Estimating fetal dose from tube current‐modulated (TCM) and fixed tube current (FTC) abdominal/pelvis CT examinations

Abstract

Purpose

The purpose of this work was to estimate scanner‐independent CTDI _vol‐to‐fetal‐dose coefficients for tube current‐modulated (TCM) and fixed tube current (FTC) computed tomography (CT) examinations of pregnant patients of various gestational ages undergoing abdominal/pelvic CT examinations.

Methods

For 24 pregnant patients of gestational age from <5 to 36 weeks who underwent clinically indicated CT examinations, voxelized models of maternal and fetal (or embryo) anatomy were created from abdominal/pelvic image data. Absolute fetal dose (D _fetus ) was estimated using Monte Carlo (MC) simulations of helical scans covering the abdomen and pelvis for TCM and FTC scans. Estimated TCM schemes were generated for each patient model using a validated method that accounts for patient attenuation and scanner output limits for one scanner model and were incorporated into MC simulations. FTC scans were also simulated for each patient model with multidetector row CT scanners from four manufacturers. Normalized fetal dose estimates, nD _fetus , was obtained by dividing D _fetus from the MC simulations by CTDI _vol. Patient size was described using water equivalent diameter (D _w ) measured at the three‐dimensional geometric centroid of the fetus. Fetal depth (DE _f) was measured from the anterior skin surface to the anterior part of the fetus. nD _fetus and D _w were correlated using an exponential model to develop equations for fetal dose conversion coefficients for TCM and FTC abdominal/pelvic CT examinations. Additionally, bivariate linear regression was performed to analyze the correlation of nD _fetus with D _w and fetal depth ( DE _f ). For one scanner model, nD _fetus from TCM was compared to FTC and the size‐specific dose estimate (SSDE) conversion coefficients (f‐factors) from American Association of Physicists in Medicine (AAPM) Report 204. nD _fetus from FTC simulations was averaged across all scanners for each patient $(\overline{n{D}_{\mathit{fetus}}})$. $\overline{n{D}_{\mathit{fetus}}}$ was then compared with SSDE f‐factors and correlated with D _w using an exponential model and with D _w and DE _f using a bivariate linear model.

Results

For TCM, the coefficient of determination (R ² ) of nD _fetus and D _w was observed to be 0.73 using an exponential model. Using the bivariate linear model with D _w and DE _f , an R ² of 0.78 was observed. For the TCM technology modeled, TCM yielded nD _fetus values that were on average 6% and 17% higher relative to FTC and SSDE f‐factors, respectively. For FTC, the R ² of $\overline{n{D}_{\mathit{fetus}}}$ with respect to D _w was observed to be 0.64 using an exponential model. Using the bivariate linear model, an R ² of 0.75 was observed for $\overline{n{D}_{\mathit{fetus}}}$ with respect to D _w and DE _f . A mean difference of 0.4% was observed between $\overline{n{D}_{\mathit{fetus}}}$ and SSDE f‐factors.

Conclusion

Good correlations were observed for nD _fetus from TCM and FTC scans using either an exponential model with D _w or a bivariate linear model with both D _w and DE _f . These results indicate that fetal dose from abdomen/pelvis CT examinations of pregnant patients of various gestational ages may be reasonably estimated with models that include (a) scanner‐reported CTDI _vol and (b) D _w as a patient size metric, in addition to (c) DE _f if available. These results also suggest that SSDE f‐factors may provide a reasonable (within ±25%) estimate of nD _fetus for TCM and FTC abdomen/pelvis CT exams.

Short abstract

1. Introduction

Performing computed tomography (CT) exams on pregnant patients is occasionally necessary. In such cases, physicians, medical physicists, and/or radiation safety officers may need to estimate the radiation dose received by the conceptus (fetus or embryo) with a reasonable degree of accuracy due to the risks associated with irradiating developing radiosensitive organs such as red bone marrow.^{1, 2, 3} Initial attempts to estimate the radiation dose a fetus would receive from a CT exam were based either on phantom measurements, Monte Carlo (MC) simulations of geometric phantoms, or a combination of the two.^{4, 5, 6, 7, 8, 9, 10, 11, 12, 13} The approach described by Felmlee et al.⁴, for example, uses anthropomorphic phantom measurements and measured Computed Tomography Dose Index (CTDI) values. Some important limitations of these early efforts relate to their use of simplified, geometric models and assumptions of nonvarying maternal anatomy in a single‐size patient model.

More recent methods for estimating radiation conceptus dose from CT exams of pregnant patients are typically based on MC simulations of pregnant patient anatomy representing a range of gestational ages and patient sizes.^{14, 15, 16, 17, 18, 19, 20} However, many of these studies were limited in that they were either not based on actual pregnant patient anatomy, applied only to a single scanner model, or did not include a range of gestational ages. In addition, most of these methods have thus far been limited to fixed tube current (FTC) CT exams of pregnant patients. For example, the study conducted by Angel et al.¹⁴ examined the effects that maternal and fetal characteristics such as maternal size, gestational age, and fetal presentation have on fetal dose using MC simulations. This work was based on voxelized patient models generated from a set of pregnant patients who underwent clinically indicated abdominal and pelvic CT examinations.¹⁴ While the results of Angel et al. addressed the limitations of earlier studies by providing size‐specific fetal dose estimates based on actual patient anatomy across a range of gestational ages, dose estimates were nevertheless limited to FTC CT exams of pregnant patients for a specific scanner model.

In the current context of CT dosimetry, nearly all CT exams are performed with attenuation‐based tube current modulation (TCM). Additionally, Turner et al. developed scanner‐independent, size‐adjusted estimates of organ dose by normalizing organ doses from MC simulations of voxelized models by CTDI_vol.^{21, 22} The work of Turner et al. was incorporated into size‐specific dose estimate (SSDE) in AAPM Report 204 and was subsequently extended in AAPM Report 220 which described the attenuation‐based size metric water equivalent diameter (D _w ), which is defined as “x‐ray attenuation of a patient in terms of a water cylinder having the same x‐ray absorption.”^{23, 24} The normalization metric of CTDI_vol and the attenuation‐based size metric of D _w are routinely used for generating predictive estimates of normalized organ dose. However, estimates reported by Angel et al. were based on a per 100 mAs normalization and patient size characteristics were described in terms of maternal perimeter.

