Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Jun 2.
Published in final edited form as: Radiat Phys Chem Oxf Engl 1993. 2025 Jun 2;237:113013. doi: 10.1016/j.radphyschem.2025.113013

Patient-specific organ dose calculation using automatic segmentation and extension of CT scans

Sergio Morató 1, Belén Juste 2, Rafael Miró 2, Agustín Santos 3, Andreu Badal 4, Choonsik Lee 1, Gumersindo Verdú 2, Matthew M Mille 1
PMCID: PMC12379956  NIHMSID: NIHMS2097535  PMID: 40873456

Abstract

The primary objective of this study is to develop a fast and automated methodology for calculating personalized radiation organ dose from computed tomography (CT) scans using anatomical models derived from the CT images of the patient obtained during the imaging procedure. To validate this method, a comparison was conducted between experimentally measured dose values in a CT scan and those obtained through Monte Carlo simulation. Multiple point dose measurements were taken within a RANDO phantom during a Siemens Somatom Emotion CT scan, employing Metal Oxide Semiconductor Field Effect Transistor dosimeters. The software tools described in this work facilitate automatic Monte Carlo simulations using DICOM metadata and images from patient CT scans. A personalized patient-specific 3D voxelized model was created based on the DICOM images, and the radiation dose was estimated using the MC-GPU Monte Carlo code, which supports parallelization on a GPU. The required inputs for the simulations were automatically extracted from the DICOM data. Additionally, this study outlines the procedure for calculating the dose in actual patients (where the scans only cover a fraction of the body) and presents a comparison of the results with NCICT dose calculator, which uses a library of computational phantoms. A previously developed anatomical extension technique was applied to realistically extend the partially captured patient anatomy. Without this technique, the mean dose to organs that are only partially included in the CT scan is likely overestimated. Furthermore, a realistic case that uses a complete CT scan anatomy considering 134 anatomical structures or organs is presented. The results demonstrate a consistent trend when comparing simulated and experimental values, with a median difference of 5.7%. When comparing the simulated results with NCICT estimations, we observe similar magnitudes of values overall. However, significant dose differences up to 51% are noted for some organs, underscoring the importance of personalizing the dose calculation.

Keywords: Computed Tomography, Dosimetry, Anatomical Predictive Extension, Monte Carlo, MC-GPU

1. Introduction

Although the benefits of x-ray computed tomography (CT) are widely recognized, in recent years there has been an increased focus on patient safety owing to the growing number of CT scans performed and rare, but highly publicized, instances of patient overexposure (Domino 2024; Zarembo 2009). Furthermore, a number of epidemiological studies have identified an associated cancer risk with the use of CT scanners (Brenner and Hall 2007; Berrington de González et al. 2009; Smoll et al. 2023; Hauptmann et al. 2023; Meulepas et al. 2019; Journy et al. 2017; Krille et al. 2015; Huang et al. 2014; Mathews et al. 2013; Pearce et al. 2012). The greatest attention is directed towards children because they are more sensitive to radiation compared to adults and have a longer life expectancy (Foucault et al. 2022; Thierry-Chef et al. 2021; Brenner et al. 2001; Brody et al. 2007).

In response, various national and international groups have recommended, or in some cases mandated, the reporting of dose from CT scans. Example mandates include the California state law (effective July 1, 2012) (Legislature 2010), Texas state law (effective May 1, 2013) (Services 2013), and the European Atomic Energy Community Treaty directed (effective February 6, 2018) (Council 2014). However, the only dose measures readily available for reporting are the CT Dose Index (CTDI) or metrics derived from the CTDI such as the volume CTDI (CTDIvol) or Dose Length Product (DLP). These parameters are useful measures of the radiation output of CT scanners, but are not the same as patient organ dose (Boone et al. 2012). Current scanners provide a generalized dose calculation that satisfies legal requirements. Nonetheless, there is a growing interest in maintaining patient-specific organ dose records which would help inform patients of their individual exposure while also supporting research on the risks and benefits of CT scans (Simon et al. 2006).

There are two main approaches for reporting patient CT organ dose. The first uses specialized software which rely on libraries of pre-calculated dose conversion coefficients for computational phantoms. Example software are ImPACT CT Patient Dosimetry Calculator (Sawyer et al. 2009; ImPACT 2011) and CT-Expo (Stamm and Nagel 2002) which use simplified stylistic phantoms, or VirtualDose CT (Ding et al. 2015) and NCICT (Lee et al. 2015; Lee, Yeom, and Folio 2022) which use more realistic, state-of-the-art phantoms. However, the patient’s anatomy inevitably differs from the computational phantom used, and any such discrepancy can result in errors in the estimated dose.

A second, more personalized dose calculation approach, is to derive an anatomical model of the patient directly from the CT images of the patient (Lee et al. 2020; Peng et al. 2020). This method allows for a tailored dose calculation, but has several challenges: (1) A personalized Monte Carlo dose calculation needs to be run for each patient and most radiation transport software are too slow to be practical for clinical applications; (2) The CT images need organ segmentations for reporting organ dose; (3) If the dose calculation is performed on a partial-body CT, then the calculation is limited to the scan coverage and does not account for scatter radiation to tissues beyond the scan range. This second approach is likely the future of personalized CT dosimetry; however, despite its potential, to our knowledge no single software comprehensively addresses all three of these technical hurdles.

To address this important gap, in this paper we implemented three technologies into CT organ dose calculations: Graphical Processing Unit (GPU)-based Monte Carlo calculation, automatic organ segmentation, and anatomy extension. We also conducted experimental validation of the Monte Carlo calculations by comparing with measurements performed by placing MOSFET dosimeters inside a physical anthropomorphic phantom.

2. Materials and methods

We developed a method to automatically calculate patient-specific organ doses directly from CT images in Digital Imaging and Communications in Medicine (DICOM) format. A voxelized model of the patient is generated to perform a personalized radiation transport simulation of the CT scan procedure.

2.1. Personalized CT dose calculations

A Graphical User Interface (GUI)-based software was developed using MATLAB (Mathworks, Natick, MA) programming language to convert user-provided DICOM CT images into an input file for performing a personalized CT dose calculation using the Monte Carlo code MC-GPU (Badal and Badano 2009; Badal et al. 2021; U.S. Food and Drug Administration 2012). The CT dose calculation is performed in three major steps as illustrated in Fig. 1: (1) generation of Monte Carlo input file and voxel patient anatomy, (2) execution of the Monte Carlo simulation, and (3) reporting of organ doses.

Figure 1.

Figure 1.

Open in a new tab

Workflow diagram showing the process for patient-specific Monte Carlo organ dose calculations using patient DICOM CT images in three steps: (1) create input files for a Monte Carlo simulation in the MC-GPU code using methods of automatic segmentation and anatomy extension, (2) execution of the MC-GPU simulation, and (3) reporting of organ dose results.

