Simulation study on the conversion between CT and CBCT dose quantities via the effective dose

In recent years, cone-beam computed tomography (CBCT) has been used in many imaging tasks traditionally performed by computed tomography (CT). This has created challenges for dosimetry, as the dose quantities in CBCT and CT, i.e. the dose-area product (DAP) and dose-length product (DLP), are not mutually convertible. Convertibility would be desirable to compare doses in similar clinical studies performed using CT or CBCT and ultimately for the application of diagnostic reference levels (DRLs). In this work, the conversion of the DAP into the DLP and vice versa via the effective dose E is investigated with the aim of finding common diagnostic reference levels. The dose calculation was performed using Monte Carlo simulations for scan regions with imaging tasks, which can be carried out either with CT or CBCT scanners. Four regions in the head and four in the trunk were chosen. The calculations resulted in conversion coefficients k=DAPDLP of 30(4) cm for the cranium, 22(4) cm for the facial bones, 24(2) cm for the paranasal sinuses, 18(2) cm for the cervical spine, 78(12) cm for the thorax, 85(13) cm for the upper abdomen, 57(6) cm for the lumbar spine and 70(12) cm for the pelvis.


Introduction
Historically, cone-beam computed tomography (CBCT) was developed from planar imaging systems in intraoperative applications such as angiography, traumatology and orthopaedics.Nowadays, it is an established 3D x-ray imaging technology alongside computed tomography (CT).The applications do not generally overlap, as each system has its advantages for specific imaging tasks.The advantages of CBCT units are their mobility, the relatively small size of the associated equipment, which is important for intraoperative applications, and their cost efficiency compared to that of CT.This is especially relevant when they are used, for example, in dedicated head scanners (Strahlenschutzkommission 2016).CT, on the other hand, has a superior low-contrast resolution, and is less prone to motion artefacts.There are, however, applications where both modalities can be employed and where diagnostic reference levels (DRLs) for CT studies have been published for optimization purposes by the German Federal Office for Radiation Protection (Bundesamt für Strahlenschutz (BfS) 2023).Some relevant clinical tasks are detailed below: (i) neurosurgical imaging of the spine to verify successful treatment.This can be performed either directly in the surgical room using CBCT or afterwards using CT.
(ii) maxillo-facial imaging in traumatology using either CBCT or CT.While CBCT is more easily available in clinics or operating rooms, CT has the advantage that it can be delegated to radiologists.
(iii) orthopaedics to verify screw positioning either in operating rooms (CBCT) or afterwards (CT).
(iv) dental imaging, either by dentists in their own practices (CBCT) or delegated to radiologists (CT).
Other applications include liver and chest imaging during treatment and imaging the pelvis in urology.These are less frequently done and they are typically performed within dedicated studies.
Until recently, CBCT imaging did not have its own separate DRLs in Germany (Bundesamt für Strahlenschutz (BfS) 2016, 2018).In the most recent publication of DRLs by the German Federal Office for Radiation Protection (BfS) dose-area product (DAP) values have been added for CBCT imaging in dental applications and the imaging of sinuses (Bundesamt für Strahlenschutz (BfS), 2023).However, the deviating metric precludes the easy comparison of doses for similar CT applications, such as dental CT or CT of the sinuses.
Common dose metrics in CT imaging are the computed-tomography dose index (CTDI) and the doselength product (DLP) (Shope et al 1981, IAEA 2007).These cannot easily be applied to CBCT because of CBCT's wide cone angle, its standing table and possible asymmetric collimations.Due to many CBCT devices having been developed from fluoroscopy units, the DAP has become a common dose metric and has been proposed for dose reporting and quality assurance by the German Commission on Radiological Protection (SSK) and the European Federation of Organisations for Medical Physics (EFOMP) (Strahlenschutzkommission 2016, de las Heras Gala et al 2017).
There are several approaches to expanding the existing formalism with the aim of unifying CT and CBCT dosimetry, including phantom-based dose metrics and dose metrics measured free-in-air.The American Association of Physicists in Medicine (AAPM) report 111 presents an in-phantom method based on work by Dixon, in which the measurement of the equilibrium dose D eq is proposed (Dixon 2003, Dixon et al 2010).Although this solves the problem of the standing table, asymmetric rotations remain an issue.Huda, on the other hand, proposes a measurement free-in-air using the kerma-area product in CT (Huda 2008).A refinement of this idea is the PAKT concept introduced by de las Heras Gala et al, who propose a unifying dosimetry concept using the incident air kerma and the irradiated area in the CT or CBCT scan (de las Heras Gala et al 2018).For dental applications, the SEDENTEXCT collaboration developed a dedicated phantom and a formalism based on measurements using this phantom (Pauwels et al 2012, Araki et al 2013).
Despite these efforts, there is no unifying formalism which is applied to all CT and CBCT modalities.The goal of this work is to establish the direct conversion of the dose quantities in CBCT and CT for similar imaging tasks with the focus on clinical indications for which both modalities can be applied.We used the effective dose as the linking quantity and thus the stochastic health risk.Therefore, in the case of CT, we consider the DLP instead of the CTDI, as it scales with the effective dose.Our approach is based on Monte Carlo simulations of anthropomorphic phantoms to calculate the DAP and the DLP, respectively, together with the effective dose.This is undertaken using realistic scan protocols from clinical partners as input, focusing on imaging of the head and trunk regions.