Gu et al. conducted a study to estimate the effects of TCM on fetal doses for three computational phantoms representing pregnant patients of the gestational ages of 3, 6, and 9 months.^{18, 19} Since no TCM data existed for these computational phantoms, TCM schemes were instead selected from actual pregnant patients of gestational ages of 15, 20, and 31 weeks, and were appropriately applied to the computational phantoms. These selected TCM schemes were longitudinal modulation (z axis) only and did not include angular (x–y axis) modulation. This approach assumed that the TCM scheme for one patient matched the anatomy of another patient. This is crucial since TCM adjusts the tube current with respect to the attenuation characteristics of the patient in question. Additionally, as shown in Angel et al., fetal dose correlates strongly with a measure of maternal patient size but does not correlate with gestational age.¹⁴

Physical phantom studies have also been used to estimate fetal dose from CT with TCM at several gestational stages using anthropomorphic phantoms and MOSFETs.^{13, 20} While these studies do consider TCM, there is an inherent issue with estimating dose using small detectors. Even when considering a FTC study, the dose distribution within a patient is substantially nonuniform.²⁵ When this effect is combined with the nonuniform exposure patterns of TCM during a helical acquisition, the point detector is sampling only a certain location in a complex dose distribution environment. MC simulations can be used to estimate average fetal dose in a way that overcomes the limitations of measuring at single points inside a nonuniform dose distribution.

However, performing MC simulation for every pregnant patient undergoing CT examinations is a time prohibitive option. Therefore, the primary purpose of this investigation was to develop a patient size‐specific method for estimating radiation dose to the embryo or fetus for pregnant patients undergoing abdominal/pelvic CT exams that accounts for the use of either TCM or FTC scanning techniques and applies to patients of different sizes and gestational ages. This investigation was meant to serve as an update to the study conducted by Angel et al. and used the same patient models. However, although the patient models used in this study were the same as those used by the study conducted by Angel et al., this study differs from the previous study in four distinct ways. (a) The study conducted by Angel et al. normalized fetal doses on a per 100 mAs basis, whereas this study used CTDI_vol as normalization metric.¹⁴ The normalization using CTDI_vol employed in this study is consistent with AAPM Report 204.²³ For TCM scans, fetal doses were normalized by CTDI_vol based on the average tube current across the entire scan. For FTC scans, the dose to the fetus was normalized by CTDI_vol for each patient. (b) This study used D _w as metric of patient size from AAPM Report 220²⁴ while the study conducted by Angel et al.¹⁴ used patient perimeter as the metric of patient size. (c) Multiple scanner models were included in the FTC simulations for this study to accommodate for the effects of scanner design. (d) The effects of TCM were included in this study, whereas the study by Angel et al. was limited to FTC. The results of this study were a set of scan technique‐independent, CTDI_vol‐to‐fetal‐dose coefficients for abdominal/pelvic CT, which can be applied to (a) either TCM or FTC scans, (b) patients of different sizes in terms of D _w , and (c) patients at various gestational stages of pregnancy. Additionally, as with Angel et al., this study also investigated using fetal depth (DE _f ) in conjunction D _w for estimating CTDI_vol‐to‐fetal‐dose coefficients.¹⁴ Lastly, the resulting fetal dose coefficients were compared to SSDE conversion coefficients (f‐factors) from AAPM Report 204 to investigate the use of SSDE as an estimate of fetal dose for either TCM or FTC scans. The SSDE f‐factors are the conversion coefficients for adjusting the scanner‐reported CTDI_vol to account for patient size. AAPM Report 204 notes “that the actual dose to any given patient may differ from the value calculated using this report by 10% to 20%”.²⁴ Although not originally intended to be a measure of fetal dose, it is expected that the SSDE f‐factors will become widely available. In addition to estimating fetal dose, this study therefore also sought to compare fetal dose from TCM and FTC to SSDE in order ascertain whether or SSDE could be used as a surrogate for fetal dose.

2. Materials and methods

2.A. Voxelized patient cohort

The patient image data used in this investigation were previously collected by Angel et al. and comprise 24 pregnant patients of gestational ages ranging from <5 to 36 weeks who underwent clinically indicated abdominal/pelvic FTC CT examinations.¹⁴ For each patient, the image data included, at a minimum, patient anatomy from the lower thorax to the pubic symphysis. Given that the patients represent a range of gestational ages, the fetus was not visible in five patients due to the pregnancies being in the early stages. In these cases, either the uterus or the gestational sac was used as a surrogate organ for the fetus. For each patient, the uterus, gestational sac, and fetus were semi‐automatically segmented from the axial images, depending on what was visible in the image data. The three‐dimensional geometric centroid of the fetus was calculated based on the segmented boundaries. For each patient, an estimate of patient size in form D _w was determined from the image data per AAPM Report 220.²⁴ The patient size was determined as the D _w measured at the image containing the three‐dimensional geometric centroid of the fetus.¹⁴ If the fetus was not present, then D _w was estimated at the image containing three‐dimensional geometric centroid of either the uterus or gestational sac.¹⁴ DE _f is defined and was measured as the distance from the anterior skin surface to the most anterior part of the fetus.¹⁴ DE _f was therefore measured at the slice containing the most anterior part of the fetus. Voxels containing the uterus were modeled as soft tissue, and the voxels of the gestational sac were modeled as water.¹⁴ All voxels containing the fetus were modeled as either soft tissue or fetal bone depending on Hounsfield number.¹⁴ The remaining voxels outside of the contoured regions were identified as a specific tissue type (lung, fat, water, muscle, bone, air) using a Hounsfield number lookup table.²⁶ Table 1 contains a list of the patient gestational ages and the organs of interest used in this study. Figure 1 shows axial images of the centroid of the region of interest demonstrating the development of voxelized patient models for three of the pregnant patients.

Table 1.

Gestational age and region of interest used for all subjects

ID	Gestational age (weeks)	Region of interest
1	<5	Uterus
2	5.0	Ges. Sac
3	5.0	Ges. Sac
4	6.6	Ges. Sac
5	7.1	Ges. Sac
6	12.1	Fetus
7	14.3	Fetus
8	14.9	Fetus
9	17.0	Fetus
10	17.1	Fetus
11	18.5	Fetus
12	20.3	Fetus
13	22.0	Fetus
14	23.7	Fetus
15	24.0	Fetus
16	24.4	Fetus
17	25.0	Fetus
18	27.0	Fetus
19	27.4	Fetus
20	27.4	Fetus
21	28.3	Fetus
22	29.4	Fetus
23	35.0	Fetus
24	35.9	Fetus

Figure 1.

Images in the first row, from left to right, represent early‐term, mid‐term, and late‐term pregnant patients. Images in the second row show the uterus (yellow), gestational sac (green), and fetus (red) segmented from the images of these pregnant patients. Adapted with permission from Fig. 3 of Angel et al.¹⁴ [Color figure can be viewed at wileyonlinelibrary.com]

2.B. Monte Carlo simulation tool

A Monte Carlo software package, MCNPX (Monte Carlo N‐Particle eXtended version 2.7.a), was utilized for all simulations.^{27, 28} The source code of MCNPX was modified to model MDCT scanner geometries and spectra.^{29, 30, 31, 32} The code is capable of selecting the appropriate energy spectrum data reflecting a range of scanner models previously generated using the “equivalent source” method by Turner et al.³³ and other user‐specified variables such as scan length and helical pitch. All simulations were conducted in photon transport mode with a 1 keV low‐energy cutoff. Additionally, all simulations were performed with 10⁷ particle histories. All MC simulations were performed with the voxelized patient models at isocenter. In the case of TCM simulations, an additional text file with information on individual table location, tube angle, and tube current value throughout the scan was utilized in the simulation. Validation of this MC simulation package under a variety of conditions including TCM has been previously reported.^{30, 34, 35, 36} This work used the computational and storage services of the Hoffman2 Shared Cluster provided by the UCLA Institute for Digital Research and Education's Research Technology Group.