2.1.1. Extraction of CT scan parameters

Our software extracts the CT scan parameters from the DICOM CT header and generates an input file for the MC-GPU code to perform a realistic simulation of the patient’s CT scan. Key geometric parameters extracted to define the CT scan include the distance between the CT x-ray source and isocenter, x-ray fan beam angle, total collimation width in the cross-plane direction, scan pitch, scan start and stop position, source rotation direction and rotation speed. Other important parameters include the x-ray tube voltage and current settings. As some parameters may be missing from the DICOM CT header, our software allows the user the option to manually enter or revise suggested parameters.

2.1.2. GPU Monte Carlo calculations

To accelerate the CT dose calculations, we used MC-GPU v1.3, a GPU-based Monte Carlo radiation transport code for simulating radiography devices (Badal and Badano 2009; Badal et al. 2021; U.S. Food and Drug Administration 2012). The code uses CUDA programming (NVIDIA corporation 2025) to leverage GPU multi-threading to significantly accelerate the Monte Carlo radiation transport simulation by processing thousands of independent x-ray tracks in parallel on each GPU. The software has options for creating synthetic CT projection x-ray images of a voxelized anatomy and for tallying dose. The continuous CT source trajectory is modeled as a discrete number of projections. By default, our software GUI assumes that the x-ray source rotates 1 degree around the patient between projections, however, this can be modified by the user. The user must also specify the number of source photon histories to simulate. At the end of the simulation, the MC-GPU code reports the average dose in each voxel and in each defined material (and their associated statistical uncertainties). MC-GPU uses the kerma approximation for dose calculation, assuming that all secondary electrons generated during photoelectric and Compton events are absorbed locally. The x-ray interaction models and cross sections are adopted from PENELOPE 2006 (Salvat, J.M., and J. 2006; NEA 2019).

A limitation of using MC-GPU to simulate CT scans is that the current version of the code does not model the scanner’s bowtie filter, which is important for CT dose calculations. The bowtie filter serves to reduce radiation exposure to the periphery of the patient while maintaining good image quality. To address this limitation, we modified the MC-GPU source code to allow for angular modulation of the x-ray emission in the horizontal direction. In the new source model with CT bowtie extension, the initial energy and emission angle of each x-ray track are sampled jointly to account for the angular intensity profile and the corresponding beam hardening caused by varying filter thickness at different angles. Instead of a one-dimensional energy spectrum, the user provides a two-dimensional distribution describing the probability of x-ray emission at a number of angular and energy bins. After random sampling the angular and energy bin, the initial x-ray energy is uniformly sampled within the bin energy interval, and the emission angle is sampled with a linear interpolation between the emission angle of consecutive bins, creating a smooth profile in the simulated flat-field projections.

2.1.3. Patient-specific voxel phantom

Our method involves generating a patient-specific voxel phantom from a user-provided CT scan. The following capabilities were combined to prepare the voxel phantom for the MC-GPU simulation.

Automatic organ segmentation:

We applied the TotalSegmentator tool (Wasserthal 2023) to the patient DICOM CT images. TotalSegmentator uses a state-of-the-art deep learning-based automatic segmentation algorithm and GPU parallelization to segment over 220 different organs and tissues throughout the entire body in ~2 minutes. The resulting segmentations have standardized names and are stored in DICOM and NIfTI file format. The resulting organ segmentation also facilitates the organ dose reporting as described in section 2.1.5.

Extending partial-body CT images:

As the patient CT is typically not a whole-body image, the CT images need to be extended to account for scatter dose to tissues beyond the scan coverage. This is also necessary for accurate reporting of dose to organs partially or fully outside the CT scan coverage. For this purpose we adopted our previously published method for extending patient CT images called Anatomical Predictive Extension (APE) (Morató et al. 2024). This method uses a library of CT images from which surrogate images are selected. Image registration is performed through comparison of the skeletons of the incomplete CT with those of selected patient or phantom from the library. The images showing closest similarity are transformed and appended to the incomplete CT. TotalSegmentator is also applied to the appended CT images so that a new DICOM RT Structure file can be written combining structures from the incomplete and extended portions of the phantom. Details on this method can be found in our previous publication (Morató et al. 2024).

Material assignment:

We assigned appropriate materials (elemental composition and physical density) to the segmented regions by matching them to the organs defined by the International Commission on Radiological Protection (ICRP 2002). Any remaining space within the patient outer body contour is assigned soft tissue or air. Cross-section files for photons up to 130 keV were generated for the various tissues based on the PENELOPE database using the auxiliary utility “MC-GPU_create_material_data.f” which was included with the MC-GPU package.

2.1.4. Absolute dose calculation

The MC-GPU code reports dose in units of eV g−1 per photon history emitted. However, the exact number of photons emitted by the CT scanner is not known. Instead, the radiation output of the scanner is typically reported by the scanner through the CTDIvol, a standardized quantity defined as:

CTDIvol(mGy)=1P13CTDI100,center+23CTDI100,periphery, (1)

where P is the scan pitch, and CTDI100, center and CTDI100, periphery are the integrated dose in mGy over a 100 mm long ionization chamber placed at the central or peripheral slots of a standard CTDI cylinder phantom (16 or 32 cm diameter for head or body imaging).

To calculate dose in units of mGy, the dose results of the patient simulation (Dpatient, MC-GPU) are normalized by that of a simulated CTDIvol calculation (CTDIvol, MC-GPU) obtained through additional Monte Carlo simulations according to the following equation:

PatientDose(mGy)=Dpatient,MC−GPUeVgperhistoryCTDIvol,MC−GPUeVgperhistory×CTDIvol,scanner(mGy)×R, (2)

where CTDIvol, scanner (mGy) is the CTDIvol of the patient’s CT scan as reported by the scanner. An estimate of the CTDIvol is often displayed on the computer console before the scan. After the examination, this value can usually be found in the DICOM CT header or DICOM Structured Dose Report. The unitless factor R represents the number of full revolutions of the x-ray source included in the simulation of the patient’s helical scan. It adjusts for the difference between the helical scan, which involves multiple revolutions, and the CTDIvol, which is based on a single axial revolution. The factor R is calculated from the CT scan parameters as R= L⁄(N × T × P), where L is the scan length, N × T is the nominal beam width for a multi-slice CT scan with N simultaneously acquired slices of thickness T (in units of mm), and P is the pitch.

