Method and device for calculating dose of radioactive microspheres, electronic equipment and medium

Abstract

The application discloses a radioactive microsphere dose calculation method and device, electronic equipment and medium, and relates to the technical field of data processing. A liver artery model is constructed through multi-modal image fusion to provide an accurate geometric basis for dose calculation. Secondly, blood flow information is extracted from two-dimensional hepatic arteriography images to calculate a hemodynamic field, which truly reflects the blood flow state of the target object. Then, the blood flow field is used to drive the simulation of particle transport and comparison with the real distribution, the particle mapping model parameters are iteratively corrected, and the prediction deviation caused by the difference between the contrast agent and the particle physical properties is eliminated. After the physical properties of the radioactive microspheres are mapped to the corrected model, the movement and deposition of the radioactive microspheres are simulated in the individualized blood flow environment, and finally, a voxel-level dose distribution map is generated, the improvement from the partition estimation to the voxel-level accurate calculation is realized, and a reliable basis is provided for the accurate formulation of a treatment plan.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for calculating the dose of radioactive microspheres, a device for calculating the dose of radioactive microspheres, a computer-readable storage medium, and an electronic device. Background Technology

[0002] Selective internal radiotherapy involves injecting radioactive microspheres (e.g., Y90) via hepatic artery intervention to deliver a high-dose local irradiation to the tumor region. In related techniques, the calculation of the radioactive microsphere dose typically employs a zoned estimation method, dividing the liver into a limited number of regions and assuming uniform microsphere distribution within each region. This leads to discrepancies between the predicted Y90 microsphere distribution and the actual situation, making it impossible to accurately determine the radioactive microsphere dose. Summary of the Invention

[0003] This application aims to at least partially address one of the technical problems in related technologies. To this end, the first objective of this application is to propose a method for calculating the dose of radioactive microspheres. This method involves acquiring multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of a target object, and constructing a hepatic artery model based on these images. The method then calculates the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography images. The hemodynamic field is used to drive simulated particle transport to obtain a predicted distribution of the simulated particles. The actual distribution of the simulated particles in the target object is obtained, and the predicted distribution is compared with the actual distribution. Based on the comparison results, the model parameters of the fluid-simulated particle mapping model are corrected to obtain a corrected fluid-simulated particle mapping model. The corrected fluid-simulated particle mapping model is further modified based on the physical properties of the radioactive microspheres to obtain a fluid-radioactive microsphere mapping model. Radioactive microspheres are released into the hepatic artery model using the hemodynamic field as the transport environment, and the fluid-radioactive microsphere mapping model is used to calculate the trajectory and deposition location of the radioactive microspheres driven by the hemodynamic field, thus obtaining the distribution of the radioactive microspheres. A voxel-level radioactive microsphere dose distribution map is generated based on the radioactive microsphere distribution.

[0004] This application constructs a hepatic artery model through multimodal image fusion, providing a precise geometric basis for dose calculation. Secondly, it extracts blood flow information from two-dimensional hepatic artery angiography images to calculate the hemodynamic field, accurately reflecting the blood flow state of the target object. Then, it uses the blood flow field to drive simulated particle transport and compares it with the actual distribution, iteratively correcting the particle mapping model parameters to eliminate prediction bias caused by differences in the physical properties of contrast agents and particles. After mapping the physical properties of radioactive microspheres to the corrected model, it simulates the movement and deposition of radioactive microspheres in an individualized blood flow environment, ultimately generating a voxel-level dose distribution map. This achieves an improvement from zonal estimation to voxel-level precise calculation, providing a reliable basis for the accurate formulation of treatment plans.

[0005] The second objective of this application is to provide a device for calculating the dose of radioactive microspheres.

[0006] The third objective of this application is to provide a computer-readable storage medium.

[0007] The fourth objective of this application is to propose an electronic device.