2.C. Modeling TCM scans

As actual TCM schemes are scanner make and model dependent, TCM modeling was limited to the specific scanner modeled previously by McMillan et al.³⁷ The method was used to estimate the TCM scheme from one manufacturer (CAREDose4D, Siemens Healthineers, Forchheim Germany). Therefore, only TCM scans using CAREDose4D³⁸ from the Definition AS64 were simulated. The methodology is summarized below in Section 2.C.1. Table 2 contains the scanning protocol for TCM CT simulations. For each pregnant patient model, the TCM curves were estimated for an abdomen/pelvis scan. A CTDI_vol value was then derived for each TCM curve based on the average mA across the scan.

Table 2.

Scanning parameters used for the tube current‐modulated (TCM) scan on Siemens Definition AS64 using CAREDose4D

Parameter	Setting
kVp	120
Quality reference mAs (QRM)	200
Rotation time (s)	0.5
Pitch	1.0
Nominal collimation (mm)	19.2 (64 × 0.6 FFS)
Measured collimation (mm)	23.8
Bowtie filter	Body
HVL (mm Al)	8.2
CTDIvol (mGy/mAs)	0.078

HVL, half‐value layer.

2.C.1. Creation of TCM functions

Using the methodology described in McMillan et al.,³⁷ TCM schemes were created for each of the pregnant patient models. As described therein, the TCM schemes generated are specific to one manufacturer's TCM algorithm. A brief summary of the methods described in McMillan et al.³⁷ are provided below.

Estimating patient size using attenuation profiles

TCM typically adapts the tube current in response to the patient attenuation characteristics from the CT localizer radiograph. However, no localizer information is available for these patients. Therefore, a simulated CT radiograph was generated for each patient using MCNPX in order to acquire estimates of attenuation.

To generate the TCM schemes used in this investigation, the first step was to create a simulated abdominal/pelvic CT scan radiograph by simulating projections along the length of each patient's image data in 1‐mm increments. When combined with a simulated in air scan, the result is the generation of patient attenuation profiles in the anterior–posterior (AP) direction. Figure 2 depicts the methodology for generating the simulated CT radiograph. From these patient attenuation profiles, an estimate of the AP dimension of patient size was generated. This involved emulating methods employed by the manufacturer and included using a moving average filter on the attenuation profile and then determining the maximum attenuation from the profile.³⁹ The AP patient dimension was calculated at each table position using the maximum attenuation normalized by the linear attenuation coefficient of water at a given beam energy (assumed to be 120 kVp and having an equivalent energy of 60 keV).

Figure 2.

(a) Voxelized representation of patient model (sagittal view) and (b) simulated computed tomography (CT) radiograph anterior–posterior (AP view). The simulated CT radiograph was generated by simulating projections at 1‐mm increments along the length of the voxelized patient model and dividing the resulting projections by a reference air scan. The legend below (a) is color coded for the material designations for each voxel. [Color figure can be viewed at wileyonlinelibrary.com]

The lateral (LAT) patient dimension was estimated using data derived from the simulated AP scan radiographs described above and applying a mathematical model that involves the elimination of outside air, the CT table, and low‐attenuation regions through the application of thresholds to the patient attenuation profile.^{39, 40} Additionally, in order to generate the LAT patient estimates from the AP estimates, the CAREDose4D estimation algorithm applies a table offset correction factor to account for the patient not being at the isocenter in the image data. Once determined, these estimates of AP and LAT dimensions of patient size were used as the inputs to the methods to estimate Siemens TCM schemes described. AP and LAT patient dimension estimates were determined at each 1‐mm step.

Estimating tube current modulation schemes from attenuation data

Estimating longitudinal tube current modulation

In the CAREDose4D automatic exposure control (AEC) algorithm, tube current is first determined by comparing the actual patient attenuation from the CT radiograph to reference patient attenuation values hardcoded in the AEC alorgithm.^{38, 41} The AP and LAT water‐equivalent estimates of patient size were determined from simulated CT radiographs in the previous section. The estimation of longitudinal modulation is based on the maximum attenuation at each table position from either the AP or LAT direction. Therefore, the maximum attenuation at each table position, i, A _max (i) is determined by taking the maximum of the calculated AP and LAT size values for each tale position as shown in Eq. (1):

$$A_{\max }(i)=max\left(exp\left({\mu }_{water,kVp}\times AP(i)\right),exp\left({\mu }_{water,kVp}\times LAT(i)\right)\right)$$ (1)

where μ _water,kVp is the linear attenuation coefficient of water for a given beam energy. For this investigation, μ _water,kVp was set to 0.2/cm for a 120‐kVp beam.

After the maximum attenuation at each table position is calculated, tube current values (mA) at each table position, i, are calculated from the corresponding patient attenuation using Eq. (2):

$$mA(i)=\frac{QRM\times pitch}{t}\times {\left(\frac{A_{max}(i)}{A_{\mathit{ref}}}\right)}^b$$ (2)

where QRM is the quality reference effective tube current‐time product (effective mAs) set on the scanner by the user,⁴¹ t is the gantry rotation time, pitch is the user selected pitch value, A _max (i) is the maximum patient attenuation at each table location i, A _ref is the protocol‐specific reference attenuation value specified by the manufacturer,⁴¹ and b is a strength parameter that can be set according to individual preferences for the tube current increase and decrease. The default strength setting for all Siemens CT scanners (including all scanners used in this investigation) is “Average” and corresponds to a b value of 0.33 for attenuation greater than A _ref and 0.5 for attenuation less than A _ref .^{41, 42} Applying Eq. (2) to the patient attenuation profiles determined at each table position from the simulated radiograph yielded an estimate of the maximum tube current at each table position and corresponds to the longitudinal modulation based upon attenuation characteristics.

Estimating angular tube current modulation

This AEC algorithm also modulates the tube current angularly according to angular attenuation measurements (i.e., angular or x–y modulation).³⁸ The only patient attenuation data used were the attenuation data derived from the simulated radiograph and was based on the AP and LAT attenuation profiles described above. Attenuation values between these ordinal positions of the tube gantry were obtained through interpolation to derive an estimate of the angular attenuation. In addition, a gantry rotation time‐dependent parameter is utilized that limits the amount of modulation allowed at a given table position.⁴³

Combining longitudinal and angular modulation

Combining the longitudinal modulation scheme from Section Estimating longitudinal tube current modulation and the angular modulation scheme from Section Estimating angular tube current modulation generated an estimated tube current profile. For the estimated tube current, the tube current at each table position, i, is based on the longitudinal modulation value multiplied by the angular modulation value.³⁷ For this work, the operating limits for tube current values were 665 mA for 120 kVp. (This was the limit for a Definition AS64, but that limit is higher for later scanners). An example TCM curve for one pregnant patient model is illustrated in Fig. 3.

