Chapters 1-6

Back to Article Outline

1. Introduction 

(1) Phantoms simulating the human body or parts thereof play a central role in radiation dosimetry, either as physical phantoms for practical measurements or as computational models (ICRU, 1992).

(2) Computational phantoms (or mathematical models) of the human body are used to evaluate the energy deposition in organs resulting from internal and external radiation exposures. Traditionally, these phantoms have been based upon mathematical expressions representing planes, and cylindrical, conical, elliptical, and spherical surfaces that describe the shape and position of idealised body organs. This type of phantom was developed at the Oak Ridge National Laboratory (Fisher and Snyder, 1967, Fisher and Snyder, 1968, Snyder et al., 1969, Snyder et al., 1978, Cristy, 1980, Cristy and Eckerman, 1987) for the Medical Internal Radiation Dose (MIRD) Committee of the Society of Nuclear Medicine. From the original adult MIRD phantom, several paediatric phantoms were derived to represent infants and children of various ages (Cristy, 1980). For these models, the organ masses and volumes were in accordance with the ICRP data of former Reference Man (ICRP, 1975). As an improvement to these hermaphrodite models, separate male and female adult mathematical models called ‘Adam’ and ‘Eva’ were introduced by Kramer et al. (1982). Subsequently, four models representing the adult female, non-pregnant and at three stages of pregnancy, were developed by Stabin et al. (1995).

(3) As an extension and improvement to these earlier models, various groups have developed a new type of anatomical phantom during the last two decades, so-called ‘tomographic’ or ‘voxel’ models. Voxel phantoms are anatomical models based on computed tomography, magnetic resonance, or other images obtained from high-resolution scans of a single individual and, thus, offer a more realistic replication of human anatomy. They consist of a large number of volume elements (voxels) and are the most detailed representation of human anatomy at the present time (Zankl et al., 1988, Zubal et al., 1994, Zubal et al., 1996, Dimbylow, 1996, Caon et al., 1999, Xu et al., 2000, Zankl and Wittmann, 2001, Petoussi-Henss et al., 2002, Zaidi and Xu, 2007). Voxel phantoms can be used for a wide spectrum of applications where the simulation of human anatomy is required. Until now, they have been used for the simulation of exposures due to ionising or electromagnetic radiation with the emphasis on radiation protection. However, being derived from a specific individual, these models do not represent the average Caucasian man or woman as defined by Publication 23 (ICRP, 1975) and Publication 89 (ICRP, 2002).

(4) Various authors have shown that the organ shapes of the MIRD-type phantoms present an over-simplification, having an influence on the energy distribution, which – for some cases – deviates systematically from that calculated for voxel models (Jones, 1996, Smith et al., 2000, Chao et al., 2001, Kinase et al., 2003, Zankl et al., 2003, Kramer et al., 2004). For external radiation, the parameters influencing the organ doses are mainly depth of the organ below the body surface, exterior shape of the trunk, and trunk diameter relative to the incoming radiation beam (Jones, 1997, Chao et al., 2001, Zankl et al., 2002, Kramer et al., 2004, Schlattl et al., 2007). For internal dosimetry, the influencing parameters are the relative position of source and target organs (for penetrating radiations subject to so-called cross-fire), and organ mass (for non-penetrating radiations) (Jones, 1998, Smith et al., 2000, Chao and Xu, 2001, Zankl et al., 2003).

(5) Following the 2007 Recommendations (ICRP, 2007), the purpose of this document is to introduce the official computational models representing the adult Reference Male and Reference Female. These phantoms will be used by ICRP in establishing radiation protection guidance, e.g. effective dose coefficients (see Fig. 1.1) and other secondary dosimetric quantities.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1.1. 

  Schematic for calculation of effective dose using sex-specific reference phantoms.

(6) Similar activities in deriving standard phantoms were performed by Dimbylow (Dimbylow, 1996), who adjusted a segmented magnetic resonance imaging whole-body data set to yield a phantom (NORMAN) with body mass and height as well as organ masses in agreement with the reference values from Publication 23 (ICRP, 1975). Recently, more efforts have been made to build standard human models (Zankl et al., 2005, Kramer et al., 2006), where the individual organ masses and densities also conform to the values of Publication 70 (ICRP, 1995) and Publication 89 (ICRP, 2002). The phantoms presented in this document have been developed by the Task Group on Dose Calculations (DOCAL) of ICRP Committee 2 in collaboration with the Helmholtz Zentrum München – German Research Centre for Environmental Health (formerly: GSF – National Research Centre for Environment and Health), and the International Commission on Radiation Units and Measurements (ICRU). The following chapters describe the specifications that these phantoms had to meet, their construction, their main characteristics, and their limitations. Finally, an overview of their potential applications and intended use is given.