Approach for the calculation of the conversion coefficients
The DAP and the DLP have to be related to a common quantity to enable a comparison.This is challenging due to the different dose concepts.The DAP is an incident dose quantity, while the DLP is measured in a phantom.One approach to link these two conceptually different quantities is to use the effective dose from both imaging studies, as it is related to the stochastic health risk.
The DAP is defined as the product of the mean air kerma K a and the area A of the field at the same distance from the source.The DLP is calculated from the volume computed-tomography dose index (CTDI vol , in the following abbreviated as CTDI) and the scan length L of the CT scan: In this work, we used CTDI 100 and chose a pitch factor of 1 for simplicity, but without limiting generality.The conversion coefficients were calculated from the DLP and the DAP of a specific scan normalized to the corresponding effective dose and denoted as DLP E and DAP E in the following: Conversion coefficients from the DAP and the DLP to E, = k E CBCT DAP and = k E CT DLP , respectively, can be defined as the reciprocal of DAP E and DLP E .This leads to the following definition for the conversion coefficient, calculated for each scan region separately: For each scan region, the conversion coefficients were calculated as the mean of the conversion coefficients for the male and female anthropomorphic phantom.

Investigated body regions
Eight body regions were chosen for investigation: four in the head and four in the trunk.These regions were chosen because there are clinical tasks where the corresponding body parts might be investigated The scan parameters are listed in tables 4 and 5.For the CT scans, a pitch factor of one was chosen for simplicity, and half a rotation was added to each side of the scan region to emulate overranging.For the CBCT scans, only partial axial scans were performed with fixed collimation.The high voltage and rotation angle reflect those used in typical scan protocols for the scanners and regions.
2.4.Monte Carlo simulations 2.4.1.General Simulations were performed using ImpactMC version 1.6.1.0.ImpactMC is a software tool for fast Monte  Carlo calculations undertaken directly on DICOM images for real-time dose assessment after computedtomography scans.This software offers the easy implementation of CT and CBCT source data and DICOM images.More information on the simulation in this work is given in table 2.
Based on the simulations, the DLP was calculated for CT scans and the effective dose E for both CT and CBCT scans.The DLP was derived from simulations of a CTDI phantom using the exposure parameters of the corresponding CT scan.The scan length was obtained from typical scan protocols in order to calculate the DLP.Dose modulation was not considered.The DAP for CBCT scans does not have to be simulated.It comprises the air kerma free-in-air at the isocentre and the corresponding area of the incident beam.The air kerma is an input parameter for the simulation of the effective dose, which is used as a normalization factor for the dose calculation and can thus be freely chosen.