Figure 3.

Estimated TCM scheme for a pregnant patient who received a clinically indicated computed tomography (CT) examination. The TCM scheme is overlaid on an image of the simulated CT localizer radiograph anterior–posterior (AP orientation) of the pregnant patient. The portion of the scan range in which the fetus is located is indicated with yellow dashed lines. [Color figure can be viewed at wileyonlinelibrary.com]

2.D. Modeling FTC scans

For the FTC scans, four different 64 slice MDCT scanner models were used. These were (a) LightSpeed VCT (GE Medical Systems Waukesha, WI), (b) Brilliance 64, (Philips Medical Systems, Cleveland, OH), (c) Aquilion 64 (Toshiba Medical Systems, Inc., Otawara, Japan), and (d) Definition AS64 (Siemens Healthineers, Forchheim, Germany). For each scanner model, the equivalent source method³³ was used to determine the equivalent spectra and equivalent bowtie filtration profile. The nominal collimation, measured beam width, half‐value layer (HVL), and CTDI_vol per mAs for each scanner are described in Table 3. The following technique was used for each scanner: 120 kVp, 400 mA, 0.5‐s rotation time, and a pitch of 1. The beam collimation used was the widest available on each scanner model. For each scanner model, physical measurements were made using the 32‐cm‐diameter CTDI phantom to determine the CTDI_vol under these scan conditions (kVp, bowtie, etc.) and were reported on a mGy/mAs basis. CTDI_vol values for each scanner were achieved by multiplying CTDI_vol per mAs values by the tube current‐rotation time product (effective mAs). These scanner models were incorporated into the MC simulation tools described in Section 2.B.

Table 3.

Scanners used for fixed tube current scans and associated parameters. The nominal collimation, measured beam width, half‐value layer (HVL), and Computed Tomography Dose Index (CTDI) per mAs for the Definition AS64 listed in Table 2 of Section 2.C are presented here for comparison

Manufacturer	Model	Nominal collimation (mm)	Measured beam width (mm)	HVL (mm Al)	CTDIvol (mGy/mAs)
GE	LightSpeed VCT	40 (64 × 0.625)	42.4	7.8	0.089
Philips	Brilliance 64	40 (64 × 0.625)	43.7	8.6	0.062
Toshiba	Aquilion 64	32 (64 × 0.5)	36.9	5.5	0.123
Siemens	Definition AS64	19.2 (64 × 0.6 FFS)	23.8	8.2	0.078

2.E. Fetal dose estimates

The nomenclature employed in this study was adapted from Turner et al.²¹ Using voxelized models of patient anatomy described in Sections 2.A, MC simulations of both TCM and FTC abdomen/pelvis CT examinations were performed to estimate the absolute dose to the fetus (D _fetus ) for each pregnant patient model. For this study, D _fetus was taken as the ratio of total energy imparted to the total fetal mass.

Normalized fetal dose estimates, nD _fetus , were obtained by dividing D _fetus from the MC simulations by CTDI_vol. The differences in nD _fetus for TCM relative to FTC were compared for one scanner (Definition AS64). nD _fetus estimates were also used (a) to investigate an exponential relationship between nD _fetus and patient size in terms of D _w (Section 2.E.1), (b) to explore a bivariate linear relationship between nD _fetus and both D _w and DE _f (Section 2.E.2), and (c) to compare nD _fetus to the SSDE f‐factors. Regression equations describing the correlations between nD _fetus with either D _w or both D _w and DE _f served as the means to generate scan technique‐independent fetal dose estimates for any patient size. Additionally, for the FTC scans, $\overline{n{D}_{\mathit{fetus}}}$ was calculated by averaging the nD _fetus across the four scanners on a per patient basis and was used to investigate an exponential relationship with D _w , a bivariate relation with D _w and DE _f , and a comparison with SSDE f‐factors. For FTC scans, the coefficient of variations (CoV) across the four scanners was also calculated on a per patient basis.

2.E.1. Fetal dose conversion coefficients as a function of D_w using an exponential model

The first approach is consistent with the observed exponential relationships between CTDI_vol‐normalized organ dose and patient size demonstrated by Turner et al.²² and used in AAPM Report 204.²³ This exponential relationship between nD _fetus and D _w is defined in Eq. (3):

$$n{D}_{fetus,1}(D_w)=A_0\times \exp (-B_0\times D_w)$$ (3)

where nD _fetus,1 (D _w ) represents fetal dose conversion coefficients as a function of D _w using the exponential model and A ₀ and B ₀ are exponential regression coefficients. A ₀ and B ₀ were determined by performing the regression of nD _fetus and D _w across all patient models and were determined separately for TCM and FTC scans. For FTC scans, nD _fetus for the FTC scans from each manufacturer and $\overline{n{D}_{\mathit{fetus}}}$ were correlated separately using Eq. (3). $\overline{n{D}_{fetus,1}}(D_w)$ was used to represent the exponential model using $\overline{n{D}_{\mathit{fetus}}}$. In order to gauge the strength of these correlations, the coefficient of determination (R ² ) was used.

2.E.2. Fetal dose conversion coefficients as a function of D_w and DE_f using a bivariate linear model

The second approach is based on Angel et al.¹⁴ wherein the relationship between nD _fetus and both D _w and DE _f is defined in Eq. (4):

$$n{D}_{fetus,2}(D_w,D{E}_f)=A_1-B_1{D}_w-C_{1}D{E}_f$$ (4)

where nD _fetus,2 (D _w ,DE _f ) represents fetal dose conversion coefficients as a function of D _w and DE _f using the bivariate linear model, and A ₁ , B ₁ , and C ₁ are regression coefficients. A _1, B ₁, and C ₁ were determined by performing the regression of nD _fetus with D _w and DE _f across all patient models and were determined separately for TCM and FTC scans. The bivariate linear regression was performed using GraphPad Prism 6.00 for Mac OS X (GraphPad Software, La Jolla, CA, USA, www.graphpad.com). A user‐defined regression model was configured for the analysis of two independent variables. This regression analysis was performed using the FTC scans from each manufacturer and $\overline{n{D}_{\mathit{fetus}}}$. $\overline{n{D}_{fetus,2}}(D_w,D{E}_f)$ was used to represent the bivariate linear model using $\overline{n{D}_{\mathit{fetus}}}$. As in Section 2.E.1, the coefficient of determination (R ² ) was used to gauge the strength of the correlation.