Back to Article Outline

2. Specifications of the computational phantoms 

(7) The voxel phantoms for calculations of energy deposition in body organs and tissues (so-called ‘target regions’) following the 2007 Recommendations (ICRP, 2007) should accommodate all organs and tissues that are relevant to the assessment of human exposure to ionising radiation for radiation protection purposes. These target regions are: active (‘red’) bone marrow, adrenals, brain, breast, colon, endosteal tissue (formerly called ‘bone surfaces’), extrathoracic (ET) airways, eye (lens), gall bladder, heart, kidneys, liver, lungs, lymphatic nodes, muscle, oesophagus, oral mucosa, ovaries, pancreas, prostate, salivary glands, skin, small intestine, spleen, stomach, testes, thymus, thyroid, urinary bladder, and uterus. Furthermore, additional target regions have been identified in the Human Respiratory Tract Model (ICRP, 1994) and Human Alimentary Tract Model (ICRP, 2006). These target regions include: alveolar-interstitium, basal cells of anterior and posterior nasal passages and pharynx, basal cells of bronchi, lymph nodes of ET and thoracic region, secretory cells of bronchi and bronchioles, and tongue and tonsils.¹

(8) When radioactive material is incorporated into the body, those organs, tissues, and body regions where radionuclides reside or pass through become source regions that irradiate other (target) regions. Many regions are both source and target regions. Additional source regions are located in the alimentary and respiratory tracts, as well as in the skeleton. Certain individual anatomical regions have to be considered differently depending on the rate with which the material passes through or is cleared from them. These additional source regions include: oral cavity, teeth surfaces, teeth volumes, oesophagus (fast, slow), stomach content, small intestine content, right colon (content, wall), left colon (content, wall), rectosigmoid colon (content, wall), gall bladder content, urinary bladder content, nasal passages (anterior and posterior surfaces), pharynx, sequestered ET₂ region, bronchi (fast, slow, bound, sequestered), bronchi, bronchioles (fast, slow, bound, sequestered), blood vessels (head, trunk, legs, and arms), cortical bone (surface, volume), trabecular bone (surface, volume), and inactive (‘yellow’) bone marrow.

(9) Due to the limited resolution (in the range of millimetres) of the tomographic data used to construct the voxel phantoms, and the very small dimensions of some source and target tissues (tens of micrometres), not all tissues could be segmented directly. Therefore, for some source and target tissues, ‘surrogate’ regions had to be found or correction factors have to be applied to the calculated doses. These limitations of the phantoms are discussed in more detail in Section 5.4.

(10) To enable the application of the reference computational phantoms for dosimetry in nuclear medicine by ICRP Committee 3 and others, further features were introduced: (1) left and right components of organ pairs were identified separately; and (2) in the kidneys, the renal cortex, medulla, and pelvis were identified separately.

(11) A list of all organs, tissues, and regions that were defined and have been assigned an individual organ identification number is given in Annex A. Annex B is a list of different tissue types of which the organs and tissues consist, and their elemental compositions. Annex C presents a list of all source organs and regions, together with the organ identification numbers by which they are represented. Annex D gives this information for the target organs and tissues.

(12) Since the tomographic data used to create the phantoms were acquired while the individuals were in a supine posture, it is obvious that the anatomy of the resulting voxel models also corresponds to this posture. That means that – due to the f act that the forces of gravity act differently in standing and lying postures – the abdomen is flatter than in a standing person, the abdominal organs are shifted upwards towards the chest, and the lungs are compressed. Furthermore, the curvature of the spine is slightly different from that seen in a standing person. Correcting for these effects would mean the necessity for extensive modifications in organ positioning in the body. On the other hand, although this effect is qualitatively obvious, there is little quantitative information on the amount of positional changes of individual organs, since no comparable examinations of the same person in different positions are usually available. A study on the effects of posture on organ doses per unit activity intake for eight radionuclides with different biokinetic behaviour revealed only moderate dose differences for the main organs (i.e. those contributing to the effective dose with the highest tissue weighting factors) for a person in upright and supine postures (Sato et al., 2007, Sato and Endo, 2008). A similar study for external photon exposure showed agreement of the organ dose conversion coefficients for a supine person and an upright person within 2–20% for photon energies above 50keV (Sato et al., 2008). Thus, it is concluded that the dosimetric impact of the person’s position is limited, and the organ position of the voxel models in the supine position is acceptable for the applications intended.

(13) Being derived from specific individuals, most models to date do not represent the average Caucasian man or woman as defined by Publication 23 (ICRP, 1975) and Publication 89 (ICRP, 2002). The method and principles applied to construct computational phantoms representing the adult Reference Male and Reference Female (ICRP, 2002, ICRP, 2007) and the data that were used during this process are outlined in Fig. 2.1 and will be described in the following chapters.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 2.1. 

  Schematic for constructing computational phantoms representing the adult Reference Male and Reference Female.

Back to Article Outline

3. Selection and segmentation of tomographic data 

(14) In order to construct voxel models representing the adult Reference Male and Reference Female (ICRP, 2002, ICRP, 2007), appropriate tomographic data sets were used as starting points, i.e. individuals with external dimensions close to the reference values, so that the required modifications remained moderate, and the problem of distorting the anatomical relations was minimal (Step 1 of Fig. 2.1).