Calculation of the CTDI
Simulations of the CTDI for CT and CBCT scans were performed on data sets of a CTDI head and body phantom that have been acquired using the GE Optima and a table as specified above.The scan parameters were taken from measurements and specifications as given in tables 4 and 5.

Calculation of the effective dose
Simulations of patient scans were performed on female and male anthropomorphic phantoms as specified in report 110 of the International Commission on Radiological Protection (ICRP) (ICRP 2009).The phantoms were segmented into 140 organs and tissues.For the red bone marrow, which is not separate tissue in the ICRP phantom, the averaged dose in the spongiosa of the skull, the ribs, the sternum and the large tubular bones in the arms and legs was used (Zankl et al 2021).The contribution of the bone surface was not considered in our calculations, as it was not segmented in the phantom.Moreover, since the total bone surface contributes to the effective dose with a weighting factor of 0.01 only (see table 3), this was not expected to have any significant influence on the results.
For the simulations, the arms of the phantom were removed from the DICOM image, as patients in real clinical imaging are positioned with the arms outside the scan area.A table made of a generic carbon fibre was added, using parameters as described above.
The effective dose is defined as the weighted sum of equivalent organ doses H T for specific tissues T in the body.The weighting factors w T reflect the radiation sensitivity of each tissue.Table 3 shows the relevant tissues and their corresponding weighting factors used to calculate the effective dose.
The effective dose was calculated from the simulated dose distribution in the following steps.A mask for each organ was generated from the anthropomorphic phantom.This mask was multiplied by the dose distribution.The dose values from all organ voxels were added and a mean organ dose was calculated by dividing the sum by the number of voxels N: Finally, all organ doses were added after being multiplied by the respective weighting factor as specified in equation (5).The calculation routine was written in python3.

Determination of source parameters
The source parameters used for the simulation of the CT and CBCT scans are summarized in tables 4 and 5.
They have been retrieved in the following way: • Spectra: The x-ray spectra of the CT scanner were established by measurements using a procedure published by (Rosendahl et al 2019).The x-ray spectra of the CBCT scans were calculated using Spekpy-generated tungsten spectra with the nominal Al and Cu filtration as the starting point (Bujila et al 2020, Poludniowski et al 2021).The filtration has been modified to match attenuation measurements.For details see (Borowski et al 2022).
• Collimation: The true collimation in the patienttable (or z) direction and an aluminium-equivalent bow-tie filter have been determined for each CT scanner and high-voltage setting according to a procedure published by (Rosendahl et al 2019).The beam collimation perpendicular (z-axis) and parallel (x-axis) to the patient table length axis for the CBCT scanners have been measured as described in (Borowski et al 2022).
• Source-isocentre distance (SID): The SID was taken from the specifications of the scanners both for CT and CBCT.
• Angular coverage: The angular coverage (start angle and rotation) in CBCT scans was taken from information displayed by the scanners.The scan lengths have been calculated as described in section 2.2.

Generic carbon-fibre table
The attenuation from the For the CBCT scanners, we compared measured and calculated transmission factors.The measured transmission factors were determined by dose measurements with and without the table using an RC0.6 Farmer chamber (RadCal) positioned at the isocentre of the scanner.The transmission factors T were calculated as a fraction of the kerma at the isocentre with table K a,att and without table K a : Table 5. CBCT source and scan parameters as applied in the simulation and taken from (Borowski et al 2022).The uncertainties estimated for the generating voltage HV and the source-isocentre distance (SID) are 1keV and 1 mm in all cases.t Al and t Al denote the experimentally determined aluminum-and copper-equivalent filtration thicknesses, respectively.The angles α start and α rot denote the start and rotation angles of the x-ray source, respectively.The x-ray tube is below the table for the reference at 0°, and the rotation direction is counterclockwise when looking from the feet to the head.