2.E.3. Comparison of nD_fetus to SSDE f‐factors

The nD _fetus estimates from TCM and FTC were compared to the SSDE f‐factors based on the 32‐cm CTDI_vol phantom from AAPM Report 204 using the f‐factors as a reference. Eq. (5), taken from AAPM Report 204 (Equation A‐1 of the report), was used to generated the conversion factors across the patient sizes investigated in this study and is as follows:

$$SSDEf-factors=A_0\times \exp (-B_0\times D_w)$$ (5)

where A ₀  = 3.70 and B ₀  = 0.037.²³ $\overline{n{D}_{\mathit{fetus}}}$ estimates were also compared to the SSDE f‐factors. The differences relative to SSDE f‐factors were expressed in terms of percentage (%) for each patient. In addition, for the exponential model regression analyses performed in Section 2.E.1, the SSDE f‐factors were included as a point of reference with shaded regions corresponding to ±20% and ±25% of the SSDE f‐factors.

3. Results

3.A. TCM comparison to FTC for a single scanner

Table 4 contains the D _w estimates, DE _f , CTDI_vol estimates for TCM scans, absolute fetal dose (D _fetus ) values for AS64 TCM simulations and AS64 FTC simulations, normalized fetal dose (nD _fetus ) values for AS64 TCM simulations and AS64 FTC simulations, and differences in TCM nD _fetus relative to FTC nD _fetus . D _w estimates ranged from 25.3 cm to 35.6 cm. DE _f estimates ranged from 3.4 cm to 10.9 cm. The CTDI_vol for TCM ranged from 6.9 mGy to 17.3 mGy. The CTDI_vol for FTC for the AS64 was 15.6 mGy across all patients. For TCM, nD _fetus differences relative to FTC ranged from −5% to 23%, with a mean of 6% across all patients.

Table 4.

For all patients listed in Table 1, this table includes patient size metric (D _w ), fetal depth (DE _f ), patient CTDI_vol estimates for TCM scans, D _fetus from tube current‐modulated (TCM) and fixed tube current (FTC), nD _fetus from TCM and FTC, as well as TCM nD _fetus difference (%) relative to FTC. The CTDI_vol for FTC for was 15.6 mGy

ID	D w (cm)	DE f (cm)	TCM CTDIvol (mGy)	Absolute fetal dose, D fetus (mGy)		Normalize fetal dose (nD fetus )		nD fetus difference (%)
				TCM	FTC	TCM	FTC
1	33.5	10.6	15.4	18.2	14.9	1.18	0.96	23
2	25.6	4.2	8.7	14.6	23.3	1.68	1.50	12
3	28.9	7.6	6.9	10.9	25.6	1.57	1.64	−5
4	29.2	10.9	11.8	14.1	16.6	1.20	1.07	12
5	27.3	5.9	9.2	17.3	25.0	1.87	1.61	16
6	25.3	4.6	9.1	16.2	28.1	1.77	1.81	−2
7	32.0	6.5	12.7	17.8	21.2	1.40	1.37	3
8	28.0	7.1	10.1	14.5	21.7	1.43	1.40	2
9	29.6	7.7	11.2	16.5	22.3	1.47	1.43	3
10	25.9	6.7	9.1	15.0	24.3	1.65	1.56	6
11	26.6	5.6	9.5	17.3	27.6	1.83	1.77	3
12	34.6	8	13.4	15.3	16.8	1.14	1.08	6
13	30.6	4.7	15.9	22.4	21.5	1.41	1.38	2
14	35.6	6.3	15.6	17.0	15.9	1.09	1.02	7
15	29.7	5.6	8.9	13.3	22.3	1.50	1.44	5
16	28.2	6.6	9.5	14.0	22.5	1.47	1.45	1
17	27.9	2.5	12.7	19.8	23.8	1.56	1.53	2
18	27.9	9	8.0	11.6	22.5	1.45	1.45	1
19	30.8	3.6	11.7	17.9	23.1	1.52	1.48	3
20	35.6	6	17.3	20.2	16.8	1.17	1.08	8
21	34.0	5.5	13.6	17.5	18.6	1.29	1.20	8
22	31.7	3.5	16.7	21.8	18.9	1.30	1.22	7
23	28.5	5.1	11.5	15.5	20.5	1.36	1.32	3
24	35.3	3.4	16.1	20.1	17.4	1.24	1.12	11
							Mean	6

3.B. Fetal dose conversion coefficients as a function of D _w using an exponential model

3.B.1. TCM and FTC exponential relationships for a single scanner

Figure 4 shows the exponential model regression analysis for both TCM and FTC simulations for the AS64 nD _fetus data tabulated in Table 4. The R ² values of the TCM and FTC was observed to be 0.73 and 0.70, respectively. The results of the regression analysis are tabulated in Table 5.

Figure 4.

nD _fetus,1 (D _w ) for the tube current‐modulated ( TCM) and fixed tube current (FTC) scans from the AS64. nD _fetus,1 (D _w ) represents the exponential model using nD _fetus and D _w for TCM and FTC scans. [Color figure can be viewed at wileyonlinelibrary.com]

Table 5.

nD _fetus,1 (D _w ) and R ² values for the AS64 tube current‐modulated (TCM) and fixed tube current (FTC) scans

Normalized dose	A 0	B 0	R 2
AS64 TCM	4.68	0.040	0.73
AS64 FTC	5.28	0.045	0.70

3.B.2. FTC exponential relationships for four scanners

For the FTC scan protocol described in Section 2.E, the CTDI_vol for the LightSpeed VCT, Brilliance 64, and Aquilion 64 scanners was 17.7 mGy, 12.5 mGy, and 24.6 mGy, respectively, across all patients. As mentioned in Section 3.A, the CTDI_vol for FTC for the AS64 was 15.6 mGy across all patients. The nD _fetus for each of the four scanners in FTC mode and $\overline{n{D}_{\mathit{fetus}}}$ are tabulated in Table 6. The CoV ranged from 10% to 14%. The mean CoV across all patients was 12%. Figure 5 contains the FTC regression analyses for the four scanners. The R ² values for the exponential model for the LightSpeed VCT, Brilliance 64, and Aquilion 64 in FTC mode were observed to be 0.63, 0.60, and 0.64, respectively. From Section 3.B.1, the R ² value of the AS64 in FTC mode was observed to be 0.70. The exponential regression coefficients and coefficients of determination for FTC fetal dose estimates from the four scanners are shown in Table 7. Figure 6 shows regression analysis of $\overline{n{D}_{fetus,1}}(D_w)$, which yielded an R ² of 0.64.

Table 6.