Calculation of CTDIvol, MC-GPU requires additional simulations for the same CT scanning technique applied to a standard CTDI cylinder phantom (IAEA 2007). Two simulations are performed with the ion chamber placed at either the central or peripheral slots of the cylinder phantom. The results from the two simulations are combined as follows:

CTDIvol,MC−GPU(eVg−1perhistory)=1P100mmN×T13Dcenter,MC−GPU+23Dperiphery,MC−GPU, (3)

where Dcenter, MC-GPU and Dperiphery, MC-GPU are the average dose in the ion chamber calculated in the MC-GPU simulation and N × T is the nominal beam width. The difference between equation 1 and 3 arises from the fact that the CTDI100 is a measure of the integrated dose over the 100 mm long ion chamber whereas the Monte Carlo simulation reports the average dose in the ion chamber volume. The calculated conversion factor for the simulated dosimetry results (equation 3) needs to be re-calculated if there is any change in the energy spectra, the bowtie filter profile, or the field-of-view (collimation). A typical value of CTDIvol, MC-GPU in this study was approximately 18 eV g−1 per photon history.

2.1.5. Organ dose reporting

The MC-GPU code reports dose to each voxel in the phantom. Two 3D matrix files are generated: one with the average dose in each voxel and the other with the corresponding Monte Carlo statistical uncertainty. In addition, MC-GPU reports the average dose in each material to the standard output. Mean organ dose and dose-volume statistics were calculated by post-processing the 3D voxel data using the organ label maps generated by TotalSegmentator. Our software GUI provides a 2D dose visualization for each slice of the CT scan and a table summarizing the calculated dose for each organ. In addition, we converted the voxel dose data to VTK format for visualization in ParaView (Ahrens et al. 2005).

2.2. Experimental validation

We validated our dose calculation method by comparing our simulation results with experimental measurements from CT scans of a physical phantom. We used a RANDO phantom representing a woman with height 163 cm and weight 54 kg. The phantom consisted of 49 slices, each 2.5 cm thick, made of several materials including a human skeleton (average density 1.61 g cm−3) and two synthetic materials imitating soft tissue (density 0.977 g cm−3) and lung (density 0.26 g cm−3).

The phantom was prepared by placing ten Metal Oxide Semiconductor Field Effect Transistor (MOSFET) dosimeters inside designated holes for measuring point dose. Two mobile MOSFET readers (Best Medical Canada, Ontario, Canada) were connected to the dosimeters. Each reader consisted of a set of five MOSFET dosimeters (model TN-502RD-H), a remote monitoring dose verification software, wall-mounted Bluetooth transceiver, and a small reader module which communicates with the MOSFET dosimeters where the user can record dose data on the computer.

The phantom was scanned on a Siemens Somatom Emotion CT scanner at Radiophysics Service, Provincial Hospital Consortium of Castellón and the resulting dose measurements and images were collected for analysis. Repeated CT scans were peformed for the pelvis and chest sections of the body phantom, respectively (Fig. 2). The locations of dosimeters for each case are shown in Fig. 3. The measurement locations were selected to allow dose comparison throughout the phantom and in all material regions. The CT scans were conducted using a tube potential of 130 kVp and current of 120 mA. The CTDIvol for the scan was 13.44 mGy. Other scan parameters included a pitch factor of 1, exposure time of 1000 ms, table feed of 19.2 mm per rotation and a table speed of 19 mm/s.

Figure 2.

Figure 2.

Open in a new tab

The experimental setup for performing the CT dose measurements for the pelvis (left) and chest (right) scans. The RANDO phantom is placed in the CT gantry with ten MOSFET dosimeters inserted for measuring dose at specific points inside the phantom.

Figure 3.

Figure 3.

Open in a new tab

The locations of the ten MOSFET dosimeters placed inside the RANDO phantom for the (a) pelvis and (b) chest scan setups. For each case color photographs of the physical phantom are shown above the corresponding CT images.

The MOSFET dosimeters used in this study were previously calibrated in terms of air kerma at the Ionizing Radiation Metrology Laboratory of the National Dosimetry Center of Spain, and the measurement errors for each dosimeter are below 5%, in accordance with the manufacturer’s specifications and calibration results. To compare with the MC-GPU results, the measurements were converted to absorbed dose in the various materials by multiplying by the ratio of the mass energy absorption coefficient in the material to that of air. As MC-GPU does not provide the x-ray spectrum at the dosimeter location, we used the mean energy of the initial x-ray spectrum as an approximation. The calculated conversion factors were 4.6, 1.07 and 1.08 for bone, soft tissue, and lung, respectively.

2.3. Example patient CT dose calculations

2.3.1. Clinical data source

To evaluate the performance of our CT dose calculation method, we applied it to two distinct cases: (1) abdominal CT images of an 8-year-old male patient (height: 144 cm, weight: 61 kg), and (2) a head-to-pelvis CT scan of a 10-year-old female patient (height: 150 cm, weight: 54.6 kg). Both anonymized datasets were obtained from the Pediatric-CT-SEG dataset which is publicly available on the National Cancer Institute’s Cancer Imaging Archive (Clark et al. 2013). The first case, involving the abdominal CT scan, served as an example for applying our anatomy extension method, simplifying the anatomy to 13 distinct material classifications. This patient was selected for his extreme size for his age so as to demonstrate the benefits of personalized dosimetry. The second case provided a more complex scenario, simulating a comprehensive anatomical model with 134 organs or structures defined.

2.3.2. CT scan parameters

For the dose calculation we assumed the CT images were acquired with a Siemens Emotion 16 scanner with a tube voltage of 120 kVp and a tube current of 243 mA. We assumed that the distance from the source to data collection center was 59.5 cm, total collimation of 16 × 1.92 mm, spiral pitch of 1, and clockwise source rotation at a speed of 1.0105 s per rotation. The CT fan beam angle was 45.58 degrees in the in-plane direction and 1.85 degrees in the cross-plane direction. The body CTDIvol reported by the scanner was 20.6 mGy. For the 8-year-old patient case, the continuous helical scan was approximated using 4,050 projections, while the 10-year-old patient case used 15,094 projections. For each case the x-ray source rotated 1 degree per projection. The start and stop positions of the scan were defined by the z-coordinates corresponding to the edges of the first and last slices in the patient’s CT scan. The scan lengths were L=21.6 cm for the 8-year-old patient and L=80.5cm for the 10-year-old patient. The 120 kVp energy spectrum was obtained using Siemens’ manufacturer-provided spectra simulator, accessible through their official website (Siemens Healthineers 2024). This simulator is based on x-ray spectra standards outlined in official publications (Fewell 1981; Boone and Seibert 1997). The bowtie filter shape was derived from information provided by the scanner manufacturer on the radiation fluence through filter by fan beam angle. We assumed the filter was made of polymethyl methacrylate (PMMA) with density of 1.19 g cm−3. The x-ray source was defined in MC-GPU with 98 energy bins (1 keV increment) and 15 angular bins (2 degree increment).

2.3.3. Patient modeling

Automatic organ segmentation:

Out of 220 organs and tissues segmented by TotalSegmentator, for the 8-year-old patient case we included 13 organs and tissues in the current study for demonstration purposes: 11 organs (spleen, kidneys, liver, stomach, aorta, pancreas, lungs, esophagus, heart, colon, and certain muscles) and 2 bone structures (vertebrae and ribs). The remaining space within the patient body was filled with soft tissue and air. The 10-year-old patient case included 134 defined organs or structures.

