Abstract
A full set of ICRP mesh-type reference computational phantoms (MRCPs), including adult, paediatric, and pregnant-female phantoms, is under development to produce dose coefficients for the next General Recommendations. The MRCPs are high-fidelity human models for radiation protection dosimetry, including all the source and target regions for effective dose calculations. The phantoms include micron-thick stem cell layers in the respiratory and alimentary tract organs, urinary bladder, and skin. The phantoms also include very detailed eye models and skeletal models. The MRCPs are developed in the fourth-generation phantom geometry, i.e. tetrahedron mesh geometry. This geometry, classified as an unstructured volume mesh, provides several key advantages. Firstly, it allows the phantoms to be directly incorporated into various Monte Carlo codes, such as Geant4, MCNP6, PHITS, and EGSnrc, without the need for voxelisation. This preserves the original high fidelity of the mesh phantoms during dose calculations, ensuring the most accurate results. The tetrahedron mesh geometry is volume representation, not surface or boundary representation, and provides the capability of suborgan/structure density variation modelling using the tetrahedra. A notable aspect of the mesh technology is its flexibility, allowing the phantoms to be easily adjusted for different body shapes and postures as required. In this article, the deformability of the MRCPs will be highlighted by reporting that the MRCPs were deformed into a library of 212 phantoms for adults and 637 phantoms for adolescents and children to represent different body sizes and shapes. It will also be reported that the phantoms were deformed into several different postures.
VOXEL-TYPE REFERENCE COMPUTATIONAL PHANTOMS FOR 2007 GENERAL RECOMMENDATIONS
Following the 2007 General Recommendations for the radiological protection system in International Commission on Radiological Protection (ICRP) Publication 103 (ICRP, 2007), ICRP first officially adopted the adult male and female reference computational phantoms in ICRP Publication 110 (ICRP, 2009). About 10 years later, the age-dependent paediatric male and female reference computational phantoms (0, 1, 5, 10, and 15 years) were released in ICRP Publication 143 (ICRP, 2020a). The P110 and P143 reference phantoms have been extensively used to produce reference dose coefficients (RDCs) for the current General Recommendations (ICRP, 2010, 2015, 2016a,b, 2017, 2019, 2020b). These reference phantoms were developed in voxel geometry (so-called voxel-type reference computational phantoms, VRCPs) and faithfully adjusted to match the reference values provided in ICRP Publication 89 (ICRP, 2002). The VRCPs, which were constructed based on computed tomography (CT) images of real persons, significantly improve both anatomical and dosimetrical features against the older-fashioned stylised phantoms (also called mathematical phantoms as modelled using quadric equations such as sphere and ellipsoid) which were conventionally used for calculating the previous RDCs (ICRP, 1979, 1988, 1990, 1991a,b, 1993, 1994a,b, 1995a,b,c, 1996, 1998, 2006) prior to the current General Recommendations. The adult and paediatric VRCPs are shown in Fig. 1.
Voxel-type reference computational phantoms (VRCPs) for adults in ICRP Publication 110 (ICRP, 2009) and children in ICRP Publication 143 (ICRP, 2020a).
Nevertheless, the VRCPs still have anatomical and dosimetrical limitations because the sizes of the voxels composing the VRCPs are limited on the order of millimetres (e.g. adult male 2.137 × 2.137 × 8.0 mm3 and adult female 1.775 × 1.775 × 4.8 mm3) and thus cannot explicitly represent thinner/smaller structures, especially those required in dose calculations (ICRP, 2009). For example, the skin of the VRCPs, which is defined by a single voxel layer only in transverse directions, is discontinuous between the individual transverse slices. Through the discontinuous parts of the skin, radiations externally incident to the body can reach internal organs/tissues superficially located (such as breasts, testes, and salivary glands), directly without penetrating the skin layer. Owing to this limitation, ICRP Publication 116 (ICRP, 2010) observed that breast doses calculated using the VRCPs were significantly overestimated for low-energy protons (<10 MeV) in ISO (isotropic) geometry. In addition, not only the skin but also other organs of walled structures (e.g. stomach and bladder) are discontinuous between the transverse slices (Yeom et al., 2013). Moreover, the skin basal cell layers of the epidermis and hair follicles, which are considered as the relevant target cells at radiogenic risk, are assumed to be a depth of 50 to 100 μm (15 years and adults) and 40 to 100 μm (0, 1, 5, and 10 years) below the skin surface (ICRP, 1977, 2010, 2015, in press), which cannot be defined in the VRCPs with the millimetre-scale voxel sizes of the phantoms. Similarly, the micron-scale radiosensitive target and source regions in the respiratory tract system (ICRP, 1994a) and those in the alimentary tract system (ICRP, 2006) cannot be defined in the VRCPs. In this circumstance, the RDCs for the current General Recommendations have been produced not only using the VRCPs, but also using ∼50 mathematical models to calculate absorbed doses to the micron-scale target regions (skin, eye lens, and respiratory and alimentary tract organs/tissues) that cannot be precisely calculated using the VRCPs (ICRP, 2010, 2015, 2016a,b, 2017).