The results below are only for fixed tube current (FTC) scans and include nD _fetus for the four scanners, $\overline{n{D}_{\mathit{fetus}}}$, and coefficient of variations (CoV) across the four scanners. The FTC nD _fetus for the AS64 was included for comparison. CoV results reflect the variation among the four scanner models on a per patient basis

ID	nD fetus				$\overline{n{D}_{\mathit{fetus}}}$	CoV (%)
	LightSpeed VCT	Brilliance 64	Aquilion 64	AS64
1	0.89	0.68	0.85	0.96	0.84	14
2	1.36	1.11	1.36	1.50	1.33	12
3	1.54	1.20	1.51	1.64	1.47	13
4	0.98	0.75	0.93	1.07	0.93	14
5	1.52	1.27	1.55	1.61	1.48	10
6	1.73	1.42	1.74	1.81	1.67	10
7	1.30	1.03	1.25	1.37	1.24	12
8	1.32	1.03	1.29	1.40	1.26	13
9	1.35	1.07	1.34	1.43	1.30	12
10	1.48	1.15	1.45	1.56	1.41	13
11	1.71	1.37	1.68	1.77	1.64	11
12	1.02	0.79	0.98	1.08	0.97	13
13	1.34	1.09	1.32	1.38	1.28	10
14	0.97	0.76	0.92	1.02	0.92	12
15	1.38	1.08	1.33	1.44	1.31	12
16	1.38	1.09	1.34	1.45	1.32	12
17	1.49	1.21	1.45	1.53	1.42	10
18	1.38	1.07	1.33	1.45	1.31	12
19	1.45	1.18	1.41	1.48	1.38	10
20	1.04	0.81	0.99	1.08	0.98	12
21	1.16	0.90	1.11	1.20	1.09	12
22	1.19	0.95	1.13	1.22	1.12	11
23	1.27	1.00	1.22	1.32	1.20	12
24	1.10	0.88	1.05	1.12	1.04	10
				Mean	1.25	12

Figure 5.

The results here are only for fixed tube current (FTC) scans. nD _fetus,1 (D _w ) for the (a) LightSpeed VCT, (b) Brilliance 64, (c) Aquilion 64, and (d) AS64. The CTDI _vol values for the four scanners were 17.7 mGy for the LightSpeed VCT, 12.5 mGy for the Brilliance, 24.6 mGy for the Aquilion, and 15.6 mGy for the AS64. nD _fetus,1 (D _w ) represents the exponential model using nD _fetus and D_w for each of the four scanners. [Color figure can be viewed at wileyonlinelibrary.com]

Table 7.

nD _fetus,1 (D _w ) and R ² for the four scanners, along with $\overline{n{D}_{fetus,1}}(D_w)$ and its R ² values. The AS64 fixed tube current (FTC) nD _fetus,1 (D _w ) regression coefficients are shown here for comparison

Conversion Coefficients	A 0	B 0	R 2
LightSpeed VCT FTC	4.78	0.044	0.63
Brilliance 64 FTC	4.09	0.046	0.60
Aquilion 64 FTC	5.21	0.047	0.64
AS64 FTC	5.28	0.045	0.70
$\overline{n{D}_{fetus,1}}(D_w)$	4.82	0.046	0.64

Figure 6.

Regression analysis for $\overline{n{D}_{fetus,1}}(D_w)$ using the exponential model. $\overline{n{D}_{fetus,1}}(D_w)$ represents the exponential model using $\overline{n{D}_{\mathit{fetus}}}$ and D _w .

3.C. Fetal dose conversion coefficients as a function of D_w and DE_f using a bivariate linear model

Table 8 contains the results for the multivariate regression analyses for the all the scanners used in this investigation. The multivariate regression using the bivariate linear model for nD _fetus from TCM was observed to have an R ² value of 0.78. The R ² values for nD _fetus,2 (D _w ,DE _f ) for the LightSpeed VCT, Brilliance 64, Aquilion 64, and AS64 in FTC mode were observed to be 0.74, 0.76, 0.75, and 0.77, respectively. The R ² value of $\overline{n{D}_{fetus,2}}(D_w,D{E}_f)$ was observed to be 0.75.

Table 8.

nD _fetus,2 (D _w ,DE _f ) regression coefficients and R ² values for the bivariate linear models. $\overline{n{D}_{fetus,2}}(D_w,D{E}_f)$ regression coefficients are also included

Conversion coefficients	A 1	B 1	C 1	R 2
AS64 TCM	3.26	0.055	0.026	0.78
LightSpeed VCT FTC	3.13	0.053	0.036	0.74
Brilliance 64 FTC	2.61	0.045	0.036	0.76
Aquilion 64 FTC	3.22	0.057	0.037	0.75
AS64 FTC	3.32	0.059	0.030	0.77
$\overline{n{D}_{fetus,2}}(D_w,D{E}_f)$	3.07	0.053	0.035	0.75

TCM: tube current‐modulated; FTC, fixed tube current.

3.D. Comparisons to SSDE f‐factors

3.D.1. SSDE f‐factor comparison to TCM and FTC nD_fetus from a single scanner

Comparison of the AS64 TCM and FTC nD _fetus to SSDE are tabulated in Table 9. Figure  7 shows the same exponential model regression analyses shown in Fig. 4 with the addition of the SSDE f‐factors as a point of reference. Shaded areas corresponding to ±20% and ±25% of the SSDE f‐factors were also added. For TCM (n = 24), 15 patients in this study had nD _fetus within ±20% of the SSDE f‐factors, and 21 patients had nD _fetus within ±25% of the SSDE f‐factors. For FTC (n = 24), 20 patients in this study had nD _fetus within ±20% of the SSDE f‐factors, and 22 patients had nD _fetus within ±25% of the SSDE f‐factors. When considering both TCM and FTC scans (n = 48), 35 instances of nD _fetus were within ±20% of the SSDE f‐factors, and 43 instances of nD _fetus were within ±25% of the f‐factors. Table 10 contains a summary of the relation of nD _fetus to the SSDE f‐factors per Fig. 7.

Table 9.

Comparison of the AS64 tube current‐modulated (TCM) and fixed tube current (FTC) nD _fetus to the size‐specific dose estimate (SSDE) f‐factors

ID	AS64 TCM nD fetus	AS64 FTC nD fetus	SSDE f‐factors	Difference from the SSDE f‐factors (%)
				AS64 TCM	AS64 FTC
1	1.18	0.96	1.08	9	−12
2	1.68	1.50	1.45	16	3
3	1.57	1.64	1.28	22	28
4	1.20	1.07	1.31	‐9	−16
5	1.87	1.61	1.36	38	19
6	1.77	1.81	1.46	21	24
7	1.40	1.37	1.14	22	19
8	1.43	1.40	1.33	8	5
9	1.47	1.43	1.25	18	15
10	1.65	1.56	1.43	16	9
11	1.83	1.77	1.40	31	27
12	1.14	1.08	1.04	10	4
13	1.41	1.38	1.20	17	15
14	1.09	1.02	1.00	9	2
15	1.50	1.44	1.25	21	15
16	1.47	1.45	1.32	11	10
17	1.56	1.53	1.33	17	15
18	1.45	1.45	1.33	9	9
19	1.52	1.48	1.19	27	22
20	1.17	1.08	1.00	16	7
21	1.29	1.20	1.06	21	13
22	1.30	1.22	1.16	12	5
23	1.36	1.32	1.30	4	1
24	1.24	1.12	1.01	23	11
			Mean	17	10

Figure 7.