Extending partial-body CT images:

The abdominal patient CT selected included only partial coverage of the lungs and heart. Therefore, we extended the patient CT images in the anterior direction using our APE method (Morató et al. 2024) for more accurate dose calculation in the chest. Surrogate images for extending the patient scan were selected from Pediatric-CT-SEG dataset (Jordan et al. 2022) which includes CT images of 359 pediatric patients. The CT scan for the 10-year-old female fully included nearly all organs of interest so no extension was necessary.

2.3.4. Dose calculations

We calculated the organ mass and mean organ doses for both the incomplete and extended patient anatomies. We also compared the results of our dose calculations with NCICT (Lee et al. 2015; Lee, Yeom, and Folio 2022), an established CT dose calculator. The NCICT dose calculation used similar scan parameters and the NCI computational phantom having closest height and weight to the patient was selected. We also performed dose calculations in MC-GPU using the same NCI phantom to evaluate the differences between our dose calculation method and NCICT.

3. Results

3.1. MC-GPU input generator

Figure 4 shows the MATLAB GUI developed to streamline the CT dose calculation workflow. The software integrates various functionalities, such as importing CT images, visualizing CT images and structures, and configuring the Monte Carlo simulation parameters. While default simulation parameter settings are suggested to the user, the users also have the flexibility to modify them as needed. The software also automates the dose calculation by calling the MC-GPU code (if installed on the user’s system) and displaying the simulation results upon completion. This user-friendly platform enhances efficiency in managing and preparing data for the MC-GPU simulations. The process begins by loading the patient’s CT images by selecting the folder containing the corresponding DICOM files. The software then automatically extracts all relevant information from the DICOM metadata to generate the input simulation file required by MC-GPU. Subsequently, the user can proceed by clicking the “Voxelized Geometry” button, which creates the geometry input file for MC-GPU. Once both input files are generated, the simulation can be initiated using the “Calculate Dose” button. CT images and the resulting dose distribution are displayed in the second tab of the interface. The third and fourth tabs provide access to basic and advanced simulation settings, respectively, which can be adjusted by the user prior to running the simulation. Upon completion, the software generates a detailed table reporting organ-specific dose values and produces a 3D visualization in VTK format, which can be viewed using ParaView (Kitware 2025).

Figure 4.

Figure 4.

Open in a new tab

Screenshots of the graphical user interface (GUI) of the MC-GPU input generator: import of CT images (left), visualization of CT images (middle), and input of Monte Carlo parameters (right).

3.2. Experimental validation

The high resolution of the CT images facilitates easy identification of the placement of the MOSFET dosimeters. Fig. 5a displays the voxelized geometries of the RANDO phantom generated by the code to be used in the Monte Carlo simulation. The voxel size was selected to maintain the resolution of the CT images, with the voxel size on the z-axis corresponding to the slice thickness. Notably, even the MOSFET cables have been voxelized for inclusion in the simulation, as evident in the images presented in Fig. 5a.

Figure 5.

Figure 5.

Open in a new tab

(a) Voxelized models of the RANDO phantom for the chest and pelvis cases. Soft tissue is displayed in blue and skeleton and MOSFET cables are shown in red. (b) A 3D visualization of the calculated dose for the CT scan in Paraview.

The calculated doses considering the specific material at each measured point are shown in Table 1 compared to the simulated ones. Point dose values used for comparison with experimental values were taken from the 3D dose distribution presented in Figure 5b. The specific slices and location in X and Y were taken from the 3D dose matrix corresponding to the location of the dosimeters during the experimental measurements.

Table 1.

Point dose values comparison between the Monte Carlo calculations and experimental measurements.

Dosimeter Channel Number 1 2 3 4 5 6 7 8 9 10
Pelvis Phantom material Bone Bone Soft tissue Soft tissue Bone Bone Soft tissue Soft tissue Bone Soft tissue
Calculated dose (mGy) 36.8 42.1 12.3 17.4 41.3 42.9 12.9 14.0 49.9 18.0
Measured dose (mGy) * 38.6 43.7 10.6 14.7 42.8 46.00 13.1 14.2 50.1 14.8
Difference (%) 4.9 3.6 16.1 18.6 3.4 6.7 1.7 1.5 0.5 21.3
Chest Phantom material Soft tissue Soft tissue Lung Bone Bone Soft tissue Lung Soft tissue Lung Soft tissue
Calculated dose(mGy) 11.4 16.7 15.1 46.9 47.8 15.8 16.1 16.7 13.1 15.1
Measured dose (mGy) * 11.6 15.7 13.4 47.9 47.9 18.4 16.1 13.1 12.3 13.1
Difference (%) 1.7 6.6 13.0 2.0 0.1 13.9 0.1 27.5 6.4 15.5
Open in a new tab
*

The measurement errors for each dosimeter are below 5%, in accordance with the manufacturer’s specifications and calibration results.

3.3. Example patient CT dose calculation

3.3.1. Patient-specific voxel phantom

Visualizations of the organ segmentations created by TotalSegmentator for the first 8-year-old obese abdominal CT case are shown in Fig. 6. The segmentation process was completed in approximately 2 min. The segmentations were used to produce a voxel phantom which was then extended using our APE method as shown in Fig. 7. The extension method in this case prioritized achieving seamless continuity for organs such as the heart, lungs, spinal cord, and esophagus, which, in turn, made achieving perfect continuity for the external body surface more challenging. Visualization of the organ segmentation for the 10-year-old female head-to-pelvis CT scan case is presented in Fig. 8.

Figure 6.

Figure 6.

Open in a new tab

(a) Axial, (b) perspective, (c) coronal, and (d) sagittal views of the 3D voxel model created from the abdominal CT scan of an 8-year-old obese male patient based on the organ segmentations generated by TotalSegmentator.

Figure 7.

Figure 7.

Open in a new tab

A 3D visualization of the 8-year-old obese male patient anatomy before and after extending the CT images. A detail of the detail of the lungs, heart, spinal cord, and esophagus is shown on the right.

Figure 8.

Figure 8.

Open in a new tab

A 3D visualization of the segmented anatomy for the 10-year-old female head-to-pelvis CT scan case.

3.3.2. Dose calculation results

A CT scan of each anatomical model was simulated using a NVIDIA V100-SXM2 GPU card in the BIOWULF cluster from the National Institutes of Health. This GPU card featured 32 GB of VRAM, 5120 cores, and 640 Tensor cores. A total of 106 photon histories were simulated and was sufficient to achieve negligible Monte Carlo statistical error for all organs of interest (<1%). The calculation time was 4 minutes for the 8-year-old male obese abdominal CT scan simulation which included 4,050 projections. The calculation time for the 10-year-old female head-to-pelvis scan simulation was 15 minutes with15,094 projections. The results of the dose calculation for 8-year-old obese male abdominal CT simulation are shown in Fig 9 and Table 2. A 3D visualization of the dose results for the 10-year-old female head-to-pelvis scan simulation are shown in Fig 10.