For more than 60 years, a computational approach for dose assessments using computational phantoms coupled with Monte Carlo (MC) radiation transport simulations has played a central role in various fields including radiation protection, medical imaging, and radiotherapy. As shown in Fig. 2, a technology in the development of computational phantoms has evolved significantly to improve the representation of the human anatomy as accurate/precise as possible for better dose assessment.
Advances in computational phantom technology from the 1950s to present.
At the beginning about 1950s, the shape of the human body was simplified to a box, cylinder, sphere, and so on. Until 2020, the International Commission on Radiation Units and Measurement (ICRU) sphere recommended the use of the ICRU sphere to derive operational quantities for external exposures (ICRU, 1985, 1988, 1993, 2020). Since 1960s, stylised phantoms (also called mathematical phantoms), classified as the first generation of computational phantoms, were developed to improve dose assessments of individual organs and tissues, although only the most general description of the position and geometry of each organ/tissue was described using simple quadratic equations and combining them with Boolean operations (Xu, 2014; Kainz et al., 2018). Stylised phantoms such as MIRD-5 (Snyder, 1978) and ADAM/EVA (Kramer et al., 1982) were used to calculate the RDCs for external and internal exposures (ICRP, 1979, 1988, 1990, 1991a,b, 1993, 1994a,b, 1995a,b,c, 1996, 1998, 2006) prior to the current General Recommendations.
Voxel phantoms: second generation
Since 1980s with the advent of medical imaging techniques such as CT and MRI, voxel phantoms (or tomographic phantoms), classified as the second generation of computational phantoms, were developed based on tomographic images that can visualise the internal structures of the real human body in 3D (Xu, 2014; Kainz et al., 2018). The earliest effort on voxel phantoms is reported by a group at Vanderbilt University, USA (Gibbs and Pujol, 1982; Gibbs et al., 1984, 1987), in which 2D x-ray images were used to create a realistic patient model and calculate doses to the patients during dental radiologic procedures. In the late 1980s, a group at GSF, Germany, first used 3D CT imaging on healthy volunteers and patients to create an 8-week-old voxel phantom (BABY) and a 7-year-old voxel phantom (CHILD) (Zankl et al., 1988). This group subsequently developed 10 phantoms more: DONNA, FRANK, HELGA, IRENE, GOLEM, GODWIN, VISIBLE HUMAN, LAURA, KLARA, and KATJA (Smith et al., 2000; Petoussi-Henss et al., 2002; Zankl et al., 2002, 2005; Fill et al., 2004; Becker et al., 2007). Later, the GOLEM and LAURA phantoms were significantly revised to create the REX and REGINA phantoms, which were finally adopted by the Commission as the adult male and female VRCPs released in ICRP Publication 110 (ICRP, 2009). Since late 1990s as computers and tomographic imaging techniques were dramatically advanced and became more easily available, the development of voxel phantoms has been increasingly popular for over three decades (Xu, 2014; Kainz et al., 2018).
While achieving significant improvement in the human anatomy over the first-generation stylised phantoms, the second-generation voxel phantoms carry inherent limitations. Firstly, it proves to be challenging when constructing models with extremely thin layers or minute structures. Specifically, to depict a thin layer in a voxel model, the size of the voxels needs to be smaller than the thickness of the layer. In practical terms, especially when dealing with human bodies where the thickness of radiosensitive target layers and source layers are few to tens of microns, it is impractical to define these thin layers using voxel geometry, as with the limitations of the VRCPs previously discussed in Section 1. Moreover, voxel geometry is not deformable, implying that it only allows for the development of a ‘rigid’ model. It also represents jagged stair-stepped surfaces.
