to TG111 w measurements on the OBI and XI systems.

3.A | CTDI AND CBD I MEASUREME NTS
Weighted CTDI w and normalized n CTDI w dose for OBI and XI clinical protocols are shown in Table 3 with weighted CBDI w and normalized n CBDI w dose shown in Table 4. CTDI w values were similar for OBI and XI pelvis protocols with XI measuring 0.24 mGy (3%) higher. The OBI thorax mode measured 0.23 (10%) mGy higher than the XI thorax mode. The varying mAs across the three OBI head modes were reflected in the respective doses, which varied by 10.17 mGy. OBI standard head mode measured 0.59 (36%) mGy higher than the XI head mode. The normalized n CTDI w dose was higher for the OBI system across pelvis, thorax, and head protocols. In particular, the pelvis modes varied by 0.47 (43%) mGy. The difference is attributed to the higher mAs for pelvis scans on the XI system which is offset by its additional beam hardening titanium filter.
The trends described above for CTDI w and n CTDI w dose also follow for CBDI as the CBDI simply upscales CTDI by dividing by chamber length rather than S-I collimation. Due to the upscaling, the CBDI values are more than double CTDI with an increase of 106% for OBI and 114% for XI protocols. These values would be higher still had additional scatter material been used for the CBDI. 14

3.B | IAEA MEASUREMEN TS
The CTDI in-air dose and their ratios for the OBI and XI systems are presented in Table 5 and the weighted IAEA w and normalized n IAEA w doses in Table 6. The OBI thorax mode could not be evaluated for the IAEA method as the UNFORS chamber would not trigger for reference beam width scans due to the low signal.
The average in-air ratio for the OBI system was 16% higher than for the XI system. The difference between OBI and XI systems is due to additional filtration in the XI system which removes lowenergy photons from the spectrum. Hence, the photon fluence in air is higher in the OBI system, and more energy is deposited within the pencil chamber.
The weighted pelvis IAEA w was 3.57 (24%) mGy higher for the OBI system compared with XI. The OBI standard head mode was 0.937 (33%) higher than the XI head mode. The higher OBI values reflect the greater variation in dose measured between the reference beam width and protocol width resulting in a larger in-air ratio for the OBI protocols when compared with the XI ones.
These differences in ratios may partially be attributed to the additional low-energy photon component that is filtered out by the titanium filter on the XI system. Normalized n IAEA w values varied by 1.30 (93%) mGy 100 mAs À1 and 0.04 (35%) mGy 100 mAs À1 across pelvis and head modes, respectively.

3.C | TG111 MEASUREME NTS
The weighted TG111 w and normalized n TG111 w doses for pelvis and thorax modes on the OBI and XI systems are given in Table 7.
The XI pelvis mode measured the highest dose, 1.22 (6%) mGy higher than the OBI pelvis protocol. The higher dose for XI can be attributed to a slightly larger transverse collimation width and a higher mAs. When normalized per 100 mAs, the OBI system delivered an additional 1.00 (47%) mGy compared with the normalized XI pelvis mode.

3.D | COMPAR ISON OF METHOD OLOGIES
Comparisons of TG111, CBDI, IAEA, and CTDI protocols for pelvis and thorax protocols are illustrated in Fig. 4(a) and 4(b). Comparison of CBDI, IAEA, and CTDI for head protocols are shown in Fig 4(c).
The TG111 methodology resulted in the highest recorded dose for the pelvis and thorax CBCT protocols. The CBDI methodology produced the second highest dose followed by the IAEA methodology, while the CTDI method yielded the lowest dose for each protocol. For pelvis modes, the CTDI w measured 56% and 57% lower than TG111 w for OBI and XI, respectively, and similarly 64% and 58% lower for the thorax protocols.
For the head protocols and noting the absence of a TG111 measurement, the CBDI measured the highest dose, followed by IAEA and CTDI for the standard head OBI mode and XI head mode. The CTDI w measured 106% and 114% lower than CBDI w for OBI and XI head protocols, respectively.