Scanner
The kerma was calculated from a bin-wise sum product of the energies E in steps of 1 keV.The normalized fluence f(E) was extracted from the x-ray spectra, and the mass-energy absorption coefficients  ( ) , also taken from the XCOM database (Berger et al 2010).The density in the attenuation term was varied until the calculated transmission factor matched the measured one.
For the CT scanners, measurements were performed using a pencil-type chamber positioned either in the upper or lower peripheral drilling of a CTDI body phantom.Such measurements were undertaken for one rotation at a generating voltage of 120 kV.The phantom was positioned at the isocentre of the scanner.The impact of the table attenuation was calculated from the ratio of two measurements: the DLP at the top and bottom position of the CTDI phantom.The setup was replicated for simulation studies using ImpactMC and the density was varied until the simulation results matched the measured results.

Table density
For the CBCT scanner, the results of the transmission measurements are listed in table 6.Two scanner types (the Planmeca Viso G7 and Morita Veraview Epocs) did not have a table, as scanning is performed with the patient positioned upright.Studies on the Ziehm Vision RFD 3D were performed outside the clinical environment in an exhibition room, thus explaining the unusually low transmission factor.The transmission factors for the other scanners are similar, ranging from 76% to 83%.The resulting densities range from 1.00 g cm −3 to 1.50 g cm −3 .
For the CT scanner, the results for the GE Optima CT 660 are displayed in figure 2. The simulated dose ratios are well represented by a straight line, and the table densities can be derived from the intersection between the measured dose ratio and the linear regression of the simulated dose ratio.The calculated densities are 2.78 g cm −3 for the GE Optima CT 660 and 3.38 g cm −3 for the Canon Aquilion ONE.

Scanner-specific and scan-region-specific conversion coefficients
The scanner-specific results for each scan region are shown in figure 3.Each point corresponds to a conversion between one specific CBCT scanner and one specific CT scanner, each with scan protocols appropriate to the scan region and settings as explained in section 2. Subgroups of CT and CBCT scanner combinations were evaluated for each scan region as not all CBCT scanners are intended for use on all body parts.The uncertainty bars were calculated for limits of generating voltage 15 kV lower and higher than the nominal chosen voltage of the CBCT scanner.These correspond to an estimated range of generating voltages covered by the automatic exposure control for patients of different size.With the Siemens Artis Zee, two different protocols were applied using either 73 kV or 109 kV generating voltages for visualizing the cranium.With the Philips Azurion, two different rotation settings were available for the head and body protocols.These were studied in order to obtain information on their influence on the resulting doses.
Two adjustments were made to obtain more realistic and comparable conversion coefficients.The width of the collimation in the x-direction (perpendicular to the patient axis) was reduced to 16 cm in the calculation of the DAP in the head region.Additionally, for the cranium, the area used to calculate the DAP was cut off at the top of the head.In this way, only the radiation field affecting the anthropomorphic phantom resulting in a contribution to the effective dose was taken into account.
Scan-region-specific conversion coefficients can be established by averaging the conversion coefficients achieved for all CBCT-CT combinations.The results are shown in figure 4 for the eight scan regions studied in this work.The values are listed in table 7. The coefficients can be grouped into two parts: one for the four head regions with values ranging from 18 cm to 30 cm and one for the four trunk regions with values ranging from 57 cm to 85 cm.