Non-uniform rational B-spline/polygon-mesh phantoms: third generation
Since the 2000s, computational phantoms based on non-uniform rational B-spline (NURBS) and/or polygon-mesh (PM) surfaces have been developed, which are classified as the third generation of computational phantoms (Xu, 2014; Kainz et al., 2018). The first effort on the anatomical modelling using the NURBS-based techniques was published by Segars (2001) about the development of the NCAT (NURBS-based Cardiac Torso) phantom. These surface phantoms describe the shape of the body and organs/tissues by using NURBS and/or PM surfaces, the advanced computational geometries used in 3D computer graphics applications such as computer-aided design (CAD). In principle, NURBS/PM surfaces, which are much more flexible and deformable than voxel geometry, have a great potential to address the limitations of the voxel phantoms. For example, the use of the surface geometry can represent the smooth boundary surface of organs and tissues. It is also possible to define very thin structures such as micron-scale radiosensitive target and source regions in the respiratory and alimentary tract systems (Yeom et al., 2013). Moreover, due to the high flexibility and deformability of the surface geometry over the voxel geometry, it is much easier to deform existing NURBS/PM phantoms into different body sizes/postures or to change the position and shape of organs/tissues within the phantoms.
Although providing significant benefits over the stylised and voxel phantoms, NURBS/PM surface phantoms still suffer from one technically critical issue: a lack of compatibility with MC particle transport codes for dose calculations. None of the existing MC particle transport codes can handle NURBS geometry as implementing Monte Carlo particle transport algorithms for NURBS geometry requires highly complex and time-consuming algebraic calculations and thus is technically very challenging (Kainz et al., 2018). PM geometry can be handled by a few of the MC codes such as Geant4 (Allison et al., 2016), for which the PM geometry must be watertight and have perfect meshes without any abnormal facets. In principle, however, transporting particles in PM geometry is significantly inefficient as compared to voxel geometry. This inefficiency for PM geometry dramatically increases as the number of facets used for PM geometry increases. A research team at Hanyang University, Korea, developed and successfully implemented a PM phantom (called PSRK-Man) in the Geant4 code but found that the PM phantom was significantly slower (e.g. 70–150 times for photons) in MC dose calculations than its counterpart voxel phantom (Kim et al., 2011). Due to the low compatibility of NURBS/PM geometry in MC codes, most of the existing NURBS/PM phantoms (Lee et al., 2007a,b, 2008, 2010; Xu et al., 2007; Zhang et al., 2008, 2009; Jeong et al., 2008; Johnson et al., 2009; Bolch et al., 2010; Cassola et al., 2010; Hurtado et al., 2012; Gardumi et al., 2013; Geyer et al., 2014) have been converted to voxel models, which is the well-known voxelisation, and then the converted voxel models are used in Monte Carlo dose calculations (Xu, 2014; Kainz et al., 2018). Consequently, this voxelisation process results in the same limitations of voxel geometry as the final geometry used in the MC codes is the voxel geometry, not NURBS/PM geometry. Note that the voxelised version of the paediatric NURBS/PM phantom series developed by the US collaboration between the University of Florida (UF) and the National Cancer Institute (NCI) was modified to create the paediatric VRCPs released in ICRP Publication 143 (ICRP, 2020a).
Tetrahedral mesh phantoms: fourth generation
The limitations of the NURBS/PM phantoms have been addressed by the dawning of a new generation of computational phantoms based on tetrahedral mesh (TM) geometry. TM geometry is achieved by filling the PM model with numerous tetrahedra through a process known as tetrahedralisation. This process is conceptually akin to voxelisation, but the main distinction is that the PM model is filled with tetrahedra instead of being voxelised. The resulting TM model does not distort the geometry of the original PM model, taking the benefits of PM geometry into account during MC dose calculations, but dramatically accelerates the MC computation speed, even at a comparable level of voxel geometry. This consequently makes the use of mesh phantoms in MC dose calculations practically reasonable. The research team at Hanyang University, Korea, first demonstrated that the TM version of the PSRK-Man phantom implemented in the Geant4 code using the G4Tet class resulted in a 2–3 order of magnitude increase in computational speed depending on particles and energies against the original PM phantom implemented in the Geant4 code using the G4TessellatedSolid class (Yeom et al., 2014). This enhancement in computational speed is because the large number of facets in PM geometry to be checked by the ray-tracing algorithm that are computationally expensive is reduced to just four facets in TM geometry. Furthermore, the utilisation of TM geometry allows for the modelling of suborgan/tissue density variation within an organ/tissue using the tetrahedra that fill the organ model. It should be importantly noted that NURBS/PM geometry is surface mesh geometry for boundary representation of the organ/tissue and therefore cannot reflect this suborgan/tissue density variation. Instead, the organ/tissue must be assumed homogenous in NURBS/PM geometry, which is not always reliable and can often pose significant limitations in dose calculations.