The same regression analyses shown in Fig. 4 accompanied by the size‐specific dose estimate (SSDE) f‐factors from American Association of Physicists in Medicine (AAPM) Report 204 as a point of reference and shaded areas corresponding to ±20% and ±25% of the SSDE f‐factors. A summary of the doses that fall within ±20% and ±25% of the SSDE f‐factors are tabulated in Table 10. [Color figure can be viewed at wileyonlinelibrary.com]

Table 10.

Summary table of nD _fetus points within the bounds of ±20% and ±25% of the size‐specific dose estimate (SSDE) f‐factors

	TCM (n = 24)	FTC (n = 24)	TCM + FTC (n = 48)
nD fetus within ±20%	15 (62.5%)	20 (83.3%)	35 (72.9%)
nD fetus within ±25%	21 (87.5%)	22 (91.7%)	43 (89.6%)
nD fetus beyond ±25%	3 (12.5%)	2 (8.3%)	5 (10.4%)

3.D.2. SSDE f‐factor comparison to $\overline{n{D}_{\mathit{fetus}}}$

Table XI contains the comparisons $\overline{n{D}_{\mathit{fetus}}}$ to the SSDE f‐factors on a per patient basis. The differences with respect to the SSDE f‐factors ranged from −29% to 17%. A mean difference of 0.4% was observed between $\overline{n{D}_{\mathit{fetus}}}$ and the SSDE f‐factors. Figure 8 shows the regression analyses for the four scanners shown in Fig. 5 with the addition of the SSDE f‐factors as a point of reference. Shaded areas corresponding to ±20% and ±25% of the SSDE f‐factors were also added to Fig. 8. All but two patients had $\overline{n{D}_{\mathit{fetus}}}$ that were within the ±20% tolerance specified by AAPM Report 204. Figure 9 shows $\overline{n{D}_{fetus,1}}(D_w)$ with the addition of the SSDE f‐factors as a point of reference and shaded regions corresponding to ±20% and ±25% of the SSDE f‐factors.

Table 11.

Comparison of $\overline{n{D}_{\mathit{fetus}}}$ to the size‐specific dose estimate (SSDE) f‐factors

ID	$\overline{n{D}_{\mathit{fetus}}}$	SSDE f‐factors	Difference from the SSDE f‐factors (%)
1	0.84	1.08	−22
2	1.33	1.45	−8
3	1.47	1.28	15
4	0.93	1.31	−29
5	1.48	1.36	9
6	1.67	1.46	15
7	1.24	1.14	8
8	1.26	1.33	−5
9	1.30	1.25	4
10	1.41	1.43	−1
11	1.64	1.40	17
12	0.97	1.04	−7
13	1.28	1.20	7
14	0.92	1.00	−8
15	1.31	1.25	5
16	1.32	1.32	0
17	1.42	1.33	7
18	1.31	1.33	−2
19	1.38	1.19	16
20	0.98	1.00	−2
21	1.09	1.06	2
22	1.12	1.16	−3
23	1.20	1.30	−8
24	1.04	1.01	2
		Mean	0.4

Figure 8.

nD _fetus,1 (D_w) for the (a) LightSpeed VCT, (b) Brilliance 64, (c) Aquilion 64, and (d) Definition AS64 shown in Fig. 5 with the size‐specific dose estimate (SSDE) f‐factors from American Association of Physicists in Medicine (AAPM) Report 204 included as a point of reference. In addition, shaded areas corresponding to ±20% and ±25% of the SSDE f‐factors are also shown. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 9.

$\overline{n{D}_{fetus,1}}(D_w)$ accompanied with the size‐specific dose estimate (SSDE) f‐factors and the shaded regions corresponding to ±20% and ±25% of the SSDE f‐factors for comparison. [Color figure can be viewed at wileyonlinelibrary.com]

4. Discussion

In this work, MC simulation methods were applied to 24 pregnant patient models (originally developed by Angel et al.¹⁴) in order to estimate fetal dose from CT abdomen/pelvis exams using TCM and FTC. TCM was applied to one scanner model for which TCM schemes could be estimated. Additionally, this scanner model and three additional scanner models were used to estimate fetal dose from FTC scans. For both TCM and FTC scans, the resulting fetal doses were normalized by scanner output (CTDI_vol) and parameterized with respect to water equivalent diameter (D _w ) to create size‐specific, scan technique‐independent fetal dose estimates. A bivariate linear model was also investigated correlating normalized fetal dose with D _w and a metric of fetal position in terms of fetal depth (DE _f ). The resulting fetal dose coefficients were then compared to the SSDE conversion coefficients (the SSDE f‐factors from AAPM Report 204).

As described in Section 2.A, D _w was measured at the image containing the three‐dimensional geometric centroid of the fetus or surrogate organ (being uterus and gestational sac), the same location used by Angel et al.¹⁴ The image data for these patients were not originally reconstructed at the maximum available field of view (FOV). Because of this, nearly all of the patients (n = 22) had small portions of peripheral anatomy outside of the FOV. Since the voxelized models were based upon the image data, the calculated D _w from the simulated topogram underestimates D _w for these cases. This underestimation of AP and LAT dimensions from anatomy outside of the FOV also affects the inputs necessary to estimate TCM schemes. Underestimated patient size yields lower tube current values. The decreased tube current values would also yield decreased CTDI_vol estimates from the average tube current across the entire scan range. However, the voxelized model itself will also be affected as there will be less tissue to attenuate the photons in the simulations. Therefore, the net effect of a slightly underestimated patient size due to small portions of peripheral anatomy outside the FOV on nD _fetus should have minimal impact to the study results.

When comparing TCM to FTC (performed for the Definition AS64 in this study), the coefficients of determination of 0.73 and 0.70, for nD _fetus,1 (D _w ) for both TCM and FTC, respectively, suggest that D _w explains much of the variation in nD _fetus . nD _fetus,2 (D _w ,DE _f ) using both D _w and DE _f for TCM yielded a coefficient of determination of 0.78. This finding suggests that a knowledge of D _w and DE _f may give a better estimate for fetal dose than D _w alone. As can be noted in Fig. 4 and in Table 4, the TCM conversion coefficients are systematically greater than the FTC conversion coefficients by roughly 6% on average. The increase in fetal dose from TCM relative to FTC is most probably due to AEC response to pelvic anatomy. The fetal extent included the pelvis for most of the patients, as is shown in Fig. 3. As such, the fetus experienced an elevated tube current due to the attenuation of the pelvis. However, TCM was only simulated for one TCM technology, so it is not clear whether this 6% would be observable for all TCM technologies. Further study would therefore be needed to ascertain if this difference exists with other AEC systems.

