Patient-specific dose estimation for pediatric chest CT

Abstract

Current methods for organ and effective dose estimations in pediatric CT are largely patient generic. Physical phantoms and computer models have only been developed for standard/limited patient sizes at discrete ages (e.g., 0, 1, 5, 10, $15\text{years}$ old) and do not reflect the variability of patient anatomy and body habitus within the same size/age group. In this investigation, full-body computer models of seven pediatric patients in the same size/protocol group (weight: $11.9–18.2\mathrm{kg}$) were created based on the patients’ actual multi-detector array CT (MDCT) data. Organs and structures in the scan coverage were individually segmented. Other organs and structures were created by morphing existing adult models (developed from visible human data) to match the framework defined by the segmented organs, referencing the organ volume and anthropometry data in ICRP Publication 89. Organ and effective dose of these patients from a chest MDCT scan protocol (64 slice LightSpeed VCT scanner, $120\mathrm{kVp}$, 70 or $75\mathrm{mA}$, $0.4\mathrm{s}$ gantry rotation period, pitch of 1.375, $20\mathrm{mm}$ beam collimation, and small body scan field-of-view) was calculated using a Monte Carlo program previously developed and validated to simulate radiation transport in the same CT system. The seven patients had normalized effective dose of $3.7–5.3\mathrm{mSv}∕100\mathrm{mAs}$ (coefficient of variation: 10.8%). Normalized lung dose and heart dose were $10.4–12.6\mathrm{mGy}∕100\mathrm{mAs}$ and $11.2–13.3\mathrm{mGy}∕100\mathrm{mAs}$, respectively. Organ dose variations across the patients were generally small for large organs in the scan coverage $(<7\%)$, but large for small organs in the scan coverage (9%–18%) and for partially or indirectly exposed organs (11%–77%). Normalized effective dose correlated weakly with body weight (correlation coefficient:$r=-0.80$). Normalized lung dose and heart dose correlated strongly with mid-chest equivalent diameter (lung: $r=-0.99$, heart: $r=-0.93$); these strong correlation relationships can be used to estimate patient-specific organ dose for any other patient in the same size/protocol group who undergoes the chest scan. In summary, this work reported the first assessment of dose variations across pediatric CT patients in the same size/protocol group due to the variability of patient anatomy and body habitus and provided a previously unavailable method for patient-specific organ dose estimation, which will help in assessing patient risk and optimizing dose reduction strategies, including the development of scan protocols.

INTRODUCTION

With the growing use of computed tomography (CT) in children^{1, 2} and the increasing awareness of CT radiation risk to this population,^{3, 4, 5} there is a greater need to accurately estimate radiation dose from CT examinations, not only for the purpose of assessing life-time cancer risk, which can be useful during discussions with healthcare providers, regulatory bodies, parents, and ethics committees, but also for the purposes of comparing and optimizing CT technologies and scan protocols. Despite the recent debate on its concept,⁶ effective dose remains the most widely used dose descriptor for radiological procedures including CT. Its calculation requires the knowledge of dose delivered to individual organs.

With no practical technique to measure organ dose directly from patients, three methods are currently being used to estimate organ and effective dose from CT examinations using patient phantoms/models: (a) experimental measurement on physical anthropomorphic phantoms,^{7, 8} (b) Monte Carlo simulation using mathematical or voxelized models of patients,^{9, 10, 11, 12, 13, 14, 15, 16} and (c) calculation of organ and effective dose from CT dose index (CTDI) or dose-length product (DLP) using conversion coefficients derived via the first two methods.^{7, 8, 9, 17, 18, 19} Current methods are limited in that they are largely patient generic. Phantoms/models have only been developed for standard/limited patient sizes at discrete ages (e.g., 0, 1, 5, 10, $15\text{years}$ old) and do not reflect the variability of patient anatomy and body habitus. Therefore, dose information for individual patients is currently not available. In a recent study,¹⁶ the effect of body weight on organ dose was studied for $15\text{year}$ old adolescents using nonuniform rational B-spline (NURBS) based computer models created at 10th, 50th, and 90th weight percentiles. The authors reported up to $\sim 30\%$ dose errors when reference patient models were used to represent overweight patients. This highlighted the need for patient-specific dose estimations.