MESH-TYPE REFERENCE COMPUTATIONAL PHANTOMS FOR NEXT GENERAL RECOMMENDATIONS
Acknowledging the limitations of the current VRCPs and taking the cutting-edge phantom technology based on TM geometry into account, the Commission launched the Task Group 103 (chaired by Prof. Chan Hyeong Kim at Hanyang University, Seoul) with the aim of the development of new reference computational phantoms based on TM geometry, so-called mesh-type reference computational phantoms (MRCPs). This task group first completed the development of the adult male and female MRCPs released to the public via ICRP Publication 145 (ICRP, 2020c). The ICRP-145 MRCPs were constructed by converting the ICRP-110 VRCPs to high-quality tetrahedral mesh, faithfully maintaining the overall anatomy of the original voxel phantoms, and at the same time by overcoming the limitations of the voxel phantoms due mainly to the limited voxel resolutions. Following the adult MRCPs, the paediatric male and female MRCPs with the five ages (0, 1, 5, 10, and 15 years) were constructed from the ICRP-143 paediatric VRCPs (ICRP, 2020a) and were released via ICRP Publication 156 (ICRP, 2024). The adult and paediatric MRCPs are shown in Fig. 3. The highlighted improvement of the MRCPs over the VRCPs is that the micron-scale radiosensitive target and source regions of the skin, lens of the eyes, urinary bladder, and alimentary and respiratory tract systems are explicitly defined in the MRCPs. It should be noted that the MRCPs in the high-quality/fidelity tetrahedral mesh format can be directly implemented into MC codes such as Geant4 (Allison et al., 2016), PHITS (Sato et al., 2013, 2018), and MCNP6 (Goorley et al., 2013, 2016), without any voxelisation process, so that the micron-scale structures defined in the mesh phantoms can be used in MC dose calculations. As a result, the use of the supplementary mathematical models to address the limitations of the VRCPs in the current ICRP dosimetry system is no longer necessary for the future ICRP dosimetry system. Very recently, the Task Group 103 has completed the development of the pregnant-female MRCPs with fetal ages (8, 10, 15, 20, 25, 30, 35, and 38 weeks) as shown in Fig. 4 (Shin et al., 2024, 2025). In contrast to the adult and paediatric MRCPs, which were converted from the adult and paediatric VRCPs, the pregnant-female MRCPs are the first ICRP reference phantoms for the mothers and foetuses, which are already being used in the RDC calculations for the current General Recommendations and will also be used for the next General Recommendations.
Mesh-type reference computational phantoms (MRCPs) for adults in ICRP Publication 145 (ICRP, 2020c) and children in ICRP Publication 156 (ICRP, 2024).
Mesh-type reference computational phantoms (MRCPs) for pregnant females and foetuses with eight fetal ages (8, 10, 15, 20, 25, 30, 35, and 38 weeks). Only male foetuses are shown for illustration, with the corresponding zoom scale indicated.
Taking the high deformability of the mesh geometry as mentioned in Section 2, the MRCPs can be flexibly transformed if necessary to various non-reference phantoms, e.g. in different body sizes and/or postures, which could be beneficial especially for dosimetry applications (e.g. accidental/emergency and or medical exposures) requiring individual-specific dosimetry rather than reference-based dosimetry. For example, radiation/nuclear accidents, although not frequently happening, could result in high radiation doses to individuals, causing serious injuries and even death. For estimation of such high doses to individuals for accidental/emergency exposure situations, the concept of the Reference Person in dosimetry (i.e. using the reference phantoms) may not be sufficient, and the individual-specific dosimetry is more desirable to implement better decision making especially for the individuals involved in the accident whose body size or posture is largely different from that of the reference phantoms. For individual dosimetry, however, the construction of individual-specific phantoms is challenging as the individual anatomy data are usually unavailable, and, even if available, it is highly impractical especially for a large-scale accident since the phantom construction is indeed a time-consuming and labour-intensive work. Given this circumstance, the dose to the individuals could be better approximated using a non-reference phantom transformed from the MRCPs whose body size or posture is close to that of the actual person.