Uncertainty
The uncertainty of the conversion coefficients is comprised of two parts: the uncertainty of k CBCT = E/DAP and k CT = E/DLP.The uncertainty for the repetition of the simulation is negligible, as each simulation was performed with an uncertainty of < 0.1%.This is very small compared to the variation in results for the different scanners and body regions (see table 2).The conversion coefficient k CBCT = E/DAP depends mainly on the following parameters: the collimations d x and d z , the spectrum via the filtration and generating voltage, the table density and the sourceisocentre distance (SID).The generating voltage was taken from a typical scan protocol while the SID was taken from the manufacturers' specifications.The other parameters were measured.For the evaluation of the uncertainty, simulations were performed on the Siemens Artis Zee scanner, which has typical scan parameter values.Simulations were performed at 73 kV for the cranium, representing the head region, and at 125 kV for the upper abdomen, representing the body region.For each scan parameter and its maximum and minimum value, simulations were performed and conversion coefficients were calculated.The parameters, their variation and the calculated uncertainties of k CBCT are given in table 8.As a result, the uncertainty for the cranium is taken as a conservative estimate, u CBCT = u cranium = 7.64% (k = 1).
The uncertainty for the conversion coefficient k CT = E/DLP was taken from Rosendahl et al (2019), who investigated the same CT scanners (GE Optima CT 660 and Canon Aquilion ONE).They derived an uncertainty of 3.1% for axial scans on a CTDI phantom and 8.0% for helical scans on an anthropomorphic phantom.In total, an uncertainty of u CT = 8.6% is estimated (k = 1).
The combined uncertainty of u k = 11.5% results from u CBCT and u CT (k = 1).

Dependence of conversion coefficients on scan settings
Distinctive ranges of conversion coefficients for the head and the trunk are apparent in table 7.However, there is a large variation between the conversion coefficients for the scan regions (see figure 4) and those for single combinations of the CT and the CBCT (see figure 3).The variation is caused by different scanner settings and by the interplay of rotations, collimations and x-ray spectra, these in turn causing changes in the organ doses.This is especially true of the conversion of the DAP in E, as can be seen in figure 3.
As an example, effective and organ doses to the male reference phantom are shown in table 9 for two thorax scans using the Philips Azurion with different angular coverage, but otherwise identical settings.Overall, an increase of 37% was recorded in the effective dose for angular scan range (b) compared to angular scan range (a).The lung doses almost correlated to the effective doses with an increase of 21%.The largest increase in organ doses was registered for the thyroid, breast and oesophagus by 139%, 85% and 37%, respectively.These organs are positioned at the front of the body.While scan (a) moves symmetrically below the table, scan (b) is more asymmetrical and thus, the front of the body receives a larger amount of the radiation.
As a consequence, a conversion coefficient based on the effective dose can only be established with great individual variability, and without the possibility of reducing this variability by parametrization in terms of rotation, energy and collimation.

Comparison of conversion coefficients to literature values
To the knowledge of the authors, no published conversion coefficients exist for the direct conversion of the DLP into the DAP.However, k CBCT and k CT have been investigated in several works.
The comparison of our calculated k CBCT values to the published ones is challenging due to the large amount of scan regions and settings, as pointed out by Al-Okshi et al (2015).Mah et al reported on a literature review of converting the DAP into effective doses for dental applications (Mah et al 2021).They found values in a broad range from 0.035μSv • mGy −1 • cm −2 to 0.31 μSv • mGy −1 • cm −2 .Our values, while not calculated specifically for dental applications, fit into this range.For example, the scan of the facial bones gives a value of 0.181 μSv • mGy −1 • cm −2 .Kim et al investigated the conversion coefficient for an Alphard VEGA (Asahi Roentgen Ind. Co., Kyoto, Japan) scanner (Kim et al 2014), a dental CBCT unit, and they found values between 0.038μSv • mGy −1 • cm −2 and 0.146 μSv • mGy −1 • cm −2 .These are somewhat smaller than our value for the scan of the facial bones.However, for the Ziehm RFD 3D, where a generating voltage similar to that from (Kim et al 2014) was applied, our calculations gave a result of 0.110 μSv • mGy −1 • cm −2 , which is well within the range of the calculations conducted by Kim et al.  was not present in the simulation of the effective dose, but was present in the measurement of the CTDI.Because different scanners and phantoms were used, the scanner parameters and scan lengths also differ.