The goal of this work is to use multi-detector array CT (MDCT) data of multiple pediatric patients in the same size/protocol group to investigate dose variations across patients due to the variability of patient anatomy and body habitus and to explore methods for patient-specific dose estimations.

MATERIALS AND METHODS

Patients

This study was approved by our institutional review board (IRB), who determined that it was in compliance with the Health Insurance Portability and Accountability Act, and did not require informed consent. The study included seven pediatric patients (three boys and four girls; median age, $2\text{years}$ old; age range, $1–6\text{years}$ old; median weight, $12.9\mathrm{kg}$; weight range, $11.9–18.2\mathrm{kg}$) who underwent 64 slice MDCT examinations (LightSpeed VCT, GE Healthcare, Waukesha, WI) of the chest, abdomen, and pelvis.

Patient-specific computer models

In CT examinations, three sources of exposure contribute to organ dose: direct primary exposure, exposure from overranging (additional scan length necessary for data interpolation in helical reconstruction), and scattered radiation. While the first source of exposure has higher contribution than the other two and can be modeled with the actual patient CT data, the other two sources contribute notably to organ dose as well and can only be adequately modeled using full-body patient models.

A full-body computer model of each patient was created to enable dose estimations for organs both inside and outside the image volume. The initial anatomy of the model was defined by segmenting the patient’s MDCT data using a software application developed in our laboratory.²⁰ The heart, liver, gall bladder, stomach, spleen, and kidneys were manually segmented by contouring from each CT slice. The lungs and bones were semi-automatically segmented using thresholding. Once a dataset was segmented, three-dimensional polygon models were generated for each structure using the marching cubes algorithm.^{21, 22} Three-dimensional NURBS surfaces were then fit to the polygon models using NURBS modeling software (Rhinoceros, McNeel North America, Seattle, WA) to create the initial patient-specific model.

Other organs and structures, not easily segmented or visible in the scan coverage, were defined by morphing an existing male or female full-body adult model (developed from visible human data)²³ to match the framework defined by the segmented organs. The morphing was performed manually using the affine transformations of Rhinoceros. The volumes of the organs and structures defined in this manner for each pediatric model were checked and scaled, if necessary, to match age-interpolated organ volume and anthropometry data in ICRP Publication 89.²⁴

The resultant full-body pediatric male and female models possessed a total of 26 and 27 organs, respectively, including most of the radiosensitive organs defined by ICRP Publication 103²⁵ (Table 1). Figure 1 illustrates surface rendered views of the three-dimensional anatomy in the computer models of the youngest ($16\text{months}$ old) and the oldest ($6\text{years}$ old) patients in our study.

Table 1.

Summary of organs included in the computer models of the seven pediatric patients.