To demonstrate this benefit of the MRCPs, ICRP Publication 145 (ICRP, 2020c) already introduced a set of posture-dependent phantoms transformed from the adult MRCPs into five non-standing postures (walking, sitting, bending, kneeling, and squatting) (Yeom et al., 2019) and also a set of body size-dependent phantoms in 10th, 50th, and 90th standing heights and body weights of Caucasian populations (Lee et al., 2019) as shown in Fig. 5. The 10th and 90th percentile phantoms were then used for the calculation of dose coefficients for industrial radiography accidents presented in Annex J of ICRP Publication 145 (ICRP, 2020c). In addition, the Task Group 113 (Reference Organ and Effective Dose Coefficients for Common Diagnostic X-ray Imaging Examinations) is planned to use the 10th and 90th percentile phantoms for the comparison with the reference phantoms to investigate the impact of the body size on organ dose calculations for common diagnostic x-ray imaging examinations (radiography, CT, and fluoroscopy). For dosimetry more precisely reflecting the body size of individuals, the percentile-dependent phantoms were expanded to a body size-dependent phantom library transformed from the adult MRCPs, comprising a total of 212 adult male and female phantoms (108 males and 104 females) in different standing heights and body weights (Choi et al., 2020). Very recently, the phantom library has been further expanded by transforming the paediatric MRCPs to a total of 637 body size-dependent paediatric phantoms (356 males and 281 females) (Kim et al., 2024). Fig. 6, as an example, shows the adult and paediatric male phantoms in the library. To improve the practical usability of the MRCPs and the extended phantom library, the Hanyang University Research Engineering Laboratory (HUREL), led by Prof. Chan Hyeong Kim (Chair of the Task Group 103 and Committee 2), is currently developing an independent GUI based MC programme (titled McSEE, Monte Carlo Simulation Programme for External Exposures) so that users can easily conduct MC dose calculations using the mesh phantoms even without any expertise of the phantoms and the MC codes. The McSEE program is planned for use in the calculation of RDCs for external exposures in emergency situations by the Task Group 112 (Emergency Dosimetry).
Posture-dependent phantoms (right upper) and body size-dependent phantoms (right lower) transformed from adult male mesh-type reference computational phantoms (left).
Body size-dependent adult and paediatric male phantoms transformed from the adult and paediatric male mesh-type reference computational phantoms: adults (upper) and children (lower).
Acknowledging the anatomical and dosimetrical limitations of the current VRCPs due to the limited voxel resolutions and adopting the cutting-edge phantom technology based on tetrahedral mesh, the Commission formulated the Task Group 103 under Committee 2 with the aim of the development of the MRCPs as the next generation of reference computational phantoms for use in the computation of reference dose coefficients for external and internal exposures for the next General Recommendations. The MRCPs have been successfully developed, including the adult male and female reference phantoms released via ICRP Publication 145 (ICRP, 2020c), the paediatric male and female reference phantoms with the five ages (0, 1, 5, 10, and 15 years) released via ICRP Publication 156 (ICRP, 2024), and even the first pregnant-female reference phantoms (mother and foetus) with the eight fetal ages (8, 10, 15, 20, 25, 30, 35, and 38 weeks) of which report for the release is currently under preparation by the Task Group 103 and expected to be published early 2026. While overcoming the anatomical and dosimetrical limitations of the current VRCPs due mainly to the limited voxel resolutions, taking the high deformability of mesh geometry into account, the MRCPs were deformed into a library of 212 phantoms for adults (Choi et al., 2020) and 637 phantoms for adolescents and children (Kim et al., 2024) to represent different body sizes. In addition, the adult MRCPs were deformed into several different postures (Yeom et al., 2019). Such non-reference phantoms transformed from the MRCPs can be beneficially used in various applications where individual-specific dosimetry is more desirable such as dose reconstructions for accident/emergency exposure scenarios.