3.E | COMPAR ISON OF METHOD OLOGIES FOR INCREASING BEAM WIDTH
Weighted doses from CTDI, CBDI, IAEA, and TG111 protocols for S-I collimation widths ranging from 2 cm to 40 cm are presented in Fig. 5. Film profiles measured in the center and periphery of the TG111 phantom are shown in Fig. 6(a) for the OBI system and Fig. 6(b) for the XI system. The weighted CTDI film values calculated from the film profiles for the OBI and XI systems are given in Table 8   of the data points between the two methodologies is apparent (Fig. 5). Additionally, DLI calculated with a 300-mm ionization chamber used in the Hu study averages dose across the full integration length, while for our approach, the dose integral was acquired from a summation of five dose integrals for each step of the pencil chamber.
Lower doses at the ends of 300-mm sensitive-length chamber would reduce the average dose compared with a stepwise approach for which far ends of the detection length were weighted, in this case by 1 5 of the dose integral as shown in eq. (7).
The CTDI values were significantly lower than the doses measured by the three alternative methodologies across all CBCT clinical protocols.
Compared to TG111 w , dose differences between 55% and 64% were observed for pelvis and thorax protocols across OBI and XI systems . The underestimation is due to an insufficient detector length to capture the full dose profile and a phantom without the required length to achieve full equilibrium scatter. The underestimation worsens with increasing beam width as the divisor in the CTDI calculation increases with minimal increase in the measured dose profile. This supports previous work from Boone who showed CTDI 100 had an efficiency of 63% compared with CTDI with infinite detection and phantom length. 13 Hu and McLean showed a 66% efficiency of a 100 mm integration length compared to a 300-mm integration. 7 Hu also demonstrated a dose difference of up to 36% between a 16-cm phantom and 45-cm phantom.
The slightly higher TG111 w measurements for the pelvis XI system compared to pelvis OBI were due to the higher mAs used for  The doses measured in this study were in agreement with studies carried out using the same methodologies and imaging systems. 5,7,24,25 However, it has to be kept in mind that, depending on imaging system, software version and methodology, doses varied greatly from study to study, and this should be taken into account when interpreting the results. 7,10,16,[25][26][27][28][29] Additionally, it must be stressed that doses presented in this work represent average air KERMA within a PMMA cylindrical phantom and should not be interpreted as patient dose. Any such conversion to patient dose would require information regarding the beam spectrum, organ site and patient size parameters. 30

| CONCLUSION
The methods evaluated in this work estimate the radiation output of two kV CBCT imaging systems as average dose to the center of a PMMA cylindrical phantom; they are therefore used as a tool to compare radiation exposures from different scanners and/or imaging protocols. This study investigated how the dose estimated by the AAPM TG111, the IAEA Report No. 5, and the Cone-Beam Dose Index protocols, which try to account for higher S-I beam widths inherent with CBCT imaging, compares to the current standard for estimation of CT radiation output, the CTDI.
It has been shown that CTDI values systematically measured lower doses when compared to the three alternative methods; in particular, they underestimated doses when wider beam widths were considered.
Amongst the protocols investigated, the TG111 method accounts for the full scatter profile using a longer cylindrical phantom than the other methods; it is therefore reasonable to consider the dose measured using the TG111 protocol as the best estimation of dose in the center of a PMMA phantom from a CBCT acquisition. This was supported by weighted average kV CBCT dose using CTDI film profile measurements. The IAEA methodology agreed with the TG111 estimations in air, but it was not able to account for the full scatter profile when measured in a phantom. In the absence of a custom-made full-length phantom, the CBDI approach gives a comparable indication of CBCT dose to the TG111 methodology using equipment more commonly found in radiotherapy departments. Future work should also involve conversion of TG111 measurements to patient dose, taking into account patient-specific imaging parameters and patient size.
As a secondary result of this study, it has been shown that, for the imaging protocols considered, the XI system consistently delivered lower dose than the OBI system due to its harder energy spectrum, in particular when values were normalized to 100 mAs.

ACKNOWLEDG MENTS
The author wishes to thank Assoc. Professor Martin Carolan and the Illawarra Cancer Care Centre for the use of their equipment and Mr.
Simon Downes and the Nelune Comprehensive Cancer Centre for lending of the CTDI phantom.

CONFLI CT OF INTEREST
The authors declare no conflict of interest.