3.1. Tomographic data for the male 

(15) A whole-body clinical computed tomography image set of a 38-year-old individual with height 176cm and mass slightly below 70kg (Reference Male: 176cm and 73kg) was selected for construction of the male reference computational phantom. The individual, who suffered from leukaemia and had to undergo whole-body irradiation, did not have obvious signs of illness that would appear in the image data. The person was lying supine with the arms parallel alongside the body. The data set consisted of 220 slices of 256 x 256 pixels. The original voxel size was 8mm in height with an in-plane resolution of 2.08mm, resulting in a voxel volume of 34.6mm³.

(16) In total, 122 individual objects were segmented (67 of these being bones or bone groups), including many – but not all – of the organs and tissues later identified in the ICRP characterisation of the anatomical reference data (ICRP, 2002). Segmentation refers to the process by which individual pixels in an image slice are given organ identification numbers instead of their original Hounsfield numbers (image pixel intensity). As the image slice refers to a certain anatomical thickness, each pixel also defines a volume element or voxel. The collection of all voxels with the same identification number defines a certain organ or tissue. The whole body is, thus, represented by a three-dimensional array of voxels that is arranged in columns, rows, and slices. The identification number of the organ to which the voxel belongs is stored at each array position (Step 2 of Fig. 2.1).

(17) One of the tissues that could not be segmented from image data of the given resolution was the bone marrow, which is contained in small cavities in the trabecular bone that are much smaller than the voxel size (in the order of a few hundred micrometres) (ICRP, 1995). Therefore, no attempt has been made to identify sub-structures of each individually segmented bone or group of bones, with the exception of cortical bone and trabecular spongiosa.

(18) Due to the limited resolution of the image data (slice thickness 8mm), it was difficult to identify small structures, such as blood vessels which are much smaller than the large main vessels in the trunk. Therefore, only a rather small proportion of the blood pool could be segmented. Furthermore, since no cartilage had been considered in the original segmented model, and due to the limited dosimetric importance of this tissue, only limited effort towards its supplementary segmentation was exerted.

(19) For the segmentation, commercial equipment dedicated to image processing purposes was used (Kontron Bildanalyse, MIPRON, Eching, Germany). The software tools used were grey value thresholding and morphological operations on binary images. The result of segmentation was the voxel phantom known as ‘Golem’ (Zankl and Wittmann, 2001).

3.2. Tomographic data for the female 

(20) The female reference computational phantom was based on the computed tomography scan of a 43-year-old individual with height 167 cm and mass 59kg (Reference Female: 163cm and 60kg), performed at a relatively high resolution preparatory to whole-body irradiation for the treatment of leukaemia. The data set consisted of 174 slices of 5-mm width (head and trunk) and 43 slices of 20-mm width (legs), each with 256 x 256 pixels. From the 20-mm-slice images, intermediate slices of 5-mm thickness were obtained by interpolation. The resulting data set consisted of 346 slices. The voxel size was then 5mm in height with an in-plane resolution of 1.875mm, resulting in a voxel volume of 17.6mm³. The patient was lying on her hands, and at the time of scanning, one shoulder was positioned higher than the other. In total, 88 objects were segmented, and the number of different bone sites was 19 (Step 2 of Fig. 2.1).

(21) The segmentation of this voxel phantom was performed with a commercial software package (Biomedical Image Resource, Analyze AVW, Rochester, MN, USA). The segmentation tools used were grey value thresholding, region growing, and manual segmentation involving Bezier spline functions that could be inherited from slice to slice, and adjusted to the actual organ outlines by moving their control points. The resulting voxel phantom was called ‘Laura’ (Zankl et al., 2005).

Back to Article Outline

4. Modifications of segmented images to create reference computational phantoms 

(22) In order to create the reference computational phantoms from the two individual segmented data sets, the following steps were followed: (1) adjustment of the body height and the skeleton mass of the original model to the reference values by voxel scaling (Steps 3 and 4 of Fig. 2.1); (2) adjustment of the organ masses to the reference values by adding or subtracting the required number of organ voxels (Step 5 of Fig. 2.1); and (3) adjustment of the whole-body mass to the reference values by adding or subtracting an appropriate number of adipose tissue voxels (Step 9 of Fig. 2.1).

(23) It was the intention to keep the modifications to the skeleton shape to a minimum in order to preserve the ‘frame’ of the body. However, it was necessary to increase the size of the skull because otherwise it was not possible for both models to accommodate the entire reference brain mass within the skull. The segmented male had a noticeably narrow head, and other organs in the head were also small compared with the reference values. Therefore, it was decided to increase the size of all voxels of the entire head and then resample this volume with the smaller voxel size of the rest of the body. Thus, the male reference phantom has a greater number of head voxels than Golem. For the female reference phantom, only the skull size was increased; the interior surface voxels of the skull were replaced by brain, and an additional layer of skull voxels was added at the exterior surface. In order not to lose a layer of the surrounding tissues, this had to be preceded by an outward movement of the surrounding muscle, adipose tissue, and skin voxels. Furthermore, an outward movement of the ribs (as would occur during breathing) was necessary to accommodate the liver. This was done while the thicknesses of tissues covering the ribs – muscle, adipose tissue, and skin – remained the same.