Figure 9.

Figure 9.

Open in a new tab

A 3D visualization of the Monte Carlo dose calculation results for the 8-year-old obese male abdominal CT scan: (a) CT image, (b) Segmented tissues displayed in different colors according to their assigned material, (c) Calculated dose from the CT scan and (d) Monte Carlo statistical error in the calculated dose.

Table 2.

Mean organ dose calculated by MC-GPU for the incomplete patient anatomy and the extended version of the patient for the 8-year-old obese male abdominal CT scan.

Calculated Dose (mGy)**
Incomplete CT Extended CT
Spleen 23.7 24.1
Kidneys * 21.2 21.4
Liver 24.7 25.4
Stomach 23.5 23.9
Pancreas 20.8 21.1
Lungs * 20.2 13.4
Esophagus 17.0 8.5
Heart 16.8 16.2
Colon 22.4 22.5
Open in a new tab
**

Monte Carlo statistical error for all organs of interest (<1%).

*

Refers to left and right combined.

Figure 10.

Figure 10.

Open in a new tab

A 3D rendering of the calculated CT dose for the 10-year-old female head-to-pelvis CT scan.

3.3.3. Comparison with NCICT

The calculated absorbed dose values, along with the corresponding organ mass values, are presented in Tables 3 and 4 for the 8-year-old obese male abdominal CT scan. In addition, a comparison was made between the simulated values and those obtained from the NCICT software for an NCI phantom having closest characteristics (145 cm height, 60 kg weight and male gender) to the selected patient (height: 144 cm, weight: 61 kg). Table 4 presents the dose values for the incomplete patient anatomy, the extended anatomy and dose calculated with the NCICT software. Figure 11 depicts the absorbed dose comparison among the incomplete patient anatomy, the extended version, and the NCI phantom calculated with MC-GPU and NCICT for the 8-year-old patient case. Figure12 present the absolute dose comparison between MC-GPU and NCICT for a patient with similar characteristics as the 10-year-old patient.

Table 3.

Calculated dose values for the 8-year-old obese male abdominal CT scan. The results for the Absolute dose values calculated with MC-GPU for the extended patient anatomy simulation were compared with NCICT software for a whole-body computational phantom with similar height and weight to the patient and to an MC-GPU simulation for the same NCI phantom.

Extended Patient Anatomy NCI Phantom
MC-GPU dose (mGy)** MC-GPU dose (mGy)** NCICT dose (mGy)**
Spleen 24.1 25.7 26.8
Kidneys * 21.4 26.4 28.9
Liver 25.4 23.0 23.9
Stomach 23.9 22.0 24.1
Pancreas 21.1 22.6 22.9
Lungs * 13.4 5.2 6.6
Esophagus 8.5 5.6 6.6
Heart 16.2 7.6 10.5
Colon 22.5 21.4 22.1
Open in a new tab
**

Monte Carlo statistical error for all organs of interest (<1%).

*

Refers to left and right combined.

Table 4.

Comparison of organ masses for the 8-year-old obese male abdominal CT scan case.

Organ Mass (g)
Incomplete Patient Anatomy Extended Patient Anatomy NCI Phantom
Spleen 168 168 75
Kidneys * 195 195 179
Liver 1033 1033 782
Stomach 152 152 190
Pancreas 59 59 57
Lungs * 404 852 382
Esophagus 6 19 16
Heart 176 271 132
Colon 231 231 220
Open in a new tab
*

Refers to left and right combined.

Figure 11.

Figure 11.

Open in a new tab

Comparison of organ doses calculated by MC-GPU for the incomplete and extended patient anatomies for the 8-year-old obese male abdominal CT scan. The results were also compared with the NCICT software for a whole-body computational phantom with similar height and weight to the patient and to an MC-GPU simulation for the same NCI phantom.

** Monte Carlo statistical error for all organs of interest (<1%).

* Refers to left and right combined.

Figure 12.

Figure 12.

Open in a new tab

Comparison of organ doses calculated by MC-GPU with the NCICT software for the 10-year-old female head-to-pelvis CT scan case.

*Monte Carlo statistical error for all organs of interest (<1%).

4. Discussion and conclusions

This study demonstrates a method for performing personalized CT organ dose calculations following patient image acquisition and could be used to create a dose registry for tracking and optimizing radiation exposure in clinical practice. Our method combines three pre-existing tools in a novel way: (1) GPU-accelerated Monte Carlo simulations, (2) automatic organ segmentation, and (3) partial-body CT image extension using our previously published method. To our knowledge, this study is the first to explicitly demonstrate the combination of these capabilities for estimating patient organ dose from CT scans. While similar work by (Peng et al. 2020) was pioneering, it did not address the challenge of extending partial-body CT images for calculating dose to organs partially or fully outside the CT scan range. In this study we applied our previously published method to extrapolate the missing anatomy (Morató et al. 2024). An additional strength of our dose calculation approach is that it provides a three-dimensional dose distribution throughout the patient, whereas most CT dose calculators, such as NCICT (Lee et al. 2015), only provide mean organ dose. By using the TotalSegmentator automatic segmentation method we could provide dose information for more organs than NCICT (up to 207 organs for TotalSegmentator depending on the extent of the anatomy vs 33 for NCICT).

We evaluated the accuracy of our method by comparing point dose measurements from the RANDO phantom experiments with corresponding MC-GPU calculations. The median dose difference between measurement and simulation was 5.7% (range 0.1% to 28%) and was within the expected measurement uncertainty. Key sources of uncertainty include unknowns such as the precise CT scanner’s trajectory (i.e. gantry starting position) and the photon energy spectrum at the point of measurement which is needed to derive the air kerma-to-absorbed dose conversion factors.

In the case of the 8-year-old male abdominal CT scan simulations, we found that the mean dose to the lung and esophagus were significantly overestimated when considering only the partial-body CT scan compared to the predicted extended anatomy (mean dose of 20.2 mGy vs 13.4 mGy for the lungs and 17.0 mGy vs 8.5 mGy for the esophagus). This discrepancy arises because large portions of these organs are outside the patient CT scan and thus receive a much lower dose from scattered photons. In our simulation, the lung mass for the partial-body CT case was 404 g versus 852 g for the predicted extended anatomy. For the esophagus these values were 6 g and 19 g, respectively. However, for the heart, the impact of extending the anatomy was less pronounced. Despite the heart’s 53% larger mass in the extended anatomy (271 g vs. 176 g), the calculated dose was only 3.6% smaller (16.2 mGy vs. 16.8 mGy). This can be explained by most of the missing heart mass being immediately adjacent to the last slice in the CT scan where the scatter radiation is still sizeable. As expected, for organs entirely within the partial-body CT scan (e.g., kidneys, liver, stomach, pancreas), the dose differences between the partial-body CT and extended CT cases were negligible (< 1 mGy) (Table 2).

