Radiation plays a critical role in modern medical diagnostics and treatments but poses significant risks to healthcare personnel. Traditional dose estimation methods, primarily based on dosimeters placed on selected body parts, neglect the varying radiosensitivity of diferent organs. In this work, we present a system that models efective radiation dose by simulating threedimensional (3D) environments including radiation sources, shielding objects, and human models segmented by MeshCNN. We employed a ray tracing algorithm to simulate radiation behavior, considering both spatial attenuation (via inverse square law) and material shielding (via the Lambert-Beer law). Our approach allows for detailed analysis of organ-specific exposures and the impact of environmental shielding. Results demonstrate the feasibility of using 3D simulation and ray tracing to achieve a more comprehensive and accurate estimation of efective radiation dose in medical environments.
the efective dose incorporates tissue weighting factors, Radiation is widely utilized in the medical field for both providing a measure of the overall risk by considering diagnostic and therapeutic purposes. Despite its clinical the varying sensitivity of diferent organs. benefits, radiation exposure poses significant health risks Exposure to ionizing radiation carries both determinto medical personnel, especially those working in envi- istic and stochastic risks. Deterministic efects, such as ronments where frequent or prolonged exposure occurs. skin burns or cataracts, occur above a threshold dose. Efective radiation protection strategies are thus essential Stochastic efects, including cancer, have no threshold to mitigate these risks. and their probability increases with dose. The Interna
Radiation can be classified into particle radiation (e.g., tional Commission on Radiological Protection (ICRP) recalpha and beta particles) and electromagnetic radiation ommends dose limits to protect workers and the general (e.g., gamma rays and X-rays). While alpha and beta par- public, emphasizing the ALARA (As Low As Reasonably ticles have limited penetration abilities, gamma rays and Achievable) principle.
X-rays can deeply penetrate tissues, potentially damag- This work proposes a novel method to estimate the ing sensitive organs. Consequently, understanding the efective dose absorbed by a human body placed in a radibehavior of radiation in complex environments and its ation environment. We designed and implemented a siminteraction with the human body is crucial for accurate ulation framework that combines 3D modeling, human risk assessment and protection. body segmentation, and ray tracing to simulate radiation
In current practice, radiation exposure is typically mon- propagation and interaction within complex scenes. Usitored using personal dosimeters, devices worn on spe- ing this framework, we can analyze radiation exposure cific body parts to record cumulative doses. However, on a per-organ basis, taking into account shielding by this method has limitations: it does not account for the environmental objects and self-shielding by body strucvarying radiosensitivity of diferent organs, the shielding tures. efects of surrounding objects, or the spatial distribution Our methodology provides a step toward more detailed of absorbed dose across the body. Moreover, improper and realistic radiation exposure assessments, with potenusage, calibration issues, and body self-shielding efects tial applications in healthcare facilities and radiological can lead to inaccurate estimations. safety evaluations.
The absorbed dose, expressed in Gray (Gy), measures the energy deposited per unit mass of tissue but does not reflect the biological impact of diferent radiation types. 2. Related works To account for this, the equivalent dose (measured in
of the methods used are based on Monte Carlo (MC)
simulation(s), a method extensively used in medical physics
applications [
However, the ability to model full particle transport ing directly to mesh edges, making it particularly suited
yields high computational complexity which makes MC for tasks such as segmentation and classification on
irsimulations prohibited for daily clinical practice. To over- regular 3D geometries.
come this problem and achieve a fast dose calculation,
several Deep Learning frameworks have been developed,
e.g. [
Although this type of research is interesting, it does not correspond exactly to the objectives of this work.
Indeed, we are more interested in the estimation of the efective dose for radiation workers.
The report [
Indeed, the current trend is to monitor a person’s
exposure doses using devices called dosimeters. They rely
on numerous physical efects and can be of several types
[
Although the use of dosimeters in estimation is very adjacent triangles, extracting local geometric patterns.
widespread, it leads to some problems. First, surveys have Downsampling is achieved via edge collapse operations,
shown that dosimeters are not always properly used [
Finally, it should be noted that all dosimeters register the required number of edges (approximately 2250) for cumulative doses, which may correspond to a high expo- MeshCNN input compatibility. The segmentation results sure over a short period of time or a low exposure over a longer period of time, but these have diferent efects on the body.
html 2https://sketchfab.com/3d-models/ study-human-female-sculpt-854fbf358991477aab518e07556da906 3https://www.blender.org/
This segmentation enables the calculation of organspecific efective doses based on which body parts are impacted by radiation in the simulation.
3D scenes that represent potential radiation exposure environments. For this purpose, we used the trimesh Python library [28], which provides eficient tools for manipulating and rendering triangular meshes.
Each scene is composed of three primary elements: Figure 3 shows examples of two constructed scenes. In the first (a), a single massive pillar stands between the radiation source and the human figure, while the second (b) depicts a more complex setting involving a table assembled from multiple primitives.