Organ/structure	Density$\left(\mathrm{g}{\mathrm{cm}}^{-3}\right)$	Material(ICRU 46)	Mass (g)14
			Mean	(Range)	CV13
Respiratory system
Pharynx-larynx1	1.03	Average soft tissue11	7.2	(4.3–10.0)	27.6%
Trachea-bronchi	1.03	Average soft tissue	4.7	(2.6–6.6)	31.2%
Lungs2	0.26	Lung (adult, healthy, inflated)	158.8	(117.6–216.6)	22.4%
Alimentary system
Esophagus3	1.03	Average soft tissue	6.7	(1.9–10.2)	50.8%
Stomach2	1.03	Average soft tissue	124.4	(61.9–219.6)	46.2%
Pancreas	1.03	Average soft tissue	33.7	(26.4–52.9)	27.2%
Liver2	1.03	Average soft tissue	484.4	(354.2–724.0)	25.7%
Gall bladder2	1.03	Average soft tissue	6.8	(2.6–12.4)	54.4%
Small intestine	1.03	Average soft tissue	238.0	(190.5–343.9)	22.6%
Large intestine	1.03	Average soft tissue	192.1	(157.6–291.7)	23.9%
Circulatory system
Heart2 4, 1n4	1.03	Average soft tissue	166.7	(133.6–248.7)	23.4%
Urogenital system
Kidneys2	1.03	Average soft tissue	109.0	(81.3–147.6)	22.5%
Urinary bladder	1.03	Average soft tissue	14.8	(10.4–22.4)	26.4%
Prostate5	1.03	Average soft tissue	1.6	(1.6–1.7)	0.8%
Testes	1.03	Average soft tissue	3.3	(3.2–3.4)	3.3%
Ovaries	1.03	Average soft tissue	2.4	(1.8–3.3)	29.5%
Uterus	1.03	Average soft tissue	3.3	(2.6–4.3)	23.3%
Vagina	1.03	Average soft tissue	1.6	(1.4–2.1)	20.8%
Skeletal system6
Compact bone2 7, 1n7	1.75	Cortical $\left(5\text{years}\right)$	2303.3	(1875.7–2989.5)	16.9%
Marrow	1.03	Red marrow (adult)12	424.5	(270.8–701.5)	38.5%
Integumentary system
Skin (torso only)8	1.03	Average soft tissue	440.8	(375.9–535.5)	12.5%
Additional organs/tissues
Brain	1.03	Average soft tissue	1161.6	(1069.4–1345.0)	8.0%
Eyes	1.03	Average soft tissue	12.0	(10.3–15.4)	14.2%
Thyroid	1.03	Average soft tissue	4.1	(2.9–7.5)	39.9%
Breasts9	0.96	Breast $(50∕50)$	3.0	(3.0–3.1)	1.1%
Thymus	1.03	Average soft tissue	35.6	(33.3–38.3)	5.5%
Spleen2	1.03	Average soft tissue	86.1	(34.2–129.7)	42.0%
Adrenal glands	1.03	Average soft tissue	7.8	(7.3–9.0)	8.3%
Residual soft tissues10	1.03	Average soft tissue	10 026.0	(7728.8–15 045.4)	25.1%

¹Dose to combined organ of pharynx and larynx was used as a surrogate for dose to salivary glands, oral mucosa, and extra-thoracic (ET) region.

²Organs individually segmented from CT images of the patients. Individually segmented compact bones included the backbone and the ribcage.

³Esophagus, combined organ of pharynx and larynx, and combined organ of trachea and bronchi were modeled as walled organs with air-fill lumens.

⁴Heart, gall bladder, alimentary tract organs (stomach, small intestine, large intestine), and urinary bladder were modeled as single homogenous organs without delineation of walls and contents.

⁵Prostate, testes, ovaries, uterus, and vagina are gender-specific organs and were included in the models of their respective genders only.

⁶The skeletons were modeled as homogeneous bone marrow encased by homogeneous compact bone. The trabecular bone was not explicitly modeled. The additional dose to bone marrow deposited by photoelectrons released in the trabecular bone was accounted for by applying dose enhancement factors. The dose enhancement factors reported by King and Spiers (Ref. ³²) were interpolated by age at $50\mathrm{keV}$, effective energy of the $120\mathrm{kVp}$ beam.

⁷Dose to compact bone was used to approximate dose to bone surface.

⁸Skin thickness, wall thicknesses of trachea and esophagus, and thicknesses of compact bones were assumed to be $3\mathrm{mm}$ given the $2\mathrm{mm}$ voxel resolution used in this study and the thin thicknesses of these structures in patients of this age range.

⁹Breast tissue is underdeveloped in patients of this age range and was not visible in the CT images. A small amount of breast tissue $\left(3.0\mathrm{g}\right)$ was “attached” to the chest of each patient model for dose estimation purposes. Dose to the breast was used to study dose variations across the patients, but was not included in the calculation of effective dose.