In table 10, values for
Previously-published conversion coefficients k CT for the head and the pelvis show substantial offsets to the coefficients presented here.This may be attributed to an offset in the scan regions.k CT for the head and neck differ significantly despite the proximity of the regions.The previously published results give, on average, 0.0021 mSv • mGy −1 • cm −1 and 0.0054 mSv • mGy −1 • cm −1 for the head and the neck, respectively.This correlates with the sensitive organs; according to ICRP report 103, the brain has a weighting factor of 0.01, while that of the oesophagus and thyroid is 0.04, thus 4 times higher.Therefore, if the scan region of this work is extended further into the neck area, a large deviation in k CT is to be expected.Additionally, differences in the red bone marrow distribution between the published results and this work  can cause large differences due to the large weighting factor of 0.12.For the pelvis, we might consider a lower part of the phantom to be scanned as compared to the published results.Therefore, the radiation might not reach the radio-sensitive inner organs, causing a large difference.The difference in the k CT values of Panzer et al (2000) and those of Deak et al suggests large variations in this region.
In conclusion, our simulations of k CBCT and k CT agree reasonably well with results published elsewhere.Differences from existing publications can be attributed to the different simulation setups and scan regions.

Application of the conversion coefficients and comparison to realistic cases
The BfS has recently published updated diagnostic reference levels (Bundesamt für Strahlenschutz (BfS), 2023).Based on these, the corresponding values for CBCT have been calculated for the investigated scan regions using the conversion coefficients from this work.This has been done to be able to make a comparison to literature values (see table 11).
These values are difficult to compare to typically published DAP values, as assessed in sections 4.1 and 4.2, because values for CBCT differ strongly depending on the scan settings.An example of the dependence on the collimation is given by Shin et al (Shin et al 2014) in a study on dental CBCT.They found DAP values for scans in the facial and dental region for an adult patient ranging between 79.9 mGy cm 2 and 3349 mGy cm 2 for fields of view between 51 mm × 51 mm and 200 mm × 200 mm.These values were obtained for scanners of the Alphard 3030 type (Asahi Roentgen Ind., Co. Ltd, Kyoto, Japan) and the Rayscan Symphony type (RAY Co., Ltd, Hwaseong, Republic of Korea).When only considering facial and head scans, values of 3349 mGy cm 2 and 1109 mGy cm 2 result for the Alphard 3030 and Rayscan Symphony scanners, respectively.These are smaller than the DAP value for facial bones computed using the conversion factors.Therefore, in this case, the CBCT application would be favoured in comparison to CT.The update of the DRLs in Germany also includes a DRL for CBCT for sinusitis at 1500 mGy cm 2 .This value is only 9% lower than the value from table 11; they agree very well.The differences are expected because of the large variety of scan settings.
For a comprehensive comparison of CBCT and CT imaging, the image quality should also be taken into account.There is a lack of information on the connection between dose reduction and image quality in CBCT, as for example assessed by Al-Okshi et al for the imaging of the facial bones (Al-Okshi et al 2015).However, a discussion of this correlation goes beyond the scope of this work.

Conclusion
In this study, we have calculated conversion coefficients from the dose-length product in CT to the dosearea product in CBCT with the aim of comparing doses to the patient for similar scans performed with both modalities.In addition, we have investigated the possibility of including CBCT devices in DRLs for CT examinations.The conversion was realized using the effective dose for an anthropomorphic phantom as the linking quantity.The effective dose was calculated from simulation studies, mimicking clinical exam situations.The variation in the result is large, which is also seen in the published literature values, and can be attributed to the intricate interplay between the scan parameters, scan regions and spectra.Published DRLs for CT have been converted to the DAP for CBCT for the eight scan regions and compared to the published literature.In comparison to a newly published DRL on sinusitis for CBCT, the calculated value agrees well with a difference of 9%.A comparison to published values for facial bones shows that the calculated DAP lies significantly higher, which in this case favours CBCT examinations.However, for a comprehensive comparison to the published literature, the large variation of results and the evaluation of the image quality has to be taken into account.

Figure 1 .
Figure 1.Cross section of the dose maps for irradiation of the male reference phantom with the Aquilion ONE for the scan regions considered in this study.The scan regions are from left to right: cranium, facial bones, paranasal sinuses, cervical spine, thorax, upper abdomen, lumbar spine, and pelvis.