4.1. Voxel scaling 

(24) Apart from these unavoidable modifications of the shape of the skeleton, the volume of the skeleton was adjusted to the reference value by voxel scaling. Since Golem’s body height corresponded to the reference value, the original voxel height was kept unmodified. Laura was taller than the Reference Female, so the voxel height for the female reference computational phantom was reduced from 5.0 to 4.84mm (Step 2 of Fig. 2.1). After the mentioned moderate changes to the skulls of both phantoms, the numbers of segmented skeleton voxels were 211,427 and 378,204 for the male and female reference phantoms, respectively. Table 4.1 shows the volumes of the skeletons of Reference Male and Reference Female as derived from the reference mass data from Publication 89 (ICRP, 2002) and mass density data of ICRU Report 46 (ICRU, 1992).

Table 4.1. Reference mass values (ICRP, 2002), mass density values (ICRU, 1992), and volumes derived for various constituents of the skeleton of the Reference Male and Reference Female.

	Mass (g)		Mass density (g/cm3)	Volume (cm3)
	Male	Female		Male	Female
Mineral bone	5500	4000	1.92	2860	2080
Cortical bone	4400	3200		2290	1670
Trabecular bone	1100	800		570	420
Cartilage	1100	900	1.10	1000	820
Active bone marrow	1170	900	1.03	1140	870
Inactive bone marrow	2480	1800	0.98	2530	1840
Miscellaneous	200	160	1.03	190	160
Total	10,450	7760	1.35	7730	5770

(25) From these values and the segmented skeleton voxel numbers, voxel volumes of 36.54 and 15.25mm³ for the male and female reference computational phantoms, respectively, were evaluated, simply by dividing the volume values aimed at by the respective numbers of skeleton voxels for both phantoms. The in-plane voxel resolution for each phantom was then derived as the square root of the in-plane voxel area (derived by dividing the voxel volume by the voxel height). This resulted in voxel in-plane resolutions of 2.137mm for the male and 1.775mm for the female reference computational phantoms. Compared with the original voxel sizes of Golem and Laura, this corresponds to an increase of 5.6% in voxel volume for the male reference computational phantom and a decrease of 13.2% for the female reference computational phantom (Step 3 of Fig. 2.1).

4.2. Individual organ volume modifications 

(26) A software package, ‘VolumeChange’, was developed to modify the organs of the original voxel models in volume, location, and shape (Becker et al., 2007). It uses the programming language IDL (‘Interactive Data Language’), and identifies each organ by its surface voxels, i.e. all voxels having at least one neighbour that does not belong to the same organ. The volumes, and consequently masses, are then modified by shifting surface voxels – inward for decreasing and outward for increasing the respective volume.

(27) VolumeChange is not a segmentation tool and does not solve the problem of laborious and very time-consuming segmentation. Only existing voxel models (and organs) can be edited; however, this task can be completed in a couple of days. Furthermore, every organ modification can be easily visualised, together with its effect on the surrounding tissues, checked, or reversed; this supports the anatomical realism and reproducibility of the modified model.

(28) Having overcome the initial difficulties with skull sizes as described above, the individual organs were adjusted one-by-one to the respective reference values, beginning with those that were larger than reference size in order to make room for those that had to be enlarged. Some very fine structures could not be adjusted exactly to the reference values, due to limitations of voxel resolution and visibility. For most organs, however, a very close approximation of the reference values could be achieved. The only limitation then was due to the fact that each organ has to consist of an integer number of voxels. That means that the resulting volumes may deviate from the desired value by, at most, half a voxel volume, i.e. approximately 18.3mm³ for the male reference computational phantom and 7.6mm³ for the female reference computational phantom (Step 5 of Fig. 2.1).

4.3. Additional organ and body region modifications 

(29) In an additional attempt to improve the models, further anatomical details were segmented in the reference computational phantoms, going back to the original computed tomography images (Step 6 of Fig. 2.1). Some effort was made to identify additional blood vessels, which was especially demanding for the male phantom due to the relatively large slice thickness resulting in decreased recognition of very fine structures. Inside the lungs, larger blood vessels were also segmented, as well as an increased amount of bronchi; thus, the reference computational phantoms have inhomogeneous lungs in contrast to Golem and Laura. The rest of the lung volume was then assigned a homogeneous tissue, the density of which was selected such that the whole mass of the lungs, including the blood, corresponds to its reference value.

(30) The individual whose medical image data were used to construct the female reference computational phantom was lying on her hands during the computed tomography scan. In order to avoid shielding of the body by the arms for posterior irradiation, the arms had to be shifted to the sides. At the height of the fingertips, movement of the left hand by 71 pixels to the left and movement of the right hand by 68 pixels to the right was necessary to separate the hands from the trunk. To preserve the realistic geometry of the shoulders and create smooth transitions for the sideward movement, the upper two slices of the arms were moved by just one pixel, the next two slices by two pixels, and so on. Where the hands had been, the trunk shape showed indentations that had to be removed. Therefore, a smooth body contour was drawn manually instead of the indented one in these slices, and this was then filled by adipose tissue. Similarly, a small indentation at the bottom of the trunk of the male phantom, due to one of the arms, had to be removed.

