Radiation dose estimation method, device, equipment, medium and product for human body

Abstract

This application relates to the field of radiation technology, providing a method, apparatus, device, medium, and product for estimating radiation dose to the human body. The method includes: generating an initial human body model based on physiological information; wherein the initial human body model is in a first pose; determining real-time pose data of the human body based on environmental images; adjusting the initial human body model based on the real-time pose data to obtain a target human body model; wherein the target human body model is in a second pose, and the first and second poses are different; and assessing the radiation dose of the target human body model based on accelerated Monte Carlo simulation. This application constructs a personalized initial model based on physiological information and dynamically adjusts it by capturing pose data from environmental images, which can accurately reflect the relationship between the human body and the radiation source in different poses. Furthermore, combined with accelerated Monte Carlo simulation technology, it can achieve dynamic radiation dose monitoring that balances high accuracy and immediacy.

Description

Technical Field

[0001] This application relates to the field of radiation technology, and in particular to methods, apparatus, equipment, media and products for estimating radiation dose to the human body. Background Technology

[0002] Digital subtraction angiography (DSA) is a key tool in interventional procedures, enabling real-time monitoring of a patient's cardiovascular system and guiding devices such as catheters within the body. As an indispensable imaging tool in procedures such as angioplasty and embolization management, DSA exposes patients and healthcare professionals to X-ray radiation, which can lead to short-term tissue damage and long-term carcinogenic risks. Accurate real-time dose estimation is crucial for minimizing radiation damage, but current technologies have significant limitations: 1. Human body models lack specificity and flexibility. Traditional methods often rely on simplified general models or simply search for matches in a pre-existing library, ignoring individual differences in anatomy and resulting in inaccurate basic geometric representations.

[0003] 2. The influence of dynamic posture is ignored. Existing technologies mostly assume that the human body is in a preset static posture, which cannot reflect the actual changes in body position during surgery. This mismatch in geometric relationships significantly amplifies the dosage estimation error.

[0004] 3. Monitoring methods are outdated and computationally inefficient. Clinically used dosimeters can only report cumulative data after the fact, lacking real-time preventative alerts. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in related technologies. To this end, this application proposes a radiation dose estimation method for the human body that can achieve both high accuracy and real-time dynamic radiation dose monitoring.

[0006] This application also proposes radiation dose estimation devices, electronic devices, storage media, and program products for the human body.

[0007] The radiation dose estimation method for the human body according to the first aspect of this application includes: generating an initial human body model based on physiological information of the human body; wherein the initial human body model is in a first pose; determining the real-time pose data of the current human body based on an environmental image; adjusting the initial human body model based on the real-time pose data to obtain a target human body model; wherein the target human body model is in a second pose, and the first pose and the second pose are different; and evaluating the radiation dose of the target human body model based on accelerated Monte Carlo simulation.

[0008] According to the radiation dose estimation method for the human body in this application embodiment, a personalized initial model is constructed based on physiological information, and the model is dynamically adjusted by capturing human posture data in real time with environmental images. This can accurately reflect the relationship between the human body and the radiation source under different actions, and therefore can be adapted to dose estimation under various human postures. At the same time, by combining accelerated Monte Carlo simulation technology, the computational efficiency is significantly improved while ensuring high physical fidelity and accuracy of radiation dose assessment, and dynamic radiation dose monitoring that balances high accuracy and immediacy is achieved.

[0009] According to one embodiment of this application, determining the real-time pose data of the current human body based on an environmental image includes: acquiring an environmental image through a depth vision sensor; and estimating the environmental image using deep learning and optimization algorithms to determine the real-time pose data of the current human body.

[0010] According to one embodiment of this application, before acquiring an environmental image through a depth vision sensor, the method further includes: calibrating the depth vision sensor to determine the transformation relationship between the coordinate system of the depth vision sensor and the coordinate system of the current environment; and adjusting an initial human body model based on real-time pose data to obtain a target human body model, including: adjusting the initial human body model according to the transformation relationship and real-time pose data so that the target human body model in the virtual space can be aligned with the human body pose in the physical space in real time.

[0011] According to one embodiment of this application, the initial human body model includes a joint-level kinematic skeleton; each joint of the kinematic skeleton includes 3 rotational degrees of freedom and 3 translational degrees of freedom; adjusting the initial human body model based on real-time pose data to obtain a target human body model includes: determining the degree of freedom information of each joint in the kinematic skeleton based on real-time pose data; adjusting the position of the initial human body model in virtual space based on real-time pose data, and adjusting the posture of the initial human body model in combination with the degree of freedom information of each joint in the kinematic skeleton to obtain the target human body model.

[0012] According to one embodiment of this application, radiation dose assessment of a human target model is performed based on accelerated Monte Carlo simulation, including: acquiring the operating parameters of an X-ray machine; simulating the propagation of X-rays in a virtual space based on the operating parameters; and, in conjunction with the second pose of the human target model in the virtual space, simulating the interaction between photons and matter using a continuous photon tracking method to determine the radiation dose distribution of X-rays.

[0013] According to one embodiment of this application, generating an initial human body model based on human physiological information includes: acquiring human physiological information; wherein the physiological information includes at least one of gender, age, height, and weight; inputting the physiological information into a parameter regression model to obtain an initial human body model output by the parameter regression model; wherein the initial human body model utilizes a three-dimensional mesh to construct the external contour of the human body, and the interior of the three-dimensional mesh has equivalent material parameters corresponding to the physical properties.

[0014] A radiation dose estimation device for a human body according to a second aspect of this application includes: a human initial model module for generating a human initial model based on physiological information of the human body; wherein the human initial model is in a first pose; a real-time pose data module for determining the real-time pose data of the current human body based on an environmental image; a human target model module for adjusting the human initial model based on the real-time pose data to obtain a human target model; wherein the human target model is in a second pose, and the first pose and the second pose are different; and a radiation dose assessment module for assessing the radiation dose of the human target model based on accelerated Monte Carlo simulation.

