Posture-Transformed Monkey Phantoms Developed from a Visible Monkey

A monkey phantom is of significant value for electromagnetic radiation (EMR) dosimetry simulations. Furthermore, phantoms in various postures are needed because living beings are exposed to EMR in various postures during their daily routine. In this study, we attempted to produce monkey phantoms based on three daily postures of a rhesus monkey. From our Visible Monkey project, we selected surface models with 177 monkey structures. In the surface models, 52 virtual joints were created to allow for changes from the anatomical position to quadrupedal and sitting positions using commercial software. The surface models of the three positions were converted into monkey voxel phantoms. These phantoms were arranged in three positions, and the number of voxels and mass of each structure were analyzed. The phantoms in anatomical, quadrupedal, and sitting positions have a total of 5,054,022, 5,174,453, and 4,803,886 voxels, respectively. The mass of 177 structures in three positions were also calculated based on the number of voxels. By comparing the monkey phantom with the phantom of a female human, we confirmed thicker skin, less fat, heavier muscle, and a lighter skeleton in monkeys than those in humans. Through posture-transformed monkey phantoms, more precise EMR simulations could be possible. The ultimate purpose of this study is to determine the effects of EMR on humans. For this purpose, we will create posture-transformed human phantoms in a following study using the techniques employed herein and the human phantoms from our previous study.

Although human phantoms are useful in various fields, they have two limitations. The first has to do with reliability. To verify the reliability of a human phantom in any simulation, it is ideal to compare the effects of a human phantom with those of an actual human being. However, EMR tests on living humans are impossible because the effects are unknown and thus tests on living subjects have been banned for ethical reasons. Therefore, as a substitute, live monkeys are used in various biological and physical experiments deemed too difficult or dangerous for humans [13][14][15][16][17][18]. In EMR fields, a living monkey can also replace a living human, and a monkey phantom can also replace a living monkey. Consequently, verification of the reliability of a human phantom is possible using monkey phantoms because we can logically deduce the relation between the human phantom and a living human through a comparison of a monkey phantom with a living monkey.
The second hindrance is the availability of only one posture in existing phantoms. Most phantoms are produced in a supine position from CT or MRI data [1][2][3][4][5][6][7][8]. To accurately simulate the EMR effects on a phantom, the phantoms must be exposed to electromagnetic waves during specific postures because the tissues and organs of living creatures may be influenced differently depending on posture [9,19,20]. In a recent study, a posturetransformed human phantom was produced for EMR simulation. However, the techniques required to create the phantom were so professional that researchers unfamiliar with mathematical algorithms cannot easily repeat the technique [9].
For verifying MRI stability, EMR simulations based on a monkey phantom of only one anatomical position has been recently produced [21]. The monkey phantom was made from data of our Visible Monkey project [22]. However, if an EMR simulation is conducted not only for medical devices but also for electronic devices on a monkey phantom, then the original posture of the phantom will be unsuitable because living monkeys are exposed to electromagnetic radiation in different postures during their daily routines. To address this limitation, a phantom was needed that can assume common postures in the everyday life of a monkey [23][24][25][26].
The primary aim of this study was to present posture-transformed monkey phantoms that are able to assume anatomical, quadrupedal, and sitting positions among the most common everyday postures used for EMR simulations. The secondary aim was to present easy methods for making posture-transformed phantoms using commercial software. To achieve these aims, in this study, surface models of Visible Monkeys in anatomical positions [27] were transformed into models in quadrupedal and sitting positions and then converted into voxel phantoms.