¹⁰Residual soft tissues include skeletal muscle, adipose tissue, cartilage, blood, lymphatic tissues, and connective tissues. Dose to residual soft tissues was used to approximate dose to skeletal muscle and lymphatic nodes.

¹¹Average soft tissue of adult male was used.

¹²Bone marrow was assumed to be composed entirely of red bone marrow for patients of this age range.

¹³CV $\left(\text{coefficient}\text{of}\text{variation}\right)=\text{standard}\text{deviation}\times 100\%$/mean.

¹⁴Organ/tissue mass after voxelization at 2 mm isotropic resolution.

Figure 1.

Surface rendered views of the three-dimensional anatomy in the computer models of the youngest ($16\text{months}$ old) and the oldest ($6\text{years}$ old) patients in our study.

A computer model of the CT table (table case and table interior) was also created via manual segmentation of the table from an adult CT image with a large scan field-of-view, referencing the dimensional data provided by the manufacturer. The model of each patient was “positioned” on the table in a supine position with arms elevated above the head to mimic actual patient posture during CT examinations.

The NURBS model of each patient with the table attached was voxelized at $2\mathrm{mm}$ isotropic resolution, resulting in a three-dimensional matrix of voxels, each assigned an integer labeling a specific organ or object.

Monte Carlo simulations

The voxelized models of the patients were used as inputs to a PENELOPE²⁶ (Universitat de Barcelona, version 2006) based Monte Carlo code, which was previously developed to simulate radiation transport in the LightSpeed VCT scanner.²⁷ The three-dimensional geometry of the three bowtie filters on the scanner and the trajectories of x-ray tube motion during axial and helical scans were explicitly modeled by the code. The accuracy of the code was previously validated against experimental measurements in terms of dose distributions in a cylindrical acrylic phantom, and the maximum dose error was found to be less than 5.4%.²⁷

Because it is impractical and inefficient to individually define all the planes and voxels in the patient model using the original geometry routine PENGEOM of PENELOPE, we developed a new geometry routine, named PENVOME (i.e.,PENGEOM for voxelized models). PENVOME conveniently labels each voxel by its matrix indices; boundary planes of the voxel are only calculated when the voxel is reached by a particle. This circumvents the need to store surface/body definitions and to sort through a genealogical tree of a large number of bodies. The accuracy of PENVOME was validated against PENGEOM in terms of simulated dose in a simple object of 18 voxels, and the results were identical within the statistical constraints of the Monte Carlo simulation.

Before incorporating the voxelized model of each patient into Monte Carlo simulation, each organ was assigned a material (Table 1). The case and the interior of the patient table were modeled as carbon fiber $(\rho =1.7\mathrm{g}∕{\mathrm{cm}}^3)$ and acrylic foam $(\rho =0.1\mathrm{g}∕{\mathrm{cm}}^3)$, respectively. The MATERIAL program of PENELOPE was used to generate material definition files based on the elemental composition and mass density information tabulated in ICRU Publication 46.²⁸

The simulated scan protocol was the standard, size-based chest scan protocol in place at our institution for the weight range of the patients: $120\mathrm{kVp}$, $70\mathrm{mA}$$(11.5–14.4\mathrm{kg})$ or $75\mathrm{mA}$$(14.5–18.4\mathrm{kg})$, $0.4\mathrm{s}$ gantry rotation period, pitch of 1.375, $20\mathrm{mm}$ beam collimation, and small body scan field-of-view (corresponding to small bowtie filter). While $40\mathrm{mm}$ beam collimation is commonly used on the LightSpeed VCT scanner to achieve fast scanning, this protocol used a $20\mathrm{mm}$ beam collimation because of its higher dose efficiency for small $(\lesssim 20\mathrm{cm})$ total scan length.