[0015] An electronic device according to a third aspect of this application includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the radiation dose estimation methods for the human body described above.

[0016] According to a fourth aspect of this application, a non-transitory computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements any of the radiation dose estimation methods for the human body described above.

[0017] A computer program product according to a fifth aspect of this application includes a computer program that, when executed by a processor, implements any of the radiation dose estimation methods for the human body described above.

[0018] The above-described one or more technical solutions in the embodiments of this application have at least one of the following technical effects: Based on physiological information, a personalized initial model is constructed and dynamically adjusted by capturing pose data from environmental images. This model can accurately reflect the relationship between the human body and the radiation source in different poses. Furthermore, by combining accelerated Monte Carlo simulation technology, dynamic radiation dose monitoring that balances high accuracy and immediacy can be achieved. Furthermore, by using depth vision sensors to acquire environmental images and combining deep learning and optimization algorithms to estimate poses, environmental adaptability and robustness can be enhanced, enabling non-contact, high-precision capture. Furthermore, calibrating the depth vision sensor can achieve spatial consistency in virtual-real fusion, thereby ensuring that the relative position of the X-ray source and the human body model in the virtual space is completely consistent with the relative position of the X-ray machine and the human body in reality.

[0019] Furthermore, the human body model adopts a 6-DOF joint-level motion skeleton, which can adjust the posture by adjusting the joint degrees of freedom, and can simulate the complex joint movements of the real human body. In addition, the model mesh is deformed by driving the skeleton, which enables the corresponding target model to be generated quickly after real-time pose data input, ensuring the real-time response capability of the system. Furthermore, by employing continuous photon tracking, it is possible to accurately simulate the scattering, absorption, and penetration effects of photons at different tissue interfaces in the human body, and accurately calculate the dose distribution map of the human body. Furthermore, the parametric regression model can generate a matching initial human model based on basic physiological parameters. This model uses a three-dimensional mesh as its appearance and is filled with equivalent materials, which forms the basis for radiation dose calculation. It can calculate the effective biological dose based on the radiation sensitivity of different tissues.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the radiation dose estimation method for the human body provided in the embodiments of this application.

[0023] Figure 2 This is a schematic diagram of the initial human body model provided in the embodiments of this application.

[0024] Figure 3 This is a schematic diagram of the overall framework of the radiation dose estimation method for the human body provided in the embodiments of this application.

[0025] Figure 4 This is a schematic diagram of the accelerated Monte Carlo simulation module provided in the embodiments of this application.

[0026] Figure 5 This is a schematic diagram of the structure of the radiation dose estimation device for the human body provided in the embodiments of this application.

[0027] Figure 6 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0028] The embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but should not be used to limit the scope of this application.

[0029] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0030] It should be noted that all actions involving the acquisition of signals, information, or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization granted by the owner of the relevant device.

[0031] Among related technologies, dose estimation and monitoring techniques have the following significant limitations in practical clinical applications: First, the construction of human body models lacks specificity and flexibility. Related technologies often rely on overly simplified human body models or generate approximate models simply by searching for matches in a pre-stored model library. This approach ignores individual anatomical differences, fails to generate a "digital twin" that truly reflects the patient's characteristics, and results in inherent errors in the basic geometric representation.

[0032] Second, the impact of dynamic changes in human posture on dosage is ignored. Related methods typically assume that the human body remains absolutely still in a pre-set bed position, or use only static postures for calculation. This is significantly inconsistent with the possible postural adjustments that patients or doctors may make during actual surgery. This geometric mismatch significantly amplifies the estimation errors of surface dose and organ radiation dose.

[0033] Third, monitoring methods lack real-time preventative measures. Clinicians primarily rely on wearable dosimeters to monitor radiation, but these devices typically only report localized (e.g., chest) cumulative radiation doses after the fact. This passive, post-hoc monitoring method cannot provide real-time spatial dose distribution information, nor can it issue immediate preventative alerts, making it difficult to effectively guide medical staff to proactively avoid high-radiation areas during surgery.

[0034] Fourth, high-precision algorithms struggle to meet real-time requirements. While traditional Monte Carlo simulations based on voxel representation and photon tracking (such as GEANT4) can provide physically accurate dose estimations, they are computationally intensive and extremely time-consuming. This bottleneck in computational efficiency prevents high-precision physical simulations from being applied to time-sensitive real-time clinical monitoring scenarios.

[0035] It can be observed that achieving accurate dose estimation requires three key elements: a precise geometric representation of the patient's body, posture tracking capable of simulating the geometric relationship between the human body and the radiation source, and radiation interaction calculations based on physical simulations. Therefore, this application provides a radiation dose estimation scheme that combines personalized human body modeling, real-time dynamic posture tracking, and efficient acceleration of physical simulations to overcome the dual deficiencies of existing technologies in terms of accuracy and real-time performance.

[0036] Please see Figure 1 , Figure 1 This is a flowchart illustrating the radiation dose estimation method for the human body provided in this application embodiment. The radiation dose estimation method for the human body may include steps S110 to S140, each step of which is detailed below: S110: Generate an initial human body model based on the physiological information of the human body; wherein, the initial human body model is in the first pose.

[0037] S120: Determine the real-time pose data of the current human body based on the environmental image.

[0038] S130: Adjust the initial human body model based on real-time pose data to obtain the target human body model; wherein, the target human body model is in the second pose, and the first pose and the second pose are different.

[0039] S140: Radiation dose assessment of human target models based on accelerated Monte Carlo simulation.