Figure 2 .
Figure2.Determination of the density of the generic carbon-fibre table for the GE Optima CT 660.The measured dose ratio in the upper and lower hole of the CTDI phantom for one rotation of the tube is shown as a red line.The corresponding simulation points for different table densities are shown as black dots and a fit of the simulation data as a blue line.The crossing of the red and blue lines is considered as the density of the CT table.
k CT published by the European Commission, Deak et al and Shrimpton et al are shown (Panzer et al 2000, Deak et al 2010, Shrimpton et al 2016).Our results agree reasonably well for the neck, chest and abdomen.The differences can possibly be attributed to the different simulation setups.While we performed simulations on the reference female and male phantoms from ICRP report 110 in this work, the simulations by Deak et al were performed using mathematical phantoms based on the ORNL-phantom series (Cristy 1980).Shrimpton et al also calculated the mean value of k CT for the female and male phantom in ICRP report 110.The phantoms of Deak et al and Shrimpton et al include arms, which were removed in our work.In the work of Shrimpton et al and Deak et al, the table

Figure 4 .
Figure 4. Conversion coefficients for each scan region averaged over the CBCT-CT scanner combinations investigated.The error bars correspond to the standard deviation of the distribution of scanner conversion coefficients.

Table 1 .
Body regions and scan ranges considered in this work.

Table 2 .
Parameters for the Monte Carlo simulations following the recommendations of AAPM task group 268 (Sechopoulos et al 2018).
a VRT-Variance reduction technique.

Table 4 .
(Rosendahl et al 2019)s applied in the simulation.The parameters are taken from specifications (source-isocentredistance (SID), anode angle) or from selected CT parameters (high voltage, nominal collimation d z,nom , bow-tie filter), or they are measured using the methods described in(Rosendahl et al 2019)(measured collimation d z,meas , aluminium-equivalent filtration t Al−equ ).
table is non-negligible.Therefore, a table was included in all simulations by adding the corresponding pixel values to the DICOM files of the CTDI and anthropomorphic phantoms.Voxel data were taken from a scan of the GE Optima CT 660 table at PTB and assigned to a generic material, identical in all simulations.Commonly, x-ray tables are made of carbon-fibre-reinforced polymers.In this study, a polyethylene polymer with a carbon-fibre volume fraction of 60% (a typical ratio of carbon-fibre composites (Grund et al 2019)) was chosen.This closely resembles a typical table material and can easily be converted into elemental fractions of carbon and hydrogen.The table material and geometry are thus identical for all simulations, but the table density was adjusted for each scanner to reflect the scannerdependent table attenuation.

Table 6 .
Measured table transmission and calculated density for each CBCT scanner.

Table 7 .
Conversion coefficients DAP/DLP averaged over the conversion coefficients for all CBCT-CT scanner combinations and conversion coefficients k CT and k CBCT averaged over the respective scanner types.The uncertainty is calculated as the standard deviation of the simulation results of the scanner types.

Table 8 .
Uncertainty estimation for the conversion coefficient k CBCT (coverage factor of k = 1).The uncertainty is estimated representatively for the Siemens Artis Zee for two scan regions: the cranium and upper abdomen.The variation of each parameter is estimated conservatively and the corresponding type B uncertainty of k CBCT is calculated by Monte Carlo simulations.Note that the statistical type A uncertainty of the Monte Carlo simulation is small (< 0.1%) and can be neglected.

Table 9 .
Organ doses for the simulation of the thorax scan of the male phantom using the Philips Azurion scanner at 120 kV with an angular scan range of (a)(α start = − 106.9°, α rot = 210.0°)and (b)(α start = 41.9°,α rot = − 210.0°).The doses are given as absolute values and relative to the dose to the lung.

Table 10 .
Published conversion coefficients from dose-length product to effective dose k CT .The values are given in (mSv • mGy −1 • cm −1 ).