The total scan length of each patient was calculated as the total image coverage plus the overranging distance. The total image coverage was defined as the distance from $1\mathrm{cm}$ above lung apex to $1\mathrm{cm}$ below lung base, typical of the clinical image coverage for a chest scan. The overranging distance was estimated from the scanner console parameters as “table speed $(\mathrm{cm}∕\mathrm{s})\times \text{total}$ scan time $\left(s\right)-$image coverage (cm).” Estimates were found to be independent of slice thickness, reconstruction interval, and image coverage. The start location of the chest scan was, therefore, $1\mathrm{cm}$ plus half of the overranging distance above lung apex, and the end location was the same distance below lung base.

In terms of simulation time, using a single processor on a $2.3\mathrm{GHz}$ Linux server with $4\mathrm{GB}$ of RAM, a $30\min$ runtime was needed to finish $10\text{million}$ photon histories, resulting in percent dose error (1 σ standard $\text{deviation}\times 100\%$/mean) of less than 1% for all organs in the scan coverage and less than 15% for other organs.

Effective dose calculations

The effective dose of each patient was calculated as the summation of radiosensitive organ dose values weighted by the tissue weighting factors defined by ICRP Publication 103.²⁵ Dose to radiosensitive organs that were not explicitly modeled was approximated by dose to neighboring organs (footnotes of Table 1). Complying with ICRP Publication 103, the weighting factor for the remainder organs was applied to the arithmetic mean dose of the 13 remainder organs for each gender.

Data analysis

Variations in organ dose across the patients were quantified by the coefficient of variation (standard$\text{deviation}\times 100\%$/mean) for selected organs both inside and outside the chest scan coverage. The coefficient of variation across the patients was also calculated for the effective dose.

Organ dose was correlated with chest size and organ volume using linear regression analysis. Chest size was expressed in terms of total scan length, a surrogate for chest length, and mid-chest equivalent diameter defined as the diameter of a circle having the same area as the mid-chest (half-way between lung apex and lung base) area of the patient model. Effective dose was correlated with chest size and body weight of the patient model using linear regression analysis.

RESULTS

Figure 2 depicts coronal dose distributions resultant from the chest MDCT scan in three $2\text{-year}$-old patients. Compact bones inside the scan coverage had the highest dose values, a result of having the highest mass energy absorption coefficient. Dose to chest organs was similar for the three patients. However, dose to abdominal organs varied substantially among the patients, affected by the locations of those organs relative to the base of the lung.

Figure 2.

Coronal dose distributions in three $2\text{-year}$-old patients, determined from the chest MDCT scan. The coronal plane was taken about half-way in between the anterior and posterior surfaces of each patient. The computer model of each patient with organs shown on a gray scale was overlaid with a semi-transparent image of the normalized dose distribution on a colored scale.

The normalized absorbed dose received by large organs in the scan coverage, i.e., the lung, the heart, and the thymus, varied very little across the patients (5.7%–6.2%) (Table 2). Greater dose variations across the patients were observed for small organs in the scan coverage, i.e., the esophagus (8.6%), the breast (17.7%), and the thyroid (9.1%), but they were generally smaller compared with those for partially or indirectly exposed organs (10.7%–76.6%) (Table 2). The correlation between dose variations and variations in organ mass was weak (Pearson correlation coefficient: $r=0.43$). For all the selected organs, the dose error was less than 1%. Therefore, the variations in organ dose reported here were due to the variations in patient anatomy and body habitus, not the uncertainty in dose estimation. The seven patients had a chest MDCT effective dose of $1.1–1.5\mathrm{mSv}$, corresponding to normalized effective dose of $3.7–5.3\mathrm{mSv}∕100\mathrm{mAs}$ (coefficient of variation: 10.8%).

Table 2.

Variations across patients in normalized organ dose from chest MDCT scan and correlations of normalized organ dose with chest size and organ volume for selected organs.