The dose results for the 8-year-old male abdominal CT scan simulations were also in good agreement with the NCICT dose calculator for a whole-body computational phantom of similar size to the patient. For the 9 organs considered (Table 2), the median dose difference was 1.9 mGy (range 0.2 to 7.5 mGy). The median relative dose difference was 11% (range 0% to 51%). Relative dose differences >20% were observed for lungs, kidneys, heart, and esophagus. Similarly, for the 10-year-old patient scan simulations the median relative dose difference for 25 compared organs was 9% (range 0% to 28%). For this case, relative dose differences >20% were observed for the eyes, small bowel, and gallbladder.

The observed dose differences between our patient simulations and NCICT can be explained in part by organ geometry and organ mass differences (Table 4), highlighting the value of personalized CT-based dosimetry over computational phantoms. We observed some small discrepancies between NCICT and MC-GPU simulations involving the same NCI computational phantom, with median organ dose differences of 1.1 mGy (range 0.3 mGy to 2.9 mGy). These can be attributed to methodological details such as uncertainties in the placement of the scan start and stop locations in NCICT. Additionally, NCICT uses a dose table method which assumes a series of axial scans, while our MC-GPU simulations correctly account for the helical trajectory of the x-ray source. The observed differences between our patient simulation results and NCICT are too large to be fully explained by such uncertainties.

It should be noted, however, that while our dose calculation method provides a more personalized estimate, the observed dose differences between our calculations and NCICT are small (typically <8 mGy). The importance of such small differences in estimated dose remains uncertain and are the subject of scientific debate (Wilson et al. 2024). Some research suggests a linear dose-response relationship for cancer incidence at doses below 100 mGy (Hauptmann et al. 2020). Cancer patients may also receive multiple CT scans during their course of treatment and combined imaging dose can be sizeable (Ding et al. 2010; Rehani et al. 2020). Nonetheless, NCICT remains reliable a tool for estimating organ doses when specific anatomical details are unavailable, particularly prior to conducting a CT scan.

Our study has several limitations. First, while we used patient images to personalize the dose estimates, beyond the available CT scan we are extrapolating the patient’s anatomy using our extension method. Therefore, we accept that there is uncertainty in our dose estimates for organs partially or fully outside the CT scan. Future work will focus on methods to quantify these uncertainties. Second, as a proof of concept study, we did not consider some technical features of modern CT scanners such as tube current modulation (Jadick et al. 2021) or overranging, referring to the automatic extension of the scan length by CT scanners for reconstruction of the first and last images of scan (Booij, Dijkshoorn, and van Straten 2017; Christner et al. 2010; Shirasaka et al. 2012).

In this study we presented a method for performing personalized CT organ dose estimates. Our study underscores the importance of accurately quantifying patient organ doses and could support epidemiological research on CT scans to better understand the risk of low doses of radiation.

Acknowledgments

The GUI interface was developed at the Universitat Politècnica de València (UPV) and is registered under software registration application number S-063-2020, dated 23/09/2020. The experimental aspects of this study were supported by IRAMED (CIPROM/2022/38) from “Conselleria d’Innovació, Universitats, Ciència i Societat Digital” provided by “Generalitat Valenciana”, under the PROMETEO 2023 program. This study was funded in part by the intramural research program of the National Institutes of Health (NIH), National Cancer Institute, Division of Cancer Epidemiology and Genetics. The image segmentation and MC-GPU computer simulations were performed using the computational resources of the NIH Biowulf high-performance computing cluster (https://hpc.nih.gov).