(31) After adjusting the masses of the organs to their reference values, the internal anatomy was fixed. At this stage, lymphatic nodes were incorporated. Since these tissues could not be identified on the medical images, they were drawn manually at locations specified in anatomical textbooks (Brash and Jamieson, 1943, Möller and Reif, 1993, Möller and Reif, 1997, GEO kompakt, 2005). Although only part of the lymphatic tissue reference mass could thus be introduced, the distribution throughout the body and the higher concentration at specified locations, such as in the groins and axillae, and to some extent the hollows of the knees, crooks of the arms, etc., was correctly mirrored (Step 8 of Fig. 2.1).

(32) The final step was to adjust the whole-body masses to 73 and 60 kg for the male and female reference computational phantoms, respectively. In both cases, the whole-body masses were lower than the reference values, so the body had to be ‘wrapped’ with additional layers of adipose tissue (Step 9 of Fig. 2.1). Towards the end of this procedure, small iterations had to be made, since each modification of the number of adipose tissue voxels resulted in small changes to the skin mass, because the number of body surface voxels was modified. Finally, the whole-body masses were adjusted to the reference values within 0.01g.

(33) Due to representing a thickness of several millimetres, the top and bottom slices of the segmented phantoms contain not only skin, but also other structures of the head and the feet, such as adipose (residual) tissue, muscle, and bone. To make sure that all surface voxels are skin, additional slices were added at the top of the head and the bottom of the feet with voxels that cover the segmented structures of the neighbouring slices. These are additional voxels, not included in the reference height and mass, and were assigned a separate identification number (141, called ‘skin at top and bottom’ in Annex A). It is at the discretion of the user and will depend on the situation under consideration whether or not these additional slices are actually used in a dose calculation. A simple way to neglect them is to assign air instead of skin as the medium for this identification number. For these organ and body region modifications, two different software tools were used (Biomedical Image Resource, Analyze AVW, Rochester, MN, USA; Becker et al., 2007).

4.4. Sub-segmentation of the skeleton 

(34) The skeleton is composed of cortical bone, trabecular bone, active (red) and inactive (yellow) bone marrow, cartilage, teeth, and miscellaneous skeletal tissues (periosteum and blood vessels). A sub-region of the bone marrow, 50μm from the bone surfaces, is further defined as the endosteal tissues. The dimensions of internal structures of most of these tissues are in the order of micrometres and therefore smaller than the resolution of a normal computed tomography scan (order of millimetres); thus, these volumes could not be segmented. However, the gross spatial distributions of the source and target volumes were represented as realistically as possible for the given voxel resolution (Zankl et al., 2007). For this purpose, the skeleton was divided into the 19 bones and bone groups for which individual data on red bone marrow content and marrow cellularity are given in Publication 70 (ICRP, 1995). These bones are: upper halves of humeri, lower halves of humeri, lower arm bones (ulnae and radii), wrists and hand bones, clavicles, cranium, upper halves of femora, lower halves of femora, lower leg bones (tibiae, fibulae, and patellae), ankles and foot bones, mandible, pelvis (os coxae), ribs, scapulae, cervical spine, thoracic spine, lumbar spine, sacrum, and sternum. These were then sub-segmented into an outer shell of cortical bone and the enclosed spongiosa part of the bone. The long bones contain a medullary cavity as a third component, which is enclosed by cortical bone. This sub-division resulted in 44 different identification numbers in the skeleton: two – cortical bone and spongiosa – for each of the 19 bones mentioned above, and a medullary cavity for each of the six long bones (upper and lower half of humeri, lower arm bones, upper and lower half of femora, and lower leg bones). Furthermore, the amount of cartilage that could be identified on the computed tomography images and could, thus, be segmented directly was attributed to four body parts – head, trunk, arms, and legs. Hence, the skeleton covers a total of 48 individual identification numbers (Step 7 of Fig. 2.1).

(35) The total volume of each bone results directly from the number of segmented voxels and the voxel volume. The cortical shell around the spongiosa was chosen to be one voxel layer; the cortical bone at the long bones’ shafts is thicker, and its thickness was adjusted such that the total cortical bone volume is in agreement with the reference value. For each of the 19 bones and bone groups, the spongiosa is composed of various proportions of trabecular bone, red marrow, and yellow marrow. Furthermore, the additional volumes of ‘miscellaneous’ tissues (ICRP, 2002) and the non-segmented cartilage had to be accommodated in the skeleton; for practicability, these were merged within the spongiosa volume of all skeletal sites.

(36) The volume of red marrow in each of the 19 bones and bone groups can be calculated from the reference values of the total amount of red marrow (ICRP, 1995, ICRP, 2002) and its percentage distribution among individual bones as given in Publication 70 (ICRP, 1995) and Publication 89 (ICRP, 2002) based on earlier data of Cristy (Cristy, 1981). The bone marrow cellularity (Cristy, 1981, ICRP, 1995) in an individual bone gives the volume proportion of the entire marrow in this bone that is still haematopoietically active, i.e. the red bone marrow fraction. From this value, the total bone marrow volume in that bone can be calculated. This method permits evaluation of the volume of yellow marrow from the red bone marrow volume for all those bones with a cellularity that is non-zero. Of course, for those bones with zero cellularity, the yellow marrow content cannot be estimated by this method. Therefore, the difference remaining between the total inactive marrow volume (ICRP, 1995, ICRP, 2002) and that assigned to individual bones using the non-zero cellularity values was distributed among those bones that do not contain red bone marrow. These are the lower halves of humeri and femora, the lower arm and leg bones, and the hand and foot bones. A portion of the remaining yellow bone marrow is contained within the segmented medullary cavities of the lower halves of humeri and femora, and the lower arm and leg bones; the rest was attributed to the spongiosa regions of all zero-cellularity bones in relation to their respective volumes. Accordingly, each of the 19 bones or bone groups has its own unique bone-specific spongiosa composition. The masses of red and yellow bone marrow in each bone (group) are given in Table 4.2, together with the masses of the endosteal tissue.

Table 4.2. Masses of red marrow, yellow marrow, and endosteum in the 19 bones and bone groups of the reference computational phantoms.

RBM, red bone marrow; YBM, yellow bone marrow.

(37) For some bones, e.g. the sternum of the male reference computational phantom, and the sacrum and sternum of the female reference computational phantom, the total bone marrow volume that is required to accommodate the reference volumes of red and yellow bone marrow (ICRP, 1995) is only marginally less than the segmented total volume of these bones. Consequently, there was not enough space left in these bones for an enclosing cortical shell. Cortical bone voxels around the sternum were, therefore, segmented in only a few slices of both models. Although it would have been possible to spare a few voxels to accommodate a small amount of cortical bone around the female phantom’s sacrum, this option was abandoned in order not to exceed the reference value of 1666.7cm³ for the total cortical bone volume. Similarly, in the ribs of both models, the space for cortical bone was limited. Only the larger portions are enclosed in a cortical shell; smaller parts – especially those with a large surface-to-volume ratio – are not given a cortical shell. This difficulty was encountered since reference masses and volumes had to be accommodated at the assigned voxel size.

(38) The segmented cortical bone has the reference mass of Publication 89 (ICRP, 2002), and although no direct segmentation of the trabecular bone and active and inactive marrow in the spongiosa was possible at the given voxel resolution, the volume and composition of the spongiosa is such that all skeletal constituents (mineral bone, cartilage, active and inactive marrow, and miscellaneous tissues) are incorporated in the skeleton at their reference masses as given in Table 4.1.

Back to Article Outline

5. Description of the adult reference computational phantoms 

5.1. Main characteristics of the phantoms 

(39) The orientation of the three-dimensional voxel array (arranged in columns, rows, and slices) describing the computational phantom is as follows. The columns correspond to the x co-ordinates, the rows correspond to the y co-ordinates, and the slices correspond to the z co-ordinates. Column numbers increase from right to left, row numbers increase from front to back, and slice numbers increase from the toes up to the vertex of the body.

(40) The main characteristics of the adult male and female reference computational phantoms are summarised in Table 5.1.

Table 5.1. Main characteristics of the adult male and female reference computational phantoms.

(41) Table 5.2 shows a list of source and target regions of the two phantoms, their segmented volumes, and resulting masses. For comparison, the reference masses (ICRP, 2002) are also shown.

Table 5.2. List of source and target regions, their segmented volumes, and resulting masses compared with the reference masses (ICRP, 2002).

Organ	Male			Female
	Volume (cm3)	Mass (g)	Reference mass (g)	Volume (cm3)	Mass (g)	Reference mass (g)
Adrenals	13.6	14.0	14	12.6	13.0	13
Blood (segmented vessels)∗	973.7	1032.1	5600	807.4	855.8	4100
Brain	1381.0	1450.0	1450	1238.1	1300.0	1300
Breast	25.6	25.0	25	511.9	500.0	500
Eyes	14.3	15.0	15	14.3	15.0	15
Eye lenses	0.4	0.4	0.4	0.4	0.4	0.4
Gall bladder	66.0	68.0	68	54.3	56.0	56
Gall bladder wall	13.5	13.9	10	9.9	10.2	8
Gall bladder contents	52.5	54.1	58	44.4	45.8	48
Gastro-intestinal tract
Stomach wall	144.2	150.0	150	134.6	140.0	140
Stomach contents	240.4	250.0	250	221.2	230.0	230
Small intestine wall	625.0	650.0	650	576.9	600.0	600
Small intestine contents	336.6	350.0	350	269.2	280.0	280
Right colon wall	144.2	150.0	150	139.4	145.0	145
Right colon contents	144.3	150.0	150	153.8	160.0	160
Left colon wall	144.2	150.0	150	139.4	145.0	145
Left colon contents	72.1	75.0	75	76.9	80.0	80
Recto-sigmoid colon wall	67.3	70.0	70	67.3	70.0	70
Sigmoid colon contents	72.1	75.0	75	76.9	80.0	80
Heart	795.4	840.0	840	587.2	620.0	620
Heart wall	314.3	330.0	330	238.1	250.0	250
Heart contents (blood)	481.1	510.0	510	349.1	370.0	370
Kidneys	295.3	310.0	310	261.9	275.0	275
Liver	1714.3	1800.0	1800	1333.3	1400.0	1400
Lungs	2891.3	1200.0	1200	2300.8	950.0	950
Lymphatic tissue†	134.0	138.0	730	76.8	79.1	600
Muscle tissue	27,619.0	29,000.0	29,000	16,666.7	17,500.0	17,500
Oesophagus	38.8	40.0	40	34.0	35.0	35
Ovaries				10.6	11.0	11
Pancreas	133.3	140.0	140	114.3	120.0	120
Pituitary gland	0.6	0.6	0.6	0.6	0.6	0.6
Prostate	16.5	17.0	17
Residual (adipose) tissue	21,535.2	20,458.4	18,200	24,838.3	23,596.4	22,500
Salivary glands	82.5	85.0	85	68.0	70.0	70
Skin	3420.2	3728.0	3300	2496.8	2721.5	2300
Skeleton	7725.3	10,450.0	10,450	5767.4	7760.1	7760
Cortical bone‡	2291.7	4400.0	4400	1666.7	3200.0	3200
Trabecular bone§	572.9	1100.0	1100	416.7	800.0	800
Cartilage¶	1000.0	1100.0	1100	818.2	900.0	900
Active marrow§	1135.9	1170.0	1170	872.9	899.1	900
Inactive marrow¶	2530.6	2480.0	2480	1836.8	1800.1	1800
Miscellaneous§	194.2	200.0	200	155.3	160.0	160
Spleen	144.2	150.0	150	125.0	130.0	130
Teeth	18.2	50.0	50	14.6	40.0	40
Testes	33.7	35.0	35
Thymus	24.3	25.0	25	19.4	20.0	20
Thyroid	19.2	20.0	20	16.4	17.0	17
Tongue	69.5	73.0	73	57.1	60.0	60
Tonsils	2.9	3.0	3	2.9	3.0	3
Ureters	15.5	16.0	16	14.6	15.0	15
Urinary bladder wall	48.1	50.0	50	38.5	40.0	40
Urinary bladder contents	192.3	200.0		192.3	200.0
Uterus				77.7	80.0	80
Total body	71,109.9	73,000.0	73,000	59,258.0	60,000.0	60,000

Some of the reference values (e.g. blood and lymphatic tissue) duplicate other mass information and are, thus, not additive.

5.2. The skeleton 

(42) For internal sources in the skeleton, the following source regions have to be considered: cortical bone (surface or volume), trabecular bone (surface or volume), and cortical and trabecular bone marrow. As in previous calculations for internal photon and neutron sources (Cristy and Eckerman, 1987a), no distinction is made between surface and volume sources in cortical and trabecular bone. The results from the respective volume sources will be applied to estimate values for the surface sources.

(43) A cortical bone volume has been segmented separately in most of the 19 bones and bone groups of the skeleton (for exceptions, see Para. 37), so the entirety of these segmented voxels can be directly used to sample a uniform source distribution.

(44) As described above, trabecular bone is one of the constituents that make up the spongiosa. Therefore, the entirety of the segmented spongiosa voxels of all bones serves as the volume which can be sampled for particle emissions. However, since the relative amount of trabecular bone in the spongiosa varies between individual bones and bone groups, the source should not simply be assumed to be homogeneously distributed; for Monte Carlo radiation transport calculations, the variation of trabecular bone concentration should be used, e.g. either to determine the probability of a starting point being selected, or to assign bone-specific initial ‘statistical weights’ to the particles starting in the spongiosa volumes of various regions of the skeleton.

(45) For cortical marrow, medullary cavities have been segmented in the shafts of the long bones, so the entirety of these segmented voxels can be directly used to sample a uniform source distribution. In the adult models considered here, the medullary cavities only contain yellow marrow.

(46) The trabecular marrow is the marrow situated in the spongiosa. Consequently, the same particle source sampling principle as for the trabecular bone should be applied for the trabecular marrow, now considering the bone-specific relative bone marrow content (red and yellow marrow).

(47) The skeletal target tissues of interest are the red bone marrow and the endosteal tissues (formerly called ‘bone surfaces’). Since the dimensions of the marrow cavities and the endosteal layer lining these cavities (assumed thickness: 50 μm) are clearly finer than the resolution of a normal computed tomography scan, they could not be directly segmented and had to be incorporated in the spongiosa volumes, as discussed. However, in contrast to the source volumes described above, it is not sufficient to consider the bone-specific relative amounts of these tissues in the spongiosa. The reason is that secondary equilibrium between the mineral bone and soft tissue components of the spongiosa does not exist; secondary particles are primarily released from mineral bone into the soft tissues during photon interactions, and these cause a dose enhancement in red bone marrow and endosteum compared with the mean dose in the spongiosa volume (Spiers, 1969, King and Spiers, 1985). Therefore, in skeletal dosimetry, specific techniques have to be applied, such as the use of correction factors (e.g. Zankl et al., 2002, Schlattl et al., 2007) or fluence-to-dose response functions that are multiplied with the particle fluence inside specific bone regions to give the dose quantities of interest to the target tissues (Cristy and Eckerman, 1987b).

5.3. Blood 

(48) It was not possible to segment the entire blood pool of the body; only the larger vessels could be identified on the computed tomography images, especially for the male where the slice thickness was greater and therefore partial volume effects blurred the boundaries. The larger part of the blood volume is situated in the small vessels and capillaries inside most of the organs, and it was not possible to segment models of the intra-organ vasculature (aside from some major vessels within the lungs). On the other hand, the elemental compositions listed in Publication 89 (ICRP, 2002) and ICRU Report 46 (ICRU, 1992) are exclusive of blood and thus are only relevant to the organ parenchyma. The blood content in each organ was, therefore, considered by including a blood portion in the elemental tissue composition of each organ. This distribution of the blood among the individual organs was again taken from Publication 89 (ICRP, 2002) for the male and female phantom separately.

(49) The following expression was used to assign blood-inclusive elemental compositions to all phantom tissues. For example, the percentage by mass of hydrogen in the liver of the reference phantom would be calculated as:

(1)

(2)

(50) A table of the percentage distribution of the entire blood volume in individual organs is given in Section 7.7.2 of Publication 89 (ICRP, 2002). The amount of blood can, thus, be evaluated for the large majority of organs. The liver, as an example, contains 10% of the blood volume; i.e. 0.53l (or 560g) for the male reference computational phantom and 0.39l (or 410g) for the female reference computational phantom. Expressions similar to Eq. (1) are applied in the data given in Annex B for the tissue compositions of the reference computational phantoms inclusive of organ blood content.

5.4. Limitations due to image resolution 

(51) In addition to the skeletal source and target tissues, other regions could not be fully segmented in the reference computational phantoms, or could not be adjusted to their reference masses at the given voxel resolution due to their small size or complex structure.

(52) The ET airways are represented by an entire voxel layer lining the airways of nose, larynx, and pharynx. This does not mirror their small dimension of only a few micrometres thickness, but locates them at their correct anatomical position. The same holds for the trachea.

(53) The bronchi were not followed down more than the very first generations of branching. Small volumes of bronchial tissues were segmented to greater depths in the lungs, although not connected by a tree-like structure.

(54) The bronchioles are too small to be segmented in a voxel phantom. These form the rest of the lungs as homogeneous tissue with a density that is the average of the higher density bronchiolar tissue and included air.

(55) The skin is represented by a voxel layer, wrapping the phantoms’ exterior. This renders it a thickness of 2.137mm for the male reference computational phantom and 1.775mm for the female reference computational phantom. The total skin masses of 3728.0g for the male phantom and 2721.5g for the female phantom thus derived are 13% and 18% higher than the reference values of 3300g and 2300g, respectively. Using the skin density of approximately 1.1g/cm³ and the reference body surface area of 1.90m² for the male and 1.66m² for the female, the reference skin thickness (epidermis and dermis) can be assumed to be approximately 1.6mm and 1.3mm for the adult male and female, respectively.

(56) The cartilage was only partly segmented due to its distributed nature and low contrast in the original images.

(57) In both reference computational phantoms, the gall bladder wall mass is higher than its reference value. The gall bladder is a small organ and, at the given voxel resolution, the number of wall voxels was not sufficient to encompass the volume of its contents. The entire mass of both the wall and contents was assigned the reference value, and the enclosing voxel layer represents the gall bladder wall.

(58) The residual tissue voxels of the reference computational phantoms were assigned as adipose tissue with the number of residual tissue voxels adjusted to permit matching of the reference total body mass for each phantom. Through this method, matching of reference adipose tissue masses was approximately achieved.

(59) The finite voxel resolution of the two reference computational phantoms limits their application to short-ranged radiations such as beta and alpha particles. For example, for assessing depth doses in the tissues of the respiratory airways of the Human Respiratory Tract Model (ICRP, 1994) or the walls of the stomach, small intestine, or colon of the Human Alimentary Tract Model (ICRP, 2006), the corresponding organs presented in these phantoms make these calculations impossible. Supplemental stylised or fine-resolution voxel tissue models must thus be used to compile absorbed fraction data relevant to particulate radiations in the body. However, in most cases, photon and neutron absorbed doses to these same tissues may be taken from radiation transport simulations in the reference computational phantoms directly or through surrogate tissue assignment.

Back to Article Outline

6. Applications and limitations of the reference computational phantoms 

(60) The phantoms presented in this document are the official computational models representing the adult Reference Male and Reference Female. These reference computational models are based on computed tomographic data of real people and, hence, represent digital three-dimensional representations of human anatomy. They are defined to enable calculations of the protection quantities – organ and tissue equivalent dose and effective dose – from exposure to ionising radiation. ICRP will publish recommended values for dose coefficients for both internal and external exposures using the male and female adult reference computational phantoms. These documents will include specific absorbed fractions for particles relevant to internal exposures, and dose conversion coefficients for external radiation fields.

(61) It should be clear that although these phantoms have organ masses of reference values, they still have individual organ topology reflecting the tomographic data used in their construction. Obviously, both models cannot represent real individuals, and thus they should not be used to assess doses for specific individuals. While the reference computational phantoms were created for the purpose of deriving radiological protection quantities, it is acknowledged that the phantoms may have broader applications. However, the specific limitations related to their intended application have to be kept in mind.