	Normalized organ dose$(\mathrm{mGy}∕100\mathrm{mAs})$			Pearson correlation coefficient $\left(r\right)$2
	Mean	(Range)	CV1 (%)	Chest size		Organvolume
				Mid-chestequivalent diameter	Totalscan length3
Directly exposed large organs
Lung	12.0	(10.4–12.6)	6.2%	$-\mathbf{0.99}$	$-0.45$	$-0.51$
Heart	12.9	(11.2–13.3)	5.7%	$-\mathbf{0.93}$	$-0.14$	$-\mathbf{0.94}$
Thymus	11.9	(10.6–12.8)	5.8%	$-0.76$	$-0.25$	$-0.70$
Directly exposed small organs
Esophagus	9.9	(8.6–10.8)	8.6%	$-0.49$	0.15	$-0.07$
Breast	9.7	(7.2–11.7)	17.7%	$-0.79$	$-0.42$	⋯4
Thyroid	9.5	(8.5–11.0)	9.1%	0.11	$-0.29$	0.17
Partially or indirectly exposed organs
Liver	8.0	(6.7–9.5)	16.3%	$-0.52$	0.47	$-0.53$
Gall bladder	3.8	(1.4–8.2)	76.6%	$-0.41$	0.40	0.18
Stomach	7.9	(4.8–10.9)	26.6%	$-0.46$	0.13	$-0.79$
Spleen	7.1	(5.3–10.0)	22.1%	0.07	0.54	$-0.11$
Kidney	2.6	(1.4–4.3)	34.9%	0.03	0.61	0.61
Marrow	2.1	(1.9–2.5)	10.7%	$-0.64$	$-0.12$	$-0.88$

¹CV $\left(\text{coefficient}\text{of}\text{variation}\right)=\text{standard}\text{deviation}\times 100\%$/mean.

²The square of r equals $R^2$, a measure of goodness-of-fit for linear regression analysis. r values larger than 0.90, corresponding to $R^2>0.81$, are highlighted in bold.

³Total scan length was used as a surrogate for chest length.

⁴Correlation of organ dose with organ volume was not calculated for the breast because all patients were arbitrarily assigned the same amount of breast tissue.

Regression analysis of organ dose with respect to chest size indicated that lung dose and heart dose correlated strongly with mid-chest equivalent diameter (lung: $r=-0.99$, heart: $r=-0.93$), but weakly with total scan length (Table 2). Normalized absorbed dose to the lung $\left(D_{\text{lung}}\right)$ and the heart $\left(D_{\text{lung}}\right)$ decreased with mid-chest equivalent diameter $\left(d_{\text{mid-chest}}\right)$ as

$$D_{\text{lung}}=22.13-0.60d_{\text{mid-chest}},$$ (1)

and

$$D_{\text{heart}}=22.26-0.56d_{\text{mid-chest}},$$ (2)

where $D_{\text{lung}∕\text{heart}}$ and $d_{\text{mid-chest}}$ are in the units of $\mathrm{mGy}∕100\mathrm{mAs}$ and cm, respectively (Fig. 3). For other organs, the correlations between normalized absorbed dose and chest size were generally weak (Table 2).

Figure 3.

Normalized absorbed dose $(\mathrm{mGy}∕100\mathrm{mAs})$ to the lung and the heart from the chest MDCT scan as a function of mid-chest equivalent diameter and the results of regression analysis. The mid-chest equivalent diameter was defined as the diameter of a circle having the same area as the mid-chest (half-way between lung apex and lung base) area of the patient model.

Regression analysis of organ dose with respect to organ volume showed that heart dose correlated strongly with heart volume $(r=-0.94)$, but for all the other organs, the correlation of dose with organ volume was generally weak (Table 2).

Normalized effective dose decreased with body weight (Fig. 4) as

$$D_{\mathrm{eff}}=6.38-0.11w$$ (3)

and with mid-chest equivalent diameter as

$$D_{\mathrm{eff}}=8.87-0.26d_{\text{mid-chest}},$$ (4)

where $D_{\mathrm{eff}}$ and w are in the units of $\mathrm{mSv}∕100\mathrm{mAs}$ and kg, respectively. But the correlations were weak, r equal to $-0.80$ and $-0.64$ for body weight and mid-chest equivalent diameter, respectively. Normalized effective dose had no apparent correlation with total scan length $(r=0.20)$.