References

  1. Ahrens James, Geveci Berk, Law Charles, Hansen C, andJohnson C. 2005. ‘36-paraview: An end-user tool for large-data visualization’, The visualization handbook, 717: 50038–1. [Google Scholar]
  2. Badal Andreu, and Badano Aldo. 2009. ‘Accelerating Monte Carlo simulations of photon transport in a voxelized geometry using a massively parallel graphics processing unit’, Medical physics, 36: 4878–80. [DOI] [PubMed] [Google Scholar]
  3. Badal Andreu, Sharma Diksha, Graff Christian G, Zeng Rongping, and Badano Aldo. 2021. ‘Mammography and breast tomosynthesis simulator for virtual clinical trials’, Computer Physics Communications, 261: 107779. [Google Scholar]
  4. Berrington de González A, Mahesh M, Kim KP, Bhargavan M, Lewis R, Mettler F, and Land C. 2009. ‘Projected cancer risks from computed tomographic scans performed in the United States in 2007’, Arch Intern Med, 169: 2071–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Booij R, Dijkshoorn ML, and van Straten M. 2017. ‘Efficacy of a dynamic collimator for overranging dose reduction in a second- and third-generation dual source CT scanner’, Eur Radiol, 27: 3618–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Boone JM, Hendee WR, McNitt-Gray MF, and Seltzer SE. 2012. ‘Radiation exposure from CT scans: how to close our knowledge gaps, monitor and safeguard exposure--proceedings and recommendations of the Radiation Dose Summit, sponsored by NIBIB, February 24–25, 2011’, Radiology, 265: 544–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Boone JM, and Seibert JA. 1997. ‘An accurate method for computer-generating tungsten anode x-ray spectra from 30 to 140 kV’, Med Phys, 24: 1661–70. [DOI] [PubMed] [Google Scholar]
  8. Brenner D, Elliston C, Hall E, and Berdon W. 2001. ‘Estimated risks of radiation-induced fatal cancer from pediatric CT’, AJR Am J Roentgenol, 176: 289–96. [DOI] [PubMed] [Google Scholar]
  9. Brenner DJ, and Hall EJ. 2007. ‘Computed tomography--an increasing source of radiation exposure’, N Engl J Med, 357: 2277–84. [DOI] [PubMed] [Google Scholar]
  10. Brody AS, Frush DP, Huda W, and Brent RL. 2007. ‘Radiation risk to children from computed tomography’, Pediatrics, 120: 677–82. [DOI] [PubMed] [Google Scholar]
  11. Christner JA, Zavaletta VA, Eusemann CD, Walz-Flannigan AI, and McCollough CH. 2010. ‘Dose reduction in helical CT: dynamically adjustable z-axis X-ray beam collimation’, AJR Am J Roentgenol, 194: W49–55. [DOI] [PubMed] [Google Scholar]
  12. Clark K, Vendt B, Smith K, Freymann J, Kirby J, Koppel P, Moore S, Phillips S, Maffitt D, Pringle M, Tarbox L, and Prior F. 2013. ‘The Cancer Imaging Archive (TCIA): maintaining and operating a public information repository’, J Digit Imaging, 26: 1045–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Council, European. 2014. “Council Directive 2013/59/Euratom: Basic Safety Standards for Protection Against the Dangers Arising from Exposure to Ionising Radiation and Repealing Directives 89/618/Euratom, 90/641/Euratom, 96/29/Euratom, 97/43/Euratom, and 2003/122/Euratom.” In.: Official Journal of the European Union L13, 57: 1–73. [Google Scholar]
  14. Ding A, Gao Y, Liu H, Caracappa PF, Long DJ, Bolch WE, Liu B, and Xu XG. 2015. ‘VirtualDose: a software for reporting organ doses from CT for adult and pediatric patients’, Phys Med Biol, 60: 5601–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Ding A, Gu J, Trofimov AV, and Xu XG. 2010. ‘Monte Carlo calculation of imaging doses from diagnostic multidetector CT and kilovoltage cone-beam CT as part of prostate cancer treatment plans’, Med Phys, 37: 6199–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Domino D 2024. ‘AuntMinnie 2010: Settlement Reached in Mad River Pediatric CT Radiation Case’, AuntMinnie.com, 26 August 2024. https://www.auntminnie.com/25-for-25/article/15682511/auntminnie-2010-settlement-reached-in-mad-river-pediatric-ct-radiation-case (accessed 26 March 2025). [Google Scholar]
  17. Fewell TR 1981. Handbook of computed tomography x-ray spectra. (US Department of Health and Human Services, Public Health Service, Food and Drug Administration, Bureau of Radiological Health, Rockville, MD.). [Google Scholar]
  18. Foucault A, Ancelet S, Dreuil S, Caër-Lorho S, Ducou Le Pointe H, Brisse H, Chateil JF, Lee C, Leuraud K, and Bernier MO. 2022. ‘Childhood cancer risks estimates following CT scans: an update of the French CT cohort study’, Eur Radiol, 32: 5491–98. [DOI] [PubMed] [Google Scholar]
  19. Hauptmann M, Byrnes G, Cardis E, Bernier MO, Blettner M, Dabin J, Engels H, Istad TS, Johansen C, Kaijser M, Kjaerheim K, Journy N, Meulepas JM, Moissonnier M, Ronckers C, Thierry-Chef I, Le Cornet L, Jahnen A, Pokora R, Bosch de Basea M, Figuerola J, Maccia C, Nordenskjold A, Harbron RW, Lee C, Simon SL, Berrington de Gonzalez A, Schüz J, and Kesminiene A. 2023. ‘Brain cancer after radiation exposure from CT examinations of children and young adults: results from the EPI-CT cohort study’, Lancet Oncol, 24: 45–53. [DOI] [PubMed] [Google Scholar]
  20. Hauptmann M, Daniels RD, Cardis E, Cullings HM, Kendall G, Laurier D, Linet MS, Little MP, Lubin JH, Preston DL, Richardson DB, Stram DO, Thierry-Chef I, Schubauer-Berigan MK, Gilbert ES, and Berrington de Gonzalez A. 2020. ‘Epidemiological Studies of Low-Dose Ionizing Radiation and Cancer: Summary Bias Assessment and Meta-Analysis’, J Natl Cancer Inst Monogr, 2020: 188–200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Huang WY, Muo CH, Lin CY, Jen YM, Yang MH, Lin JC, Sung FC, and Kao CH. 2014. ‘Paediatric head CT scan and subsequent risk of malignancy and benign brain tumour: a nation-wide population-based cohort study’, Br J Cancer, 110: 2354–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. IAEA. 2007. ‘Dosimetry in Diagnostic Radiology: An International Code of Practice. Technical Reports Series No. 457, Vienna, Austria.’. [Google Scholar]
  23. ICRP. 2002. ‘Basic anatomical and physiological data for use in radiological protection: reference values: ICRP Publication 89’, Annals of the ICRP, 32: 3–4. [PubMed] [Google Scholar]
  24. ImPACT. 2011. ‘ImPACT CT patient dosimetry calculator Version 1.0.4. London, United Kingdom.’. http://www.impactscan.org/ctdosimetry.htm (accessed 26 March 2025). [Google Scholar]
  25. Jadick G, Abadi E, Harrawood B, Sharma S, Segars WP, and Samei E. 2021. ‘A scanner-specific framework for simulating CT images with tube current modulation’, Phys Med Biol, 66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Jordan P, Adamson PM, Bhattbhatt V, Beriwal S, Shen S, Radermecker O, Bose S, Strain LS, Offe M, Fraley D, Principi S, Ye DH, Wang AS, van Heteren J, Vo NJ, and Schmidt TG. 2022. ‘Pediatric chest-abdomen-pelvis and abdomen-pelvis CT images with expert organ contours’, Med Phys, 49: 3523–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Journy NM, Lee C, Harbron RW, McHugh K, Pearce MS, and Berrington de González A. 2017. ‘Projected cancer risks potentially related to past, current, and future practices in paediatric CT in the United Kingdom, 1990–2020’, Br J Cancer, 116: 109–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Kitware, Inc. . 2025. ‘ParaView (Version 5.13.3) [Computer software]. https://www.paraview.org/’.