[0008] To achieve the above objectives, the first aspect of this application proposes a method for calculating the dose of radioactive microspheres. This method involves acquiring multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object, and constructing a hepatic artery model based on these images. The method then calculates the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography images. The hemodynamic field is used to drive the transport of simulated particles to obtain a predicted distribution of the simulated particles. The actual distribution of the simulated particles in the target object is obtained, and the predicted distribution is compared with the actual distribution. Based on the comparison results, the model parameters of the fluid-simulated particle mapping model are corrected to obtain a corrected fluid-simulated particle mapping model. The corrected fluid-simulated particle mapping model is further modified based on the physical properties of the radioactive microspheres to obtain a fluid-radioactive microsphere mapping model. Radioactive microspheres are released into the hepatic artery model using the hemodynamic field as the transport environment, and the fluid-radioactive microsphere mapping model is used to calculate the trajectory and deposition location of the radioactive microspheres driven by the hemodynamic field, thus obtaining the distribution of the radioactive microspheres. Finally, a voxel-level radioactive microsphere dose distribution map is generated based on the radioactive microsphere distribution.

[0009] According to one embodiment of this application, a hepatic artery model is constructed based on multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images, including: registering the multimodal three-dimensional liver images based on a preset registration method to obtain spatially aligned multimodal three-dimensional liver images; inputting the spatially aligned multimodal three-dimensional liver images into a preset segmentation model to output an initial hepatic artery model; and fusing the initial hepatic artery model with the two-dimensional hepatic artery angiography images to obtain a hepatic artery model.

[0010] According to one embodiment of this application, the preset registration method includes rigid registration and non-rigid registration. Rigid registration is used to characterize the elimination of positional differences between multimodal three-dimensional liver images by finding rotation and translation matrices. Non-rigid registration is used to characterize the correction of physiological deformation of the liver caused by respiration or changes in body position by calculating deformable deformation fields.

[0011] According to one embodiment of this application, fusing an initial hepatic artery model with a two-dimensional hepatic artery angiography image to obtain a hepatic artery model includes: establishing an initial projection relationship between the initial hepatic artery model and the two-dimensional hepatic artery angiography image based on the C-arm rotation angle parameters of the hepatic artery angiography device; performing virtual X-ray projection on the initial hepatic artery model based on the initial projection relationship to generate a simulated two-dimensional hepatic artery angiography image; performing a similarity measurement between the simulated two-dimensional hepatic artery angiography image and the real two-dimensional hepatic artery angiography image to obtain a similarity measurement result; and adjusting the position and angle of the initial hepatic artery model based on the similarity measurement result until the simulated two-dimensional hepatic artery angiography image matches the real two-dimensional hepatic artery angiography image to obtain a hepatic artery model.

[0012] According to one embodiment of this application, calculating the hemodynamic field within a hepatic artery model based on two-dimensional hepatic artery angiography images includes: dividing the hepatic artery model into a computational grid and assigning blood physical parameters to the computational grid; extracting the time-density curve of the contrast agent's first passage through the hepatic artery vascular tree from the two-dimensional hepatic artery angiography images, and calculating the waveform of blood flow velocity or volumetric flow rate over time based on the time-density curve as the inlet boundary condition of the hepatic artery vascular tree; setting the outlet resistance or outlet flow rate ratio of the hepatic artery vascular tree outlet based on the diameter of the outlet branch of the hepatic artery vascular tree and tissue perfusion requirements, using Murray's law or preset perfusion parameters as the outlet boundary condition of the hepatic artery vascular tree; and numerically solving the computational grid with assigned blood physical parameters based on the inlet boundary condition, outlet boundary condition, Navier-Stokes equations, and mass conservation equations to obtain the hemodynamic field within the hepatic artery model.

[0013] According to one embodiment of this application, the model parameters of the fluid-simulated particle mapping model include an inertia coefficient, a sedimentation coefficient, and a size exclusion coefficient. The inertia coefficient is used to characterize the degree to which the simulated particles deviate from the streamline due to inertia at the bifurcation or bend of the blood vessel. The sedimentation coefficient is used to characterize the deposition tendency of the simulated particles in the direction of gravity. The size exclusion coefficient is used to characterize the probability that the simulated particles are stuck and embolized by small blood vessels.