Materials and Methods
In our Visible Monkey project, real-color sectioned images of the whole body of a female rhesus monkey (intervals of 0.05 and 0.5 mm, pixel size of 0.024 mm, and color depth of 48 bits) were produced [22]. Based on the sectioned images, surface models of 177 structures in an anatomical position were created [27]. Thus, 177 surface models in a stereolithography (STL) format were adopted for this study.
Using Maya version 2018 (Autodesk, Inc., San Rafael, CA, USA), the posture of the surface models in STL format was changed as follows. In total, 52 virtual joints were created-i.e., all intervertebral spaces (7 cervical, 12 thoracic, 7 lumbar, 1 sacral, and 13 caudal vertebral joints), shoulder (2 joints), elbow (2 joints), wrist (2 joints), hip (2 joints), knee (2 joints), and ankle (2 joints)-using the Create Joints tool (Figure 1a). After creating the virtual joints, all surface models were bound to the joints using the Bind Skin tool (Figure 1b). The influence values of the virtual joints in the Paint Skin Weights tool were adjusted to create realistic shapes of the surface models when rotated on an axis at the joint ( Figure 1c). All surface models were changed from the anatomical position to the quadrupedal position by controlling the virtual joints using the Move and Rotate tools (Figure 1d). In the same manner, the postures of the surface models were changed from an anatomical position to a sitting position. The shape of quadrupedal and sitting positions were repeatedly revised and examined according to posture descriptions of primate studies targeting multiple subjects [24][25][26]. In the surface models of three positions, abnormal shapes were found and repaired using the Move, Smooth, Relax, Grab, and Flatten tools in Maya as follows. First, in the surface models of the anatomical position (Figure 2a), the skin surfaces (meshes) of an arm around the axilla overlapped with those of the thorax despite no such overlap in reality. For this reason, when the upper limbs surface models were flexed, the skin meshes expanded abnormally and unrealistically (Figure 2b). To resolve this, we manually modified the meshes near the area (Figure 2c). Second, in the surface models of the sitting position, some meshes of the liver became intruded or merged on themselves when only a vertebral joint was flexed (Figure 2e). To minimize these invasions, we flexed all joints in every vertebra ( Figure 2f). Third, in the surface models of the sitting position, meshes of the femur invaded those of the skin on the knee joint during the flexion of the lower limbs ( Figure 2h). To prevent such an invasion, the influence value in the Paint Skin Weights tool in Maya was manually repaired again ( Figure 2i). The surface models were saved in STL format to create the 177 surface models in both the quadrupedal and sitting positions ( Figure 1e). In Vstudio version 2.0 (copyright ETRI, Daejeon, Korea), STL files of the surface models in three positions were loaded with a structure list consisting of the assigned identification numbers (ID) for each structure [28]. After converting the loaded surface models into voxel models, the shapes of the voxel models were checked stereoscopically and cross-sectionally. Using the Export Voxel tool, voxel information (pixels, intervals, and ID) of the 177 structures in the voxel models were saved in TXT format to create the monkey phantom for the three positions ( Figure 1f) [28,29].
In the TXT files of the three phantom positions, the ID of the 177 structures and the coordinates (x-, y-, and z-axes) of each ID were recorded using the x-and y-axes for the pixel size and the z-axis for the intervals in the voxels. To calculate each ID in the TXT files, a house-developed software was used [29]. Because one ID was written for each voxel in the TXT files, the software calculated the total number of voxels in a body by checking the total for each ID.
The number of voxels for each structure was multiplied by its tissue density using the Gabriel list to determine the mass of the 177 structures [3,30]. To determine the relation between humans and monkeys in the EMR simulation, as a preliminary step, the mass of the structures of the female monkey were compared with those of a VK-phantom of a female human [29].
The study protocol of Visible Monkey was reviewed and approved by the Institutional Review Board (IRB) of the National Primate Research Center of Korea, Research Institute of Bioscience and Biotechnology (IRB No. KRIBB-AEC-18087) [22].     (Table 1). In the TXT files for the three positions, the coordinates (x-, y-, z-axes) and ID of each structure were recorded. Using these TXT files, the researchers could simulate the EMR research of the monkeys [7,10].

Results
Using monkey phantoms stored as TXT files, the number of voxels of each structure were analyzed. The total number of voxels of the anatomical, quadrupedal, and sitting positions were 5,054,022, 5,174,453, and 4,803,886 voxels, respectively. The discrepancies in the total number of voxels among the three positions were under 5%, which was similar to a previous study [31]. In the phantoms, the muscle tissue had the greatest number of voxels (1,699,435 anatomical position voxels, 1,767,197 quadrupedal position voxels, and 1,645,241 sitting position voxels), and the superficial temporal vein had the lowest number of voxels (5 anatomical position voxels, 3 quadrupedal position voxels, and 2 sitting position voxels). However, the number of voxels of the arteries and veins, including the superficial temporal vein, were inaccurate because their segmentations had not yet been completed. Excluding the arteries and veins, the cornea had the smallest number of voxels (18 anatomical position voxels, 22 quadrupedal position voxels, and 19 sitting position voxels) (see the Appendix A), which was the same for cornea of the VK-phantom of a female human [29].
In general, structures with a large or small number of voxels tend to either be heavy or light, respectively. However, the rank order of numbers of voxels does not always correspond to the rank order of mass because their densities differ depending on the tissue type. For example, in the anatomical position, the gallbladder ranked 66th in number of voxels, whereas it ranked 94th in mass. The differences in the total number of voxels between changed positions and the anatomical position were set at under 5%. Unfortunately, these differences (i.e., under 5%) were unavoidable because the changes in the positions caused spatial changes in the surface models, leading to changes in the volumes of the surface models. Eventually, volume changes in the surface models resulted in differences in the number of voxels (Appendix A).
Most experiments using monkeys as subjects are conducted for human health. As a preliminary step to determine the relation between humans and monkeys in EMR simulations, we compared the proportion of mass of each structure to the total mass of the human phantoms from a previous study and the monkey phantoms [29]. The results showed that a monkey phantom has larger proportions of skin (7.62%) and muscle (33.84%) than a VK-phantom (3.09% and 16.12%, respectively), whereas the proportion of bone (9.7%) of a monkey phantom is smaller than that of a VK-phantom ( Figure 4, Appendix A). It is reasonable for monkeys to have thicker skin, less fat, heavier muscle, and lighter skeletons than humans because they are arboreal animals living in a wild environment [32].
The monkey phantoms also have a larger proportion of tissues in the alimentary system, including the teeth (0.2%), tongue (0.14%), esophagus (0.2%), stomach (9.32%), small intestine (7.03%), large intestine (9.5%), and liver (2.71%) compared with a VK-phantom ( Figure 4, Appendix A). It is also logical for monkeys to have larger gastrointestinal (GI) organs than humans because they eat foods that are low in digestibility such as plants, which requires a relatively robust GI system with elaborate fermenting chambers [33]. However, the GI system of this phantom was larger than those of actual monkeys because gas was generated in the GI organs during the time between sacrifice and freezing of the monkey.
Not surprisingly, in the central nervous system, the monkey phantom had a smaller proportion of gray matter (1.13%), cingulate gyrus (0.02%), and cerebellum (0.09%) than the VK-phantom, whereas the white matter (0.39%), midbrain (0.03%), pons (0.03%), medulla oblongata (0.03%), and spinal cord (0.16%) of the monkey were larger (Figure 4, Appendix A). It is reasonable for monkeys to have a large paleencephalon (e.g., brainstem) and a relatively small neoencephalon compared to humans given that humans have a more evolved brain (e.g., gray matter in the cerebral cortex). Moreover, it is natural for monkeys to have a smaller proportion of cerebellar tissue than humans because humans use their hands more often and more elaborately by virtue of their bipedalism.  [29]. The mass proportions of skin, muscle, and structures of alimentary system of the monkey phantom are larger than those of the VK-phantom. The monkey phantom has larger proportions of cerebral white matter, brainstem, and spinal cord than those of the human.
The monkey phantoms also have a larger proportion of tissues in the alimentary sys tem, including the teeth (0.2%), tongue (0.14%), esophagus (0.2%), stomach (9.32%), smal intestine (7.03%), large intestine (9.5%), and liver (2.71%) compared with a VK-phantom (Figure 4, Appendix). It is also logical for monkeys to have larger gastrointestinal (GI organs than humans because they eat foods that are low in digestibility such as plants which requires a relatively robust GI system with elaborate fermenting chambers [33] However, the GI system of this phantom was larger than those of actual monkeys becaus gas was generated in the GI organs during the time between sacrifice and freezing of the monkey. Not surprisingly, in the central nervous system, the monkey phantom had a smalle proportion of gray matter (1.13%), cingulate gyrus (0.02%), and cerebellum (0.09%) than the VK-phantom, whereas the white matter (0.39%), midbrain (0.03%), pons (0.03%), me dulla oblongata (0.03%), and spinal cord (0.16%) of the monkey were larger (Figure 4 Appendix). It is reasonable for monkeys to have a large paleencephalon (e.g., brainstem and a relatively small neoencephalon compared to humans given that humans have a more evolved brain (e.g., gray matter in the cerebral cortex). Moreover, it is natural fo monkeys to have a smaller proportion of cerebellar tissue than humans because human use their hands more often and more elaborately by virtue of their bipedalism.

Discussion
Monkey phantoms are beneficial for research on the safety of EMR from ordinary electronic and medical devices because the effect of EMR on humans can be deduced based on the results of monkey phantoms [7,[10][11][12]. In addition, the monkey phantoms o daily postures are necessary to accurately examine the EMR effect because the results o the EMR are sensitive to various phantom postures [19,20]. However, strictly speaking, to  [29]. The mass proportions of skin, muscle, and structures of alimentary system of the monkey phantom are larger than those of the VK-phantom. The monkey phantom has larger proportions of cerebral white matter, brainstem, and spinal cord than those of the human. (* Cerebral gray matter including thalamus, hypothalamus, and hippocampus; § Brainstem including midbrain, pons, and medulla oblongata).