For the Definition AS64, TCM conversion coefficients of the fetus and surrogate organs (when the fetus was not present) were observed to be greater than the SSDE conversion coefficients by 17%, as can be seen from Fig. 7 and Table 9. The higher conversion coefficients would imply a higher absolute fetal dose relative to SSDE for a given CTDI_vol. One potential reason for the conversion coefficients being higher than the SSDE f‐factors is that the f‐factors were based on the average absorbed dose to organs located in the abdomen using MC simulations of FTC abdomen protocols for voxelized phantom models.^{21, 22, 23} The soft tissue organs within are effectively water equivalent in composition, in contrast to fetal anatomy, which is comprised both of water equivalent soft tissue voxels and bone voxels. The mass energy‐absorption coefficients of bone to water is greater than unity, meaning the absorbed dose to the fetus should be higher than absorbed dose to any of the abdominal organs,²⁰ so for a given CTDI_vol, the nD _fetus should be greater than the normalized dose to abdominal organs. On the other hand, the surrogate organs, the uterus and gestational sac (5 of the 24 patients), also experienced normalized doses higher than SSDE conversion coefficients. These two organs are, as described in Section 2.B, comprised of soft tissue and water, respectively, unlike fetal anatomy. Therefore, what these results suggest is that the variance from the SSDE f‐factors may be scanner specific. This conclusion is additionally supported by the variability observed for nD _fetus for FTC simulations of the other scanners in Section 3.B.²¹ These imply variance from the SSDE f‐factors can be related to properties of the scanner itself, such as the scanner x‐ray source.

However, for the Definition AS64, the majority of patients in this study, for both TCM and FTC, had nD _fetus within ±20% of the SSDE f‐factors and within ±25% of the SSDE f‐factors, as Table 10 shows. SSDE was never intended to be applied to fetal dose estimates. However, AAPM Report 204 stipulates a 10%–20% tolerance of estimated patient dose from size‐specific, scan technique‐independent conversion factors, and actual patient dose.²³ Although there were some patients that had differences from the SSDE f‐factors greater that 20%, results from this study indicate that normalized doses from both TCM and FTC are mostly within this tolerance range. This suggests that, within the patient size range used in this study, the SSDE f‐factors can provide a reasonable (within ±25%) estimate of normalized fetal dose estimates for both TCM and FTC abdominal/pelvis scans.

Additionally, this study also investigated nD _fetus for FTC averaged across the four different scanner manufacturers $(\overline{n{D}_{\mathit{fetus}}})$. In the case of $\overline{n{D}_{fetus,\mathit{1}}}(D_w)$, the coefficient of determination was observed to be 0.64 with respect to D _w , meaning D _w explains roughly two‐thirds of the variation seen in $\overline{n{D}_{\mathit{fetus}}}$. The regression analyses using the exponential model performed for each scanner shown in Fig. 5 highlight the observed variability in normalized dose across scanners as discussed above,²¹ albeit within 15%.

$\overline{n{D}_{fetus,\mathit{2}}}(D_w,D{E}_f)$ yielded a coefficient of determination of 0.75. Only two patients (ID1 and ID4), both of which are early‐term patients, had $\overline{n{D}_{\mathit{fetus}}}$ beyond the tolerance specified in AAPM Report 204. One possible explanation for this observation is that, for these two patients, the DE _f was larger relative (10.6 and 10.9 cm, respectively) to the other patients. As can be seen in Fig. 10, the larger DE _f implies that the organs of interest for these two patients are positioned deeper within the pelvis as compared with two other early‐term patients (ID5 and ID3) and are thus provided with more inherent shielding, which decreased their normalized dose. Given this variation in fetal position in early‐term patients, the inclusion of DE _f in the bivariate model may explain the improvement of correlative ability over the exponential model.

Figure 10.

Axial and sagittal images showing the variability in early‐term maternal anatomy. For ID1 in (a) and ID4 in (b), the greater DE _f means that the uterus (yellow) and gestational sac (green), respectively, are situated deeper within the pelvis and hence provided the fetus more shielding. For ID5 in (c), the uterus and gestation sac extend anteriorly and for ID3 in (d), a distended bladder (outlined in cyan) pushes uterus more anteriorly. [Color figure can be viewed at wileyonlinelibrary.com]

Across patients, the mean $\overline{n{D}_{\mathit{fetus}}}$ difference from the SSDE f‐factors was observed to be 0.4%. Furthermore, the curves representing $\overline{n{D}_{fetus,1}}(D_w)$ and the SSDE f‐factors are fairly similar within the patient size range of this study, as can be seen in Fig.  9. These points further buttress the conclusion made above, namely, that within the patient size range used in this study, for FTC scans, the SSDE f‐factors can provide a reasonable estimate of nD _fetus . This result is intuitive given that AAPM Report 204 was concerned with estimating the dose to the central region of abdomen CT exams using FTC and derived the CTDI_vol conversion coefficients by incorporating normalized dose values across multiple scanner manufacturers.²³ However, this being said, for early‐term patients, the position of the uterus or gestational sac can vary due to a variety of factors, as highlighted in Fig. 10. Therefore, a wider tolerance from SSDE may be necessary to account for the variability in early‐term maternal anatomy.

5. Conclusion

Results from this study suggest that the fetal dose from both TCM and FTC CT scans of pregnant patients of various gestational ages and patient sizes may be reasonably estimated using models that incorporate (A) scanner‐reported CTDI_vol and (B) with D _w as a metric for patient size metric to account for patient size variation. Moreover, more accurate estimates of fetal dose can be obtained with knowledge of DE _f if it is available. The results from this study also imply that the SSDE f‐factors can provide a reasonable (within ±25%) estimate of normalized fetal dose across scanners for the range of patient sizes investigated herein.

There are still a few important limitations worth mentioning. The first is that, as detailed in Section 2.E, the TCM data for the patients in this study were based off the attenuation characteristics from a simulated CT localizer and estimates from one manufacturer's AEC algorithm. Ideally, patient D _w information and TCM data directly from the scanner would be available. However, since they were not available for these patients, estimates of patient D _w from simulations were generated. Moreover, estimating the AEC algorithm of other manufacturers is beyond the scope of this work. The second limitation is that, as mentioned above, the variability in early‐term maternal anatomy can have an effect on the ability of SSDE to serve as a surrogate for fetal dose. This study, however, only considered five early‐term patients. The data in this study suggest that a detailed investigation of fetal dose in early‐term pregnant patients is warranted and hence will be the subject of future work. A third limitation is that only scans of the abdomen/pelvis region were considered. Head and chest scans of the patients were not considered for this study as the pregnant patient models used in this study did not include this anatomy. To extend this work to scans of other anatomic regions, whole‐body patient models of maternal anatomy would be needed such as the RPI pregnant patient models.¹⁸ However, fetal dose contributions from head or chest scans are expected to be negligible.^{44, 45} Lastly, the available image resolution was not sufficient to investigate dose to specific fetal organs such as the thyroid and red bone marrow. This study averaged the dose across the entire fetal volume and thus included the dose to individual developing organs into a single estimate of fetal dose. Extending this work could therefore also include investigating dose to developing fetal organs, provided that the resolution of fetal anatomy is sufficient enough to make this a possibility.