Figure 4.

Normalized effective dose $(\mathrm{mSv}∕100\mathrm{mAs})$ from the chest MDCT scan as a function of patient model weight and the result of regression analysis. Because the weights of the patient models were not perfectly matched to the actual patient weights (discrepancy $<5\mathrm{kg}$), the weights of the patient models were used here for the regression analysis.

For all the correlation relationships studied above, the trends of the data did not suggest the use of nonlinear models.

DISCUSSION

In this work, patient-specific computer models were constructed using the patients’ actual MDCT data, which enabled the calculation of dose values that were patient specific. Using computer models of multiple pediatric patients in the same size/protocol group, we obtained knowledge of dose variations across patients due to the variability of patient anatomy and body habitus, which could not be obtained using previous methods of DLP, Monte Carlo simulation, or physical phantom measurement. We found that, in chest MDCT, variations in normalized organ dose across patients were generally small for large organs in the scan coverage, but large for small organs in the scan coverage and for partially or indirectly exposed organs. Furthermore, dose variations correlated weakly with variations in organ mass, hence organ volume, indicating that variations in organ shape, organ location, and body habitus (e.g., breadth) also contribute to the dose variations across patients. It is necessary to further examine these trends of organ dose variations in future studies of abdominal and pelvis MDCT scans.

Our work also provided a method to determine patient-specific dose information for any other patient in the same size/protocol group (those not included in the study). The formula relating lung dose to mid-chest equivalent diameter (derivable from mid-chest circumference) allows patient-specific lung dose to be estimated with high accuracy. The relationship between heart dose and mid-chest equivalent diameter may also be used for a broad estimation of patient-specific heart dose. The correlation between normalized dose to large organs in the scan coverage and mid-chest equivalent diameter was not surprising. Nickoloff et al.²⁹ proposed and verified an exponential relationship between CT dose index (CTDI) and phantom diameter over a wide range of phantom sizes ($5–30\mathrm{cm}$ in diameter, representing newborn to large adult). The linear relationships that we found in this study may be understood as an approximation to the exponential relationship over a narrow range of body diameters. The weak correlations between organ dose and patient dimensions (mid-chest equivalent diameter and chest length) for small organs in the scan coverage and for partially or indirectly exposed organs suggested that dose to those organs may only be obtained with high accuracy using patient-specific computer models.

Normalized effective dose was found to decrease with patient weight, consistent with the finding of DeMarco et al.¹⁴ who studied the effective dose from a chest scan for eight patients with a much wider weight range $(5–100\mathrm{kg})$. In either study, the correlation between normalized effective dose and patient weight was weak (DeMarco study: $R^2=0.41$ or $r=-0.64$, our study: $r=-0.79$) and did not warrant inferences to other patients.

We note that this investigation did not use the concept of “effective dose” as originally defined. Effective dose is defined in ICRP Publication 103²⁵ for a reference person; the tissue weighting factors are mean values representing an average over many individuals of different genders and age groups. Our application of the concept of effective dose to individual patients, however, while not being exactly “correct,” is in line with other studies in literature¹⁴ and provides a surrogate for patient risk, using a concept that the medical imaging community is familiar with. For a more proper estimation of patient-specific risk, Martin³⁰ has recommended the use of age-, sex-, and organ-specific risk coefficients. Perhaps the recently proposed concept of “effective risk”³¹ could substitute effective dose for future patient-specific risk estimations.

CONCLUSION

Using computer models of multiple pediatric patients in the same size/protocol group created from the patients’ clinical MDCT data, we reported the first assessment of organ and effective dose variations across patients in the same size/protocol group due to the variability of patient anatomy and body habitus. More importantly, this work provided a previously unavailable method for patient-specific organ dose estimations that should prove useful in assessing patient risk and optimizing CT technologies and scan protocols.