Discussion
Monkey phantoms are beneficial for research on the safety of EMR from ordinary electronic and medical devices because the effect of EMR on humans can be deduced based on the results of monkey phantoms [7,[10][11][12]. In addition, the monkey phantoms of daily postures are necessary to accurately examine the EMR effect because the results of the EMR are sensitive to various phantom postures [19,20]. However, strictly speaking, to examine the precise impact of EMR on living humans, living humans should be used in EMR tests; however, doing so is impossible. Therefore, human phantoms have been employed instead of actual humans [3,4,6,34]. However, we still cannot help but consider the reliability of such human phantoms because they have never been compared to living humans in relation to EMR. To verify the reliability of human phantoms, we chose to use monkeys, which are commonly applied as human surrogates. Using a monkey, researchers can directly check the impact of EMR on both living monkeys and monkey phantoms. If the reliability of the monkey phantoms is verified through comparisons with living monkeys, we can deduce logically the relation between human phantoms and living humans without the EMR testing of actual humans.
Because this study was the first step into monkey phantom research, we developed posture-transformed monkey phantoms (Figure 3b,d,f). To develop the posturetransformed monkey phantoms, three points must be considered.
First, monkey phantoms with precise anatomy are required to obtain accurate results with regard to EMR. Although phantoms are easily created using CT and MRI scans, small structures cannot be identified in detail in such phantoms [35][36][37][38]. By contrast, it is difficult to create phantoms from the sectioned images owing to the need for a high resolution and true color [39][40][41][42][43]. Nevertheless, we have successfully created VK-phantoms for 583 male whole-body structures, 459 female whole-body structures, and 300 head structures from the sectioned images of humans [29,44]. Using not only the image data of Visible Monkey but also the experience and software techniques employed for the VK-phantoms, precise phantoms of rhesus monkeys were produced in this study (Figure 3b,d,f) [22,27].
Second, posture-transformed monkey phantoms are required to obtain simulation values in situations identical to real life. EMR simulations usually use phantoms in the anatomical position [3,6,45]. However, in the real world, living things assume different postures, and are thus exposed to EMR in different postures. Consequently, phantoms in various positions are needed for more precise measurements. Although posture-transformed phantoms of whole bodies have been employed in some EMR and SAR studies (the hands of the phantoms were even placed over their ears), these studies did not provide voxel phantoms for other researchers [7,9,10]. Most importantly, not only a monkey phantom but also posture-transformed phantoms with detailed structures were applied. Therefore, in this study, considering the most common postures in the daily routine of a monkey, we chose anatomical, quadrupedal, and sitting positions, and produced phantoms of these three positions (Figure 3 and Table 1 and Appendix A). For reference, monkeys usually walk quadrupedally, and sleep and relax on tree branches in a sitting position.
Third, in the posture-transformed procedure, manual works are required to obtain phantoms with an anatomical correct shape. In the procedures applied in this study, abnormal expansions and intrusions of surface models occurred (Figure 2). These issues led to abnormal shapes of the phantoms, and the abnormal shapes of phantoms led to differences in the number of voxels of structures between positions. Eventually, the differences will affect the reliability and accuracy of the phantoms in EMR and SAR simulations. Therefore, to make reliable and accurate posture-transformed phantoms, researchers who plan to use the methods described in this study must consider these technical issues to minimize the differences in the numbers of voxel of structures between positions, even if manual works for such issues are difficult.
In addition, the total masses of the anatomical, quadrupedal, and sitting positions were approximately 5.5, 5.6, and 5.2 kg, respectively. Because the body mass of the real monkey was 4.3 kg, the discrepancy between the total mass of the anatomical position and the real body mass was approximately 1.2 kg (Appendix A) [22]. This was caused mainly for three reasons. First, the tissue density of Gabriel, which was used for the calculation of the monkey's body mass (Appendix A), has values of a human not a monkey. Second, in a real monkey, the inner spaces of the organs such as the stomach and intestines are filled with air, whereas the inner space of the organs in the phantom are regarded as the organs not air. Finally, because the segmentations had not yet been completed, the muscle, fat, and blood vessels were not divided in the phantoms. In the near future, we will be able to decrease the discrepancy of the masses because many researchers will find the correct tissue densities and will segment more structures. At that time, we will make new phantoms based on the new densities and new segmented images available.

Conclusions
Posture-transformable volume models are necessary for EMR simulations. Consequently, in the study, we presented the first monkey phantom of a small voxel size with 177 structures in the anatomical (Figure 3b), quadrupedal (Figure 3d), and sitting (Figure 3f) positions by changing the postures of the surface models and converting the surface models into volume models. It will be absolutely imperative to compare living monkeys with a virtual monkey (phantom) in EMR simulations in our following studies. Because the ultimate purpose of this research is to determine the effects of EMR on a living human body, posture-transformed human phantoms will be made in a future study using the techniques employed herein along with the VK-phantoms of male and female humans [29,44]. The surface models and phantoms of a rhesus monkey in the three positions applied will be useful not only in EMR simulations but also in anatomical and biological experiments. Informed Consent Statement: Not applicable.

Data Availability Statement:
Publicly available datasets were analyzed in this study. This data can be found here: https://www.data.go.kr/en/data/15074162/fileData.do (accessed on 10 May 2021).

Conflicts of Interest:
The authors have no potential conflict of interest to disclose.