[0040] This application proposes a dynamic and personalized radiation dose estimation method. By combining human physiological information, real-time pose capture, and accelerated Monte Carlo simulation, it achieves high-precision estimation of radiation dose received by the human body in complex environments, such as medical radiotherapy, nuclear industry, and space radiation scenarios. The core of this embodiment lies in fusing a static human body model with dynamic pose data, and then using an efficient simulation algorithm for dose assessment, significantly improving the accuracy and practicality of the estimation.

[0041] In step S110, physiological information may include a person's height, weight, age, gender, and body fat percentage. Based on this information, a virtual 3D model, i.e., the initial human body model, can be constructed that closely approximates the subject in terms of body shape and weight. The initial human body model provides a personalized anatomical basis for subsequent dose simulation, ensuring that the model reflects individual differences and thus improving the specificity of dose estimation. The initial human body model is in the first position, which is an initial standard posture, such as lying flat, but not the subject's current actual posture.

[0042] This embodiment generates an initial model based on physiological information, which means that the model is not a general dummy, but is constructed for specific patient characteristics, further improving the estimation reference value for specific individuals.

[0043] In step S120, computer vision technology can be used to process the images captured by the environmental camera. This step not only requires capturing environmental images, but also requires extracting key skeletal points or posture parameters of the human body from the images using algorithms, such as the degree to which the elbow is bent or the degree to which the body is tilted to the left.

[0044] In step S130, based on the real-time pose data obtained in the above steps, the initial human body model can be driven to transform from the first pose to the second pose, thereby obtaining the target human body model. The second pose is the subject's current actual posture.

[0045] It should be noted that during the adjustment process, it is necessary to maintain the topological consistency of the human anatomy structure and ensure that the position and shape of organs and tissues are reasonably deformed with changes in posture, such as the compression of the lungs when bending over and the stretching of muscles when rotating limbs, so as to generate a human target model that matches the real-time posture of the examinee.

[0046] Optionally, to achieve accurate simulation of the human body's realistic posture in virtual space, skeletal-driven mesh deformation technology can be used to adapt the initial human body model to the desired posture. Through methods such as twisting, rotating, and local adjustments, it can be made into a target human body model that perfectly matches the posture of the examinee in reality. Based on this, a digital avatar that perfectly matches the examinee's body shape and posture can be constructed in the virtual environment.

[0047] For example, the use of skeleton-driven mesh deformation technology may specifically include the following steps: The process involves acquiring a 3D mesh corresponding to the initial human body model and a pre-defined hierarchical skeleton. The 3D mesh can be a multi-layered nested mesh of the human body's surface skin mesh and internal organs. Each vertex of the 3D mesh can be configured with a skin weight corresponding to the hierarchical skeleton. Based on real-time acquired environmental images, the current spatial joint feature points of the subject are extracted, and the motion transformation matrix of each bone node in the hierarchical skeleton is calculated based on the current spatial joint feature points. Finally, based on the motion transformation matrix and skin weights, the target spatial coordinates of each vertex in the 3D mesh are calculated to drive the initial human body model to transform from its initial pose to the current human target model of the subject's true pose.

[0048] In radiation dose estimation under moving conditions, the human body model can be simplified as a rigid object, and basic positioning data can be obtained by tracking its overall spatial translation and rotation. However, this embodiment can further adjust the human body model to the subject's current real posture in real time by using skeleton-driven mesh deformation technology. This not only simulates the spatial movement of the surface, but also accurately reproduces non-rigid deformations, including torsion, stretching, and compression.

[0049] In real-world medical settings, human movements often involve complex muscle stretching and trunk twisting, rather than simple mechanical translation. The skeletal-driven mesh deformation method used in this embodiment can achieve a high degree of consistency between the initial human model and the real-time posture of the subject, providing reliable geometric and posture data support for subsequent accurate instantaneous radiation dose estimation.

[0050] Unlike estimation methods based on static models in related technologies, this embodiment captures the real-time dynamic pose of the human body through environmental images. This allows the radiation dose calculation to be based on the actual irradiated posture of the human body, avoiding errors in the irradiated area and angle of organs caused by postural differences (such as bending over or turning to the side), thereby greatly improving the realism and accuracy of the estimation.

[0051] In step S140, Monte Carlo simulation is a probabilistic statistical numerical method that calculates the radiation energy deposition distribution by simulating the random transport process of a large number of particles (such as photons, neutrons, and electrons) in a medium, thereby assessing the dose. Traditional Monte Carlo simulations (such as MCNP and Geant4) are highly accurate but computationally intensive. This embodiment employs accelerated Monte Carlo simulation, which can significantly improve the simulation speed and achieve near real-time dose assessment, thereby determining the radiation dose received by various organs of the human body in a specific current posture.

[0052] While the Monte Carlo method of related technologies has high accuracy, it is computationally intensive. This embodiment introduces an acceleration mechanism to ensure high-precision physical simulation while meeting the time efficiency requirements of clinical or real-world scenarios, and has the potential for real-time or near-real-time estimation.

[0053] Some methods typically rely on preset, generic human features or simple geometry for matching and coarse scattering calculations, failing to reflect the differences in body shape and tissue distribution among individual patients. This embodiment, however, uses a parametric model to generate a patient-specific initial human model, accurately reproducing the patient's true anatomical structure and tissue density characteristics. This specific model not only improves the fit of pose tracking but also, in subsequent dose calculations, ensures that the assessment of radiation attenuation and scattering within the body is no longer based on empirical formulas or static tables but on the patient's actual physical attributes, thereby enhancing the personalized accuracy of radiation dose estimation.

[0054] In interventional surgeries or complex imaging scenarios, patients' postures often undergo dynamic changes (such as chest and abdominal movements due to breathing, irregular limb movements, etc.). While related technologies may employ real-time pose monitoring, they are prone to errors when handling complex deformations. This embodiment combines an initial human model carrying patient-specific parameters with real-time posture tracking depth, enabling accurate capture of both rigid displacements and real-time tracking of local non-rigid deformations. This tracking method based on the initial human model provides precise dynamic physical boundary conditions for subsequent dose calculations, ensuring that the patient's true spatial state in the radiation field is updated in real-time and accurately through the human target model whenever any posture change occurs.

[0055] Monte Carlo simulations consider complex particle interactions such as the photoelectric effect and Compton scattering, achieving extremely high physical accuracy. However, they are computationally very time-consuming and cannot be used for real-time monitoring. While lookup methods based on empirical formulas or tables are fast, they lack accuracy when dealing with complex non-uniform media (such as the specific internal structure of a patient's body) and multiple scattering environments. This embodiment, however, obtains high-precision real-time patient status through a human target model, introducing accelerated Monte Carlo simulations. While maintaining the physical-level accuracy of particle transport simulations, it utilizes a pre-constructed parameterized model and real-time pose data to limit the computational area, significantly reducing invalid particle tracking and thus significantly improving computational efficiency.

[0056] In summary, this application embodiment constructs a personalized initial model based on physiological information and dynamically adjusts the model by capturing human posture data in real time using environmental images. This accurately reflects the relationship between the human body and the radiation source under different actions, thus adapting to dose estimation under various human postures. At the same time, by combining accelerated Monte Carlo simulation technology, the computational efficiency is significantly improved while ensuring high physical fidelity and accuracy in radiation dose assessment, achieving dynamic radiation dose monitoring that balances high accuracy and immediacy.

[0057] It is important to emphasize that the aforementioned individualized parametric model and real-time, accurate patient tracking are the key foundations for achieving the accelerated Monte Carlo simulation described in this application. It is precisely this accurate combination of individualized modeling and dynamic tracking that allows Monte Carlo simulation to avoid blind calculations across the entire space, instead focusing on targeted simulation acceleration for key areas of change. This deep integration of the three effectively overcomes the technical limitations of traditional dose estimation, which struggles to simultaneously achieve both high precision and high real-time performance. In real-world medical scenarios, operators can instantly adjust their operating positions and protective strategies based on the real-time updated, high-precision dose distribution results.

[0058] In some embodiments, the step of generating an initial human model based on human physiological information may specifically include: Obtain physiological information of the human body; wherein, the physiological information includes at least one of gender, age, height and weight; input the physiological information into a parametric regression model to obtain an initial human body model output by the parametric regression model; wherein, the initial human body model uses a three-dimensional mesh to construct the external contour of the human body, and the interior of the three-dimensional mesh has equivalent material parameters with corresponding physical properties.

[0059] In this embodiment, the input physiological information may include at least one of gender, age, height, and weight. These parameters are statistically significant descriptors of human body shape and are relatively easy to obtain.

[0060] Please see Figure 2 , Figure 2 This is a schematic diagram of the initial human body model provided in the embodiments of this application. It can be seen that there are significant differences in body shape between children, thin adults, and heavier adults.

[0061] After obtaining the above physiological information, an initial human body model can be constructed using a parametric regression model.

[0062] Among them, the parametric regression model is a machine learning model that can establish a mapping relationship from physiological parameters to human geometry and material parameters through training data. The initial human body model uses a three-dimensional mesh to construct the external contour of the human body, and the interior of the three-dimensional mesh has equivalent material parameters corresponding to the physical properties.

[0063] Specifically, the three-dimensional mesh defines the boundary between the human body and the surrounding air, as well as the geometric surface on which rays enter the human body. Furthermore, the interior of the three-dimensional mesh includes equivalent material parameters, which are crucial for radiation physics simulations: in accelerated Monte Carlo simulations, it is necessary to determine what substance the photons are hitting; therefore, the human body model in this embodiment needs to map its geometric volume to physical density and elemental composition.

[0064] For example, the initial human body model is predefined with different anatomical regions (such as the skeletal region, soft tissue region, and lung region), each corresponding to different equivalent material parameters. For instance, the skeletal region is set with a higher density, the soft tissue region with a medium density, and the lung region with a lower density.

[0065] In this embodiment, a parametric regression model is used to generate an initial human body model with a three-dimensional mesh and equivalent material parameters. The parametric regression model can generate a matching initial human body model based on basic physiological parameters. This model uses a three-dimensional mesh as its appearance and is filled with equivalent materials, which is the basis for radiation dose calculation. The effective biological dose can be calculated based on the radiation sensitivity of different tissues.

[0066] In some embodiments, the step of determining the real-time pose data of the current human body based on the environmental image may specifically include: The system acquires environmental images using a depth vision sensor; it then uses deep learning and optimization algorithms to estimate the environmental images and determine the real-time pose data of the human body.

[0067] In this embodiment, environmental images can be acquired using a depth vision sensor, which can be an RGBD camera.

[0068] Unlike ordinary cameras that can only acquire two-dimensional planar images, the depth vision sensor in this embodiment can additionally acquire depth maps or point cloud data. Optionally, the depth vision sensor can be a ToF camera, a structured light camera, or a binocular stereo camera. The environmental images acquired by the depth vision sensor can provide additional geometric information in three-dimensional space.

[0069] In addition, this embodiment also employs deep learning and optimization algorithms, which can extract human features from complex environmental backgrounds. Even under conditions of uneven lighting or partial occlusion, it can infer the spatial coordinates of human joints through a data-driven approach.

[0070] In this embodiment, a depth vision sensor is used to acquire environmental images. The depth vision sensor can acquire not only color information but also depth (distance) information, which enables accurate capture of human contours and postures even under changing lighting conditions or complex backgrounds, thus enhancing environmental adaptability and robustness. Furthermore, by combining deep learning and optimization algorithms to estimate poses, and utilizing the powerful feature extraction capabilities of deep learning algorithms, complex human skeletons and posture data can be efficiently and automatically parsed from images without the need for sensors to be worn on patients, reducing interference with medical procedures, improving user experience, and achieving non-contact, high-precision capture.

[0071] In some embodiments, the steps prior to acquiring environmental images via a depth vision sensor may further include: The depth vision sensor is calibrated to determine the transformation relationship between the coordinate system of the depth vision sensor and the coordinate system of the current environment; the initial human body model is adjusted based on real-time pose data to obtain the target human body model, including: adjusting the initial human body model according to the transformation relationship and real-time pose data so that the target human body model in the virtual space can be aligned with the human body pose in the physical space in real time.

[0072] This embodiment also includes the calibration of the depth vision sensor, which determines the transformation relationship between the coordinate system of the virtual space and the coordinate system of the current real physical space. Calibration ensures that the human body model in the virtual world not only has the correct posture, but also that its absolute position in three-dimensional space completely coincides with that of the real patient.

[0073] Alternatively, the transformation relationship can be represented in the form of a transformation matrix.

[0074] The above calibration of the depth vision sensor enables spatial consistency in virtual-real fusion, ensuring that the relative positions of the radiation source and the human body model in the virtual space are completely consistent with the relative positions of the X-ray machine and the human body in reality. This embodiment is a crucial step in ensuring the accuracy of radiation dose estimation values, as it guarantees the precision of spatial positioning and directly determines the physical validity of the final dose calculation.

[0075] For example, the RGBD camera is calibrated using a reference target placed at a known location on the patient table. This aims to calculate the camera's intrinsic parameters and its extrinsic parameters relative to the DSA world coordinate system, thereby mapping the measurement data to the DSA physical coordinate system. A key step in this process is ground plane registration: by performing planar fitting on the point cloud of the empty table surface and combining it with known height readings from the DSA control system, a precise correlation is established between the depth sensor coordinate system and the DSA physical space.

[0076] At the start of pose estimation, synchronized RGB and depth data of the patient while supine are acquired. In the preprocessing stage, the background is removed using known examination table geometry priors, and the patient target is segmented from the point cloud. Subsequently, a pre-trained neural network is deployed on the RGB image to predict the rotation and translation parameters of the model joints, generating initial pose hypotheses. Based on this, an iterative optimization algorithm is used to fine-tune the pose parameters, minimizing the error between the generated model surface and the measured 3D point cloud. Furthermore, this embodiment also features dynamic tracking capabilities; once a change in posture is detected, a re-estimation process is triggered, ensuring continuous alignment of the patient's pose with the DSA physical space throughout the process.

[0077] In some embodiments, the initial human body model includes a joint-level kinematic skeleton; each joint of the kinematic skeleton includes 3 rotational degrees of freedom and 3 translational degrees of freedom; the step of adjusting the initial human body model based on real-time pose data to obtain the target human body model may specifically include: Based on real-time pose data, the degrees of freedom of each joint in the motion skeleton are determined; the position of the initial human body model in virtual space is adjusted based on the real-time pose data, and the posture of the initial human body model is adjusted by combining the degrees of freedom of each joint in the motion skeleton to obtain the target human body model.

[0078] In this embodiment, the initial human body model includes a joint-level kinematic skeleton. This is not simply a collection of scattered bones, but a kinetic chain system with parent-child relationships. This hierarchical structure ensures the anatomical rationality of human movement and avoids deformations that violate human physiology when the model adjusts its posture.

[0079] For example, the upper arm is the parent node of the forearm, and the forearm is the parent node of the hand. When the upper arm moves, the forearm and hand will follow suit, but the independent movement of the forearm will not affect the upper arm.

[0080] Each joint has 3 rotational degrees of freedom (Rx, Ry, Rz) and 3 translational degrees of freedom (Tx, Ty, Tz).

[0081] Optionally, rotation can be used to match postures. Rotation can correspond to flexion, extension, abduction, adduction, internal and external rotation of human joints, such as turning the head or bending the elbow.

[0082] Optionally, translation can be used to match body proportions. Translation is one of the core aspects of this embodiment. In the real human body, bone length is fixed, and joints generally only rotate. However, in the parametric modeling of this embodiment, the introduction of translational degrees of freedom can accommodate different body shapes, i.e., the stretching and contraction of bone length, and simulate complex joint sliding, such as the sliding of the scapula on the ribcage, or the minute displacement of the knee joint.

[0083] Therefore, this embodiment can convert real-time pose data input from the outside into degrees of freedom parameters inside the skeleton, which can not only adjust the overall position of the human body model, but also adjust the local state of each joint in the human body model, i.e., the human posture.

[0084] The human body model uses a 6-DOF joint-level motion skeleton, which can adjust the posture by adjusting the joint degrees of freedom, and can simulate the complex joint movements of the real human body. In addition, the model mesh is deformed by driving the skeleton, which enables the corresponding target model to be generated quickly after real-time pose data is input, ensuring the real-time response capability of the system.

[0085] Human models, including target models and initial models, are equipped with a kinematic skeleton that defines the joint hierarchy. Each joint represents rotation with three degrees of freedom (DoF) and translation with another three, thus achieving a fully parametric description of body motion. This parametric design allows the model's pose to be updated by optimizing joint parameters. Pose changes propagate along the kinematic chain, driving deformation of the relevant surface meshes to form a new body configuration. By adjusting these pose parameters, the model can be aligned with the actual patient pose observed by an RGBD camera.

[0086] In some embodiments, the step of assessing radiation dose to a human target model based on accelerated Monte Carlo simulation may specifically include: The X-ray machine's operating parameters are obtained; based on these parameters, the propagation of X-rays is simulated in a virtual space; combined with the second pose of the human target model in the virtual space, a continuous photon tracking method is used to simulate the interaction between photons and matter, in order to determine the radiation dose distribution of X-rays.

[0087] In the above embodiment, an accurate virtual patient has been constructed through personalized human body modeling and real-time dynamic posture tracking. Therefore, the main purpose of this embodiment is to conduct a real X-ray irradiation in the virtual world.

[0088] The purpose of obtaining the operating parameters of an X-ray machine is to digitize the physical characteristics of a real-world X-ray machine. Optionally, operating parameters may include tube voltage, the product of tube current and exposure time, source-image distance, filter material and thickness, and collimator settings. Using these parameters, a virtual X-ray source that is completely identical to the real equipment can be reconstructed in a computer.

[0089] Since the human target model has been adjusted to the actual posture of bending over, turning to the side, or raising an arm based on the real-time environmental images, the anatomical path that the virtual ray takes when passing through the human target model (e.g., passing through the arm bones first and then through the lungs) will be completely consistent with reality. This is a prerequisite for ensuring accurate dose distribution.

[0090] Finally, accelerated Monte Carlo simulation was employed as the computational engine to obtain macroscopic results by simulating the random behavior of a massive number of individual photons. It should be noted that continuous photon tracking is a specific algorithmic strategy. Unlike voxel-based discrete stepping, continuous tracking allows photons to move within the geometry at arbitrary step sizes, accurately calculating the intersection points of photons and matter boundaries.

[0091] For example, when simulating the operation of an X-ray machine based on its working parameters, virtual photons are first emitted. Each photon is assigned a specific energy value that conforms to the actual X-ray energy spectrum distribution and a flight direction that conforms to the collimator's field of view. Subsequently, the photons fly in a straight line in virtual space, and the intersecting trajectory with the human target model in a second pose (including the three-dimensional mesh boundaries of tissues and organs such as skin, fat, bones, and organs) is calculated in real time.

[0092] When a photon enters the human body model and encounters an equivalent material with corresponding physical properties, the algorithm determines the collision type based on a physical probability model (based on cross-sectional data): if the photoelectric effect occurs, the photon is completely absorbed, and all its energy is deposited at that point and converted into a dose; if Compton scattering occurs, the photon loses some energy and changes its flight direction, producing scattered rays that may irradiate other parts of the body; if it is penetrating, the photon does not have any effect and passes directly out of the human body.

[0093] Once energy deposition occurs, the absorbed energy can be recorded at the corresponding coordinates. If scattering occurs, the photon will continue to fly with new energy and direction, repeating the above "propagation-interaction" process until the photon energy is exhausted or it flies out of the human body's boundaries.

[0094] After simulating a sufficient number of photons (e.g., millions), all recorded energy deposition data can be statistically summarized, and the absorbed dose can be calculated by dividing by the tissue mass at the corresponding location. Ultimately, a three-dimensional, color radiation dose distribution map is generated, visually reflecting the radiation exposure of various internal organs of the human body in the current posture.

[0095] Please see Figure 3 , Figure 3 This is a schematic diagram of the overall framework of the radiation dose estimation method for the human body provided in the embodiments of this application.

[0096] This embodiment provides a comprehensive framework for real-time patient dose monitoring. This framework consists of three main components: 1) Patient-Specific Geometric Model Generator: Taking biological parameters (sex, age, weight, height) as input, it generates an anatomically accurate standard patient model (i.e., the initial human model) through a parametric regression model. The parametric regression model outputs a 3D mesh representing the unique external anatomical structure of the patient, providing a personalized representation basis for downstream simulations.

[0097] Optionally, the parametric regression model can be trained on a human mesh model library to generate a personalized patient mesh model in a standard position and posture based on input physiological parameters.

[0098] 2) Real-time posture tracking system: This system uses RGBD sensors to acquire real-time observation data (i.e., environmental images) of the surgical environment, and processes this data through a deep neural network to estimate the patient's dynamic posture (including joint positions and angles). The real-time posture tracking system can dynamically align the patient's digital twin with the coordinates of the X-ray source in the simulated space, generating a posture-aligned patient model (i.e., the aforementioned human target model), thereby achieving continuous spatial registration during the surgical procedure.

[0099] Optionally, an RGBD camera captures the patient's visual and depth information in the surgical environment in real time. This data is processed using deep learning and optimization algorithms to estimate the patient's real-time pose parameters. Subsequently, these parameters are input into a pose regressor to dynamically adjust the position and posture of the patient model to match the observed reality, thereby achieving high-precision registration between the digital model and the physical space.

[0100] 3) Accelerated Monte Carlo Simulation Module: This module simulates the propagation of X-ray photons with physical precision, modeling key photon-matter interactions (such as photoelectric absorption and Compton scattering). Unlike grid-based methods in related technologies, the Accelerated Monte Carlo Simulation Module employs continuous photon tracing technology, decoupling computational complexity from spatial resolution limitations, thereby significantly improving computational efficiency while maintaining accuracy.

[0101] Please see Figure 4 , Figure 4 This is a schematic diagram of the accelerated Monte Carlo simulation module provided in the embodiments of this application.

[0102] The accelerated Monte Carlo dose simulation module employs a system kinematics model, a beam simulator, a grid-ray intersection tester, and a physics-based photon tracing algorithm. These functional modules / algorithms work collaboratively, decoupling computational complexity and spatial resolution through continuous photon tracing. This achieves real-time dose estimation while maintaining physical accuracy, ultimately outputting a three-dimensional dose distribution map of the patient in a specific posture. The specific details of each functional module / algorithm are as follows: 1. System kinematic model: replicate the forward kinematics of an X-ray machine or X-ray imaging system (such as a floor-standing or ceiling-mounted C-arm), and dynamically update the position and orientation of the light source based on the system motion parameters.

[0103] 2. Beam Simulator: Generates virtual light rays (e.g., cone beams) that conform to the actual optical path, with photon energy distribution following the output of a spectral simulator based on parameters such as tube voltage and filter type.

[0104] 3. Mesh-ray intersection tester: efficiently detects the interaction position between photons and the patient mesh model (i.e., the human target model mentioned above).

[0105] 4. Physics-based photon tracking algorithm: Simulates key interactions between photons and tissues (such as photoelectric effect and Compton scattering), records energy deposition, and calculates dose by statistically analyzing a large number of photon trajectories.

[0106] The patient mesh model is uniformly filled with tissue-equivalent material possessing relevant physical properties. A mesh-ray intersection tester is used to determine the interaction points between tentative photon paths and the mesh model. Subsequently, a physics-based photon tracing algorithm determines whether virtual photons interact with the material and pinpoints their locations. If the sampling is a scattering event, the photon's direction and energy are adjusted accordingly based on the governing physics formulas. Escaping or absorbed photons trigger the beam simulator to generate new photons, while scattered photons continue to be simulated through iterative state updates until a termination condition is met.

[0107] In terms of dose calculation, the patient surface dose is determined by the X-ray air kerma reaching the grid surface; the internal dose is the energy absorbed within the patient grid model; and the environmental dose is estimated based on the environmental dose equivalent defined by the International Committee on Radiation Units and Measurements (ICRU). The final estimate can be further corrected by empirical measurements.

[0108] This embodiment integrates personalized modeling, real-time posture tracking, and efficient physical simulation to construct a comprehensive system suitable for intraoperative real-time dose monitoring, specifically including: In terms of personalized patient modeling, a parametric model is used to receive the patient's physiological information and generate a mesh-based, patient-specific geometric model. This model is initialized in a standard position and pose.

[0109] To bridge the gap between digital models and physical reality, an RGBD camera and a deep learning model are integrated: the camera captures real-time observation data of the surgical environment, and the deep learning model estimates the patient's dynamic pose based on this data, adjusting the position and pose of the patient model in real time. This process effectively improves spatial registration accuracy, thereby further reducing dose estimation errors.

[0110] In dose calculation, a photon tracing method based on ray-grid intersection is employed, combined with empirical measurement data and Monte Carlo simulations, to achieve physically accurate modeling of X-ray photon transport and key interactions (such as photoelectric absorption and Compton scattering). This design supports continuous photon tracing, decoupling computational complexity from spatial resolution, thereby enabling real-time and comprehensive dose estimation while maintaining accuracy.

[0111] Ultimately, this embodiment can generate a three-dimensional radiation dose distribution map reflecting the irradiation status of internal organs under the patient's current posture, providing visualization and quantitative support for intraoperative radiation safety monitoring.

[0112] On the other hand, embodiments of this application also provide a radiation dose estimation device for the human body. The radiation dose estimation device for the human body provided in this application will be described below. The radiation dose estimation device for the human body described below can be referred to in correspondence with the radiation dose estimation method for the human body described above.

[0113] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a radiation dose estimation device for the human body provided in an embodiment of this application. In this embodiment, the radiation dose estimation device for the human body may include: a human body initial model module 510, a real-time pose data module 520, a human body target model module 530, and a radiation dose assessment module 540.

[0114] The initial human body model module 510 is used to generate an initial human body model based on the physiological information of the human body; wherein, the initial human body model is in the first pose.

[0115] The real-time pose data module 520 is used to determine the real-time pose data of the human body based on the environmental image.

[0116] The human target model module 530 is used to adjust the initial human model based on real-time pose data to obtain the human target model; wherein the human target model is in the second pose, and the first pose and the second pose are different.

[0117] Radiation dose assessment module 540 is used to assess radiation dose to a human target model based on accelerated Monte Carlo simulation.

[0118] In summary, this application embodiment constructs a personalized initial model based on physiological information and dynamically adjusts the model by capturing human posture data in real time using environmental images. This accurately reflects the relationship between the human body and the radiation source under different actions, thus adapting to dose estimation under various human postures. At the same time, by combining accelerated Monte Carlo simulation technology, the computational efficiency is significantly improved while ensuring high physical fidelity and accuracy in radiation dose assessment, achieving dynamic radiation dose monitoring that balances high accuracy and immediacy.

[0119] In some embodiments, the real-time pose data module 520 is specifically used to: acquire environmental images through a depth vision sensor; estimate the environmental images through deep learning and optimization algorithms to determine the current real-time pose data of the human body.

[0120] In some embodiments, the real-time pose data module 520 is further configured to: calibrate the depth vision sensor to determine the transformation relationship between the coordinate system of the depth vision sensor and the coordinate system of the current environment; and adjust the initial human body model based on the real-time pose data to obtain the target human body model, including: adjusting the initial human body model according to the transformation relationship and the real-time pose data so that the target human body model in the virtual space can be aligned with the human body pose in the physical space in real time.

[0121] In some embodiments, the initial human body model includes a joint-level kinematic skeleton; each joint of the kinematic skeleton includes 3 rotational degrees of freedom and 3 translational degrees of freedom; the human target model module 530 is specifically used to: determine the degree of freedom information of each joint in the kinematic skeleton based on real-time pose data; adjust the position of the initial human body model in virtual space based on real-time pose data, and adjust the posture of the initial human body model in combination with the degree of freedom information of each joint in the kinematic skeleton to obtain the human target model.

[0122] In some embodiments, the radiation dose assessment module 540 is specifically used to: acquire the operating parameters of the X-ray machine; simulate the propagation of X-rays in a virtual space based on the operating parameters; and, in conjunction with the second pose of the human target model in the virtual space, use a continuous photon tracking method to simulate the interaction between photons and matter in order to determine the radiation dose distribution of X-rays.

[0123] In some embodiments, the initial human body model module 510 is specifically used to: acquire physiological information of the human body; wherein the physiological information includes at least one of gender, age, height and weight; input the physiological information into a parameter regression model to obtain an initial human body model output by the parameter regression model; wherein the initial human body model uses a three-dimensional mesh to construct the external contour of the human body, and the interior of the three-dimensional mesh has equivalent material parameters corresponding to the physical properties.

[0124] Furthermore, embodiments of this application also provide an electronic device. Please refer to [link / reference]. Figure 6 , Figure 6This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of this application. The electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute the following method: An initial human body model is generated based on the physiological information of the human body; the initial human body model is in the first pose; the real-time pose data of the current human body is determined based on the environmental image; the initial human body model is adjusted based on the real-time pose data to obtain the target human body model; the target human body model is in the second pose, and the first pose and the second pose are different; the radiation dose of the target human body model is assessed based on accelerated Monte Carlo simulation.

[0125] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0126] In another aspect, embodiments of this application disclose a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the computer program instructions are executed by a computer, the computer is able to perform the methods provided in the above-described method embodiments, such as including: An initial human body model is generated based on the physiological information of the human body; the initial human body model is in the first pose; the real-time pose data of the current human body is determined based on the environmental image; the initial human body model is adjusted based on the real-time pose data to obtain the target human body model; the target human body model is in the second pose, and the first pose and the second pose are different; the radiation dose of the target human body model is assessed based on accelerated Monte Carlo simulation.

[0127] In another aspect, embodiments of this application also provide a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments, including, for example: An initial human body model is generated based on the physiological information of the human body; the initial human body model is in the first pose; the real-time pose data of the current human body is determined based on the environmental image; the initial human body model is adjusted based on the real-time pose data to obtain the target human body model; the target human body model is in the second pose, and the first pose and the second pose are different; the radiation dose of the target human body model is assessed based on accelerated Monte Carlo simulation.

[0128] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0129] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of software products. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate this application and are not intended to limit this application. Although this application has been described in detail with reference to the embodiments, those skilled in the art should understand that various combinations, modifications, or equivalent substitutions of the technical solutions of this application do not depart from the spirit and scope of the technical solutions of this application and should be covered within the scope of the claims of this application.

Claims

1. A method for estimating radiation dose to the human body, characterized in that, include: An initial human body model is generated based on the physiological information of the human body; wherein, the initial human body model is in the first pose; Determine the real-time pose data of the human body based on environmental images; The initial human body model is adjusted based on the real-time pose data to obtain a target human body model; wherein the target human body model is in a second pose, and the first pose and the second pose are different. Radiation dose assessment was performed on the human target model based on accelerated Monte Carlo simulation.

2. The radiation dose estimation method for the human body according to claim 1, characterized in that, The step of determining the real-time pose data of the human body based on the environmental image includes: The environmental image is acquired using a depth vision sensor; The environmental image is estimated using deep learning and optimization algorithms to determine the real-time pose data of the current human body.

3. The radiation dose estimation method for the human body according to claim 2, characterized in that, Before acquiring the environmental image via a depth vision sensor, the method further includes: The depth vision sensor is calibrated to determine the transformation relationship between the coordinate system of the depth vision sensor and the coordinate system of the current environment; The step of adjusting the initial human body model based on the real-time pose data to obtain the target human body model includes: The initial human body model is adjusted according to the transformation relationship and the real-time pose data, so that the human body target model in the virtual space can be aligned with the human body pose in the physical space in real time.

4. The radiation dose estimation method for the human body according to claim 1, characterized in that, The initial human body model includes a joint-level kinematic skeleton; each joint of the kinematic skeleton includes 3 rotational degrees of freedom and 3 translational degrees of freedom. The step of adjusting the initial human body model based on the real-time pose data to obtain the target human body model includes: Based on the real-time pose data, the degree of freedom information of each joint in the motion skeleton is determined; Based on the real-time pose data, the position of the initial human body model in the virtual space is adjusted, and combined with the degree of freedom information of each joint in the motion skeleton, the posture of the initial human body model is adjusted to obtain the target human body model.

5. The method for estimating radiation dose to the human body according to any one of claims 1 to 4, characterized in that, The radiation dose assessment of the human target model based on accelerated Monte Carlo simulation includes: Obtain the operating parameters of the X-ray machine; Based on the aforementioned operating parameters, the propagation of X-rays is simulated in virtual space; By combining the second pose of the human target model in the virtual space, a continuous photon tracking method is used to simulate the interaction between photons and matter in order to determine the radiation dose distribution of X-rays.

6. The method for estimating radiation dose to the human body according to any one of claims 1 to 4, characterized in that, The process of generating an initial human model based on human physiological information includes: Obtain the physiological information of the human body; wherein the physiological information includes at least one of gender, age, height and weight; The physiological information is input into a parameter regression model to obtain an initial human body model output by the parameter regression model; wherein, the initial human body model uses a three-dimensional mesh to construct the external contour of the human body, and the interior of the three-dimensional mesh has equivalent material parameters corresponding to physical properties.

7. A radiation dose estimation device for the human body, characterized in that, include: The initial human body model module is used to generate an initial human body model based on the physiological information of the human body; wherein, the initial human body model is in the first pose; The real-time pose data module is used to determine the current real-time pose data of the human body based on the environmental image. The human target model module is used to adjust the initial human model based on the real-time pose data to obtain a human target model; wherein the human target model is in a second pose, and the first pose and the second pose are different; The radiation dose assessment module is used to assess the radiation dose of the human target model based on accelerated Monte Carlo simulation.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the radiation dose estimation method for the human body as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the radiation dose estimation method for the human body as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the radiation dose estimation method for the human body as described in any one of claims 1 to 6.

Images