Scene complexity can significantly influence radiation propagation, with factors such as object shape, size, material composition, and spatial arrangement playing key roles. Therefore, our framework allows flexible scene generation to evaluate a wide range of shielding scenarios and their efects on radiation dose distribution. (a) (b)
To simulate the propagation of radiation within the constructed scenes, we developed a ray tracing algorithm tailored to radiological modeling. Unlike traditional ray tracing used in computer graphics for visual rendering, our method focuses on modeling energy deposition and attenuation due to interaction with materials.
The radiation source emits a large number of rays uniformly distributed in space. Each ray propagates until it either exits the scene or is absorbed by an object. For • Radiation Source: Modeled as a point source each ray, intersections with scene elements are detected with configurable position and emitted intensity. using trimesh ray-mesh intersection routines. • Human Model: Selected from the segmented If a ray strikes the human mesh, the intersected trimale or female meshes described previously, angle is identified, allowing assignment of the absorbed placed at a user-defined location within the scene. dose to a specific anatomical region. If a ray encounters • Environmental Objects: A configurable set of an object, the radiation intensity is attenuated according objects such as pillars, tables, or shields, each to the material properties before continuing propagation. characterized by its position, size, and material For simplicity, scattering phenomena were neglected in properties (e.g., attenuation coeficients). this implementation. Figure 4 illustrates an example of rays propagating through a scene with a human model and environmental objects. 5.2.1. Distance Decay The intensity decreases with the square of the distance from the source, following the inverse square law: = 0 2 where 0 is the source intensity and is the distance traveled. 5.2.2. Material Attenuation When a ray passes through an object, its intensity is reduced according to the Lambert–Beer law:
= 0 · − where is the linear attenuation coeficient of the material and is the thickness traversed. Attenuation coefifcients were obtained from the NIST database [ 29] for relevant materials.
To compute the efective dose: 1. For each human body part, the energy deposited by rays intersecting that region is accumulated. 2. The absorbed dose (in Gy) is computed based on deposited energy and local mass. 3. A radiation weighting factor ( = 1) is applied, appropriate for gamma and X-ray radiation.
Tissue Bone-marrow (red), colon, lung, stomach, breast
Gonads Bladder, oesophagus,
liver, thyroid Bone surface, brain, salivary glands, skin
Tissue weighting factor
103 [31] are used to adjust contributions from diferent body regions, reflecting their varying radiosensitivity (see Table 1).
skin that are distributed across multiple anatomical parts, weighting adjustments are made based on literaturereported fractional distributions [32, 30, 33] (see Tables 2, 3, and 4).
Thus, the final efective dose is a weighted sum of contributions from all body regions, accurately reflecting both spatial and biological factors influencing radiation risk.
involving diferent human models (male and female), shielding objects, and materials. Radiation propagation and efective dose were computed for each configuration, allowing analysis of shielding efectiveness and the impact of spatial arrangement.
For all experiments, we used three common materials with known linear attenuation coeficients, summarized in Table 5.
in shielding radiation. In these experiments, a cubic object was placed in front of the torso of the human model.
Figures 5a–f visualize the absorbed dose distribution for each configuration.
The numerical results, reported in Tables 6, 7, and 8, 6.2. Self-Shielding and Body Orientation show that: Efects
In all cases, unprotected regions such as the head and arms still absorb considerable radiation, highlighting the importance of whole-body analysis.
• Plexiglass provides limited shielding, resulting in Next, we investigated how body orientation relative high total absorbed doses. to the radiation source afects exposure through self• Particleboard achieves moderate reduction in ab- shielding mechanisms. In one configuration, the human sorbed dose. model faced the source directly; in the other, it was turned • Concrete demonstrates superior shielding, reduc- sideways.
ing the absorbed dose to approximately one-third Figures 6a–d depict the absorbed dose distributions for of that without any shielding. these two configurations. Numerical results are shown in Tables 9 and 10.
As expected, the torso absorbed the highest dose when facing the source. In the side-facing configuration, total absorbed dose was reduced by approximately 50%, demonstrating the significant protective efect of body
Finally, we evaluated scenarios involving more complex shielding, such as a dining table modeled using multiple primitives. Figure 7 illustrates this setup.
The results (Table 11) demonstrate how lower limbs and parts of the torso were partially shielded, significantly altering the absorbed dose distribution compared to the unshielded case.
In this work, we developed a system for estimating effective radiation dose absorbed by a human body placed within a 3D environment containing a radioactive source and various shielding objects. By integrating ray tracing techniques with 3D modeling and human body segmentation via MeshCNN, our method allows for spatially resolved and organ-specific dose calculations, taking into account both distance-based attenuation and materialdependent shielding efects.
Experimental results demonstrate the ability of the system to capture important phenomena such as selfshielding, diferential material absorption, and the influence of complex object geometries on dose distribution. Comparative analyses across diferent shielding materials and body orientations underline the importance of detailed scene modeling for accurate radiation protection assessments. Overall, this work represents a step toward more comprehensive, flexible, and accurate tools for radiation exposure assessment, with potential applications in healthcare worker protection, medical imaging facility design, and radiological emergency response planning.
used ChatGPT, Grammarly in order to: Grammar and spelling check, Paraphrase and reword. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content. 35–42