  29. Krille L, Dreger S, Schindel R, Albrecht T, Asmussen M, Barkhausen J, Berthold JD, Chavan A, Claussen C, Forsting M, Gianicolo EAL, Jablonka K, Jahnen A, Langer M, Laniado M, Lotz J, Mentzel HJ, Queißer-Wahrendorf A, Rompel O, Schlick I, Schneider K, Schumacher M, Seidenbusch M, Spix C, Spors B, Staatz G, Vogl T, Wagner J, Weisser G, Zeeb H, and Blettner M. 2015. ‘Risk of cancer incidence before the age of 15 years after exposure to ionising radiation from computed tomography: results from a German cohort study’, Radiat Environ Biophys, 54: 1–12. [DOI] [PubMed] [Google Scholar]
  30. Lee C, Kim KP, Bolch WE, Moroz BE, and Folio L. 2015. ‘NCICT: a computational solution to estimate organ doses for pediatric and adult patients undergoing CT scans’, J Radiol Prot, 35: 891–909. [DOI] [PubMed] [Google Scholar]
  31. Lee C, Liu J, Griffin K, Folio L, and Summers RM. 2020. ‘Adult patient-specific CT organ dose estimations using automated segmentations and Monte Carlo simulations’, Biomed Phys Eng Express, 6: 045016. [DOI] [PubMed] [Google Scholar]
  32. Lee C, Yeom YS, and Folio L. 2022. ‘CT organ dose calculator size adaptive for pediatric and adult patients’, Biomed Phys Eng Express, 8 – 6 (DOI 10.1088/2057-1976/ac9845). [DOI] [PubMed] [Google Scholar]
  33. Legislature, State of California. 2010. “Chapter 521, Sections 115111, 115112, and 115113 of the California Health and Safety Code. Radiation control: Health facilities and clinics—records. http://www.leginfo.ca.gov/pub/09-10/bill/sen/sb_1201-1250/sb_1237_bill_20100929_chaptered.html (accessed 26 March 2025).” In. [Google Scholar]
  34. Mathews JD, Forsythe AV, Brady Z, Butler MW, Goergen SK, Byrnes GB, Giles GG, Wallace AB, Anderson PR, Guiver TA, McGale P, Cain TM, Dowty JG, Bickerstaffe AC, and Darby SC. 2013. ‘Cancer risk in 680,000 people exposed to computed tomography scans in childhood or adolescence: data linkage study of 11 million Australians’, Bmj, 346: f2360. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Meulepas JM, Ronckers CM, Smets Amjb, Nievelstein RAJ, Gradowska P, Lee C, Jahnen A, van Straten M, de Wit MY, Zonnenberg B, Klein WM, Merks JH, Visser O, van Leeuwen FE, and Hauptmann M. 2019. ‘Radiation Exposure From Pediatric CT Scans and Subsequent Cancer Risk in the Netherlands’, J Natl Cancer Inst, 111: 256–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Morató Sergio, Lee Choonik, Griffin Keith T, Saha Monjoy, Lee Choonsik, and Mille Matthew M. 2024. ‘Realistic extension of partial-body pediatric CT for whole-body organ dose estimation in radiotherapy patients’, Radiation Physics and Chemistry, 226: 112194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. NEA. 2019. ‘PENELOPE-2018: A Code System for Monte Carlo Simulation of Electron and Photon Transport, OECD Publishing, Paris. DOI: 10.1787/32da5043-en’. [DOI] [Google Scholar]
  38. NVIDIA corporation. 2025. ‘CUDA C++ Programming Guide (Release 12.8).’. https://docs.nvidia.com/cuda/cuda-c-programming-guide/ (accessed 26 March 2025).
  39. Pearce MS, Salotti JA, Little MP, McHugh K, Lee C, Kim KP, Howe NL, Ronckers CM, Rajaraman P, Sir Craft AW, Parker L, and Berrington de González A. 2012. ‘Radiation exposure from CT scans in childhood and subsequent risk of leukaemia and brain tumours: a retrospective cohort study’, Lancet, 380: 499–505. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Peng Z, Fang X, Yan P, Shan H, Liu T, Pei X, Wang G, Liu B, Kalra MK, and Xu XG. 2020. ‘A method of rapid quantification of patient-specific organ doses for CT using deep-learning-based multi-organ segmentation and GPU-accelerated Monte Carlo dose computing’, Med Phys, 47: 2526–36. [DOI] [PubMed] [Google Scholar]
  41. Rehani MM, Yang K, Melick ER, Heil J, Šalát D, Sensakovic WF, and Liu B. 2020. ‘Patients undergoing recurrent CT scans: assessing the magnitude’, Eur Radiol, 30: 1828–36. [DOI] [PubMed] [Google Scholar]
  42. Salvat F, Fernández-Varea JM, and Sempau J 2006. ‘PENELOPE-2006: A code system for Monte Carlo simulation of electron and photon transport. NEA Report 6222.’. http://www.nea.fr/html/science/pubs/2006/nea6222-penelope.pdf. [Google Scholar]
  43. Sawyer LJ, Whittle SA, Matthews ES, Starritt HC, and Jupp TP. 2009. ‘Estimation of organ and effective doses resulting from cone beam CT imaging for radiotherapy treatment planning’, Br J Radiol, 82: 577–84. [DOI] [PubMed] [Google Scholar]
  44. Services, Texas Department of State Health. 2013. “Texas Administrative Code, Title 25, Part 1, Chapter 289.227: Use of Radiation Machines in the Healing Arts. Effective 1 May 2013. https://www.dshs.texas.gov (accessed 26 March 2025).” In. [Google Scholar]
  45. Shirasaka T, Funama Y, Hayashi M, Awamoto S, Kondo M, Nakamura Y, Hatakenaka M, and Honda H. 2012. ‘Reduction of the unnecessary dose from the over-range area with a spiral dynamic z-collimator: comparison of beam pitch and detector coverage with 128-detector row CT’, Radiol Phys Technol, 5: 53–8. [DOI] [PubMed] [Google Scholar]
  46. Siemens Healthineers, https://bps.healthcare.siemens-healthineers.com/booneweb/index.html. 2024. ‘X-ray Spectra Simulation Tool’.
  47. Simon SL, Kleinerman RA, Ron E, and Bouville A. 2006. ‘Uses of dosimetry in radiation epidemiology’, Radiat Res, 166: 125–7. [DOI] [PubMed] [Google Scholar]
  48. Smoll NR, Brady Z, Scurrah KJ, Lee C, Berrington de González A, and Mathews JD. 2023. ‘Computed tomography scan radiation and brain cancer incidence’, Neuro Oncol, 25: 1368–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Stamm G, and Nagel HD. 2002. “CT-Expo - ein neuartiges Programm zur Dosisevaluierung in der CT [CT-expo--a novel program for dose evaluation in CT].” In RöFo, 1570–76. 10.1055/s-2002-35937. [Article in German]. [DOI] [PubMed] [Google Scholar]
  50. Thierry-Chef I, Ferro G, Le Cornet L, Dabin J, Istad TS, Jahnen A, Lee C, Maccia C, Malchair F, Olerud HM, Harbron RW, Figuerola J, Hermen J, Moissonnier M, Bernier MO, Bosch de Basea MB, Byrnes G, Cardis E, Hauptmann M, Journy N, Kesminiene A, Meulepas JM, Pokora R, and Simon SL. 2021. ‘Dose estimation for the european epidemiological etudy on pediatric computed tomography (EPI-CT)’, Radiat Res, 196: 74–99. [DOI] [PubMed] [Google Scholar]
  51. U.S. Food and Drug Administration. 2012. ‘MCGPU: GPU-accelerated Monte Carlo X-ray Imaging Simulator’, (RST24MD09.01). https://cdrh-rst.fda.gov/mcgpu-gpu-accelerated-monte-carlo-x-ray-imaging-simulator (accessed 26 March 2025).
  52. Wasserthal J, Breit H-C, Meyer MT, Pradella M, Hinck D, Sauter AW, Heye T, Boll D, Cyriac J, Yang S, Bach M, Segeroth M,. 2023. ‘TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images. ‘, Radiology: Artificial Intelligence. 10.1148/ryai.230024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Wilson C, Adams GG, Patel P, Windham K, Ennis C, and Caffrey E. 2024. ‘A Review of Recent Low-dose Research and Recommendations for Moving Forward’, Health Phys, 126: 386–96. [DOI] [PubMed] [Google Scholar]
  54. Zarembo A 2009. ‘Hospital error leads to 206 radiation overdoses’, The Los Angeles Times, 13 October 2009. https://www.latimes.com/archives/la-xpm-2009-oct-13-me-cedars13-story.html (accessed 26 March 2025). [Google Scholar]

RESOURCES