[0014] According to one embodiment of this application, generating a voxel-level radioactive microsphere dose distribution map based on the distribution of radioactive microspheres includes: determining the radioactivity within each voxel of a hepatic artery model based on the distribution of radioactive microspheres to obtain an activity distribution map; calculating the absorbed radiation dose using a graphics processor-accelerated Monte Carlo simulation method for a first region in the activity distribution map to obtain the dose distribution of the tumor region, wherein the first region is used to characterize the tumor region; calculating the absorbed radiation dose using a dose convolution kernel method for a second region in the activity distribution map to obtain the dose distribution of the normal liver tissue region, wherein the second region is used to characterize the normal liver tissue region; and merging the dose distribution of the tumor region with the dose distribution of the normal liver tissue region to obtain a voxel-level radioactive microsphere dose distribution map.

[0015] To achieve the above objectives, a second aspect of this application provides a device for calculating the dose of radioactive microspheres, comprising: an acquisition module for acquiring multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of a target object; a construction module for constructing a hepatic artery model based on the multimodal three-dimensional liver images and the two-dimensional hepatic artery angiography images; a first calculation module for calculating the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography images; a driving module for driving simulated particle transport using the hemodynamic field to obtain a predicted distribution of the simulated particles; and a first correction module for acquiring the true distribution of the simulated particles of the target object and comparing the predicted distribution with the true distribution to obtain a predicted distribution of the simulated particles. The system first corrects the model parameters of the fluid-simulated particle mapping model based on the comparison results to obtain the corrected fluid-simulated particle mapping model. The second correction module is used to correct the corrected fluid-simulated particle mapping model based on the physical properties of the radioactive microspheres to obtain the fluid-radioactive microsphere mapping model. The second calculation module is used to release radioactive microspheres into the hepatic artery model using a hemodynamic field as the transport environment, and to calculate the motion trajectory and deposition position of the radioactive microspheres under the drive of the hemodynamic field using the fluid-radioactive microsphere mapping model to obtain the distribution of radioactive microspheres. The generation module is used to generate a voxel-level radioactive microsphere dose distribution map based on the distribution of radioactive microspheres.

[0016] To achieve the above objectives, a third aspect of this application provides a computer-readable storage medium storing a radioactive microsphere dose calculation program thereon, which, when executed by a processor, implements the aforementioned radioactive microsphere dose calculation method.

[0017] To achieve the above objectives, a fourth aspect of this application provides an electronic device, including a memory, a processor, and a radioactive microsphere dose calculation program stored in the memory and capable of running on the processor. When the processor executes the radioactive microsphere dose calculation program, it implements the aforementioned radioactive microsphere dose calculation method.

[0018] According to the method, apparatus, electronic device, and medium for calculating the dose of radioactive microspheres according to embodiments of this application, multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object are acquired, and a hepatic artery model is constructed based on the multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images; the hemodynamic field within the hepatic artery model is calculated based on the two-dimensional hepatic artery angiography images; the hemodynamic field is used to drive the transport of simulated particles to obtain the predicted distribution of simulated particles; the actual distribution of simulated particles of the target object is obtained, and the predicted distribution is compared with the actual distribution, and the model parameters of the fluid-simulated particle mapping model are corrected based on the comparison results to obtain the corrected fluid-simulated particle mapping model; the corrected fluid-simulated particle mapping model is further corrected based on the physical properties of radioactive microspheres to obtain a fluid-radioactive microsphere mapping model; radioactive microspheres are released into the hepatic artery model using the hemodynamic field as the transport environment, and the motion trajectory and deposition position of the radioactive microspheres driven by the hemodynamic field are calculated using the fluid-radioactive microsphere mapping model to obtain the distribution of radioactive microspheres; a voxel-level radioactive microsphere dose distribution map is generated based on the distribution of radioactive microspheres.

[0019] This application constructs a hepatic artery model through multimodal image fusion, providing a precise geometric basis for dose calculation. Secondly, it extracts blood flow information from two-dimensional hepatic artery angiography images to calculate the hemodynamic field, accurately reflecting the blood flow state of the target object. Then, it uses the blood flow field to drive simulated particle transport and compares it with the actual distribution, iteratively correcting the particle mapping model parameters to eliminate prediction bias caused by differences in the physical properties of contrast agents and particles. After mapping the physical properties of radioactive microspheres to the corrected model, it simulates the movement and deposition of radioactive microspheres in an individualized blood flow environment, ultimately generating a voxel-level dose distribution map. This achieves an improvement from zonal estimation to voxel-level precise calculation, providing a reliable basis for the accurate formulation of treatment plans. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a method for calculating the dose of radioactive microspheres according to some embodiments of this application; Figure 2 This is a block diagram of a device for calculating the dose of radioactive microspheres according to some embodiments of this application; Figure 3 This is a block diagram of an electronic device according to some embodiments of this application. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0022] The following describes in detail, with reference to the accompanying drawings, the method, apparatus, electronic device, and medium for calculating the dose of radioactive microspheres according to embodiments of this application.

[0023] Figure 1 This is a flowchart illustrating a method for calculating the dose of radioactive microspheres according to some embodiments of this application. (Refer to...) Figure 1 The method for calculating the dose of radioactive microspheres in this application embodiment may include the following steps: S110: Acquire multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object, and construct a hepatic artery model based on the multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images. The target object can be a patient to be treated or an animal used in experiments. In this embodiment, a patient to be treated is used as an example for illustration, but this is not intended to limit the scope of this application.

[0024] Specifically, multimodal three-dimensional liver images of the target object can be acquired through CT (Computed Tomography), MRI (Magnetic Resonance Imaging), and CBCT (Cone Beam Computed Tomography), while two-dimensional hepatic artery angiography images of the target object can be acquired through DSA (Digital Subtraction Angiography).

[0025] After acquiring multimodal 3D liver images of the target object, the multimodal 3D liver images of the target object can be registered to obtain the target 3D liver image. Then, the target 3D liver image is input into a preset segmentation model to output the initial hepatic artery model of the target object. Then, the initial hepatic artery model is fused with the 2D hepatic artery angiography image to obtain the hepatic artery model of the target object. For example, the 3D model needs to be digitally reconstructed first, and its vessel centerline and key branch points need to be extracted. At the same time, the 2D angiography image is enhanced to highlight the vascular structure. On this basis, the vessel bifurcation points are used as feature markers. Through registration algorithms such as iterative nearest point or similarity measure optimization, the optimal projection matrix of the 3D model on the 2D image plane is calculated. The 3D vascular skeleton is then mapped to the 2D image plane using this matrix. After the mapping is completed, the mapped vascular skeleton is superimposed on the 2D angiography image in the form of a semi-transparent color projection map or outline, and finally the hepatic artery model of the target object is obtained.

[0026] S120, calculates the hemodynamic field within the hepatic artery model based on two-dimensional hepatic artery angiography images.

[0027] Specifically, after constructing the hepatic artery model, the temporal-spatial information of blood flow is extracted from the two-dimensional hepatic artery angiography image and transformed into physical boundary conditions on the hepatic artery model. Then, the hemodynamic field of the hepatic artery is solved through fluid dynamics equations.

[0028] S130 uses a hemodynamic field to drive simulated particle transport in order to obtain the predicted distribution of simulated particles.

[0029] S140, obtain the true distribution of simulated particles of the target object, compare the predicted distribution with the true distribution, and correct the model parameters of the fluid-simulated particle mapping model based on the comparison results to obtain the corrected fluid-simulated particle mapping model.

[0030] Specifically, simulated particles refer to mimics of radioactive microspheres, such as MAA (Macroaggregated Albumin). MAA particles have a similar particle size range to radioactive microspheres (e.g., Y90 microspheres) and exhibit similar movement in the bloodstream, mimicking the distribution pattern of Y90 microspheres in the hepatic artery. Furthermore, MAA is a biodegradable macroaggregated albumin that is broken down and cleared from the body within hours of injection, preventing permanent vascular embolism. MAA is also labeled with technetium-99m, allowing for accurate in vivo distribution images obtained via SPECT (Single Photon Emission Computed Tomography) or CT scans, providing a reference standard for subsequent model corrections.

[0031] Therefore, before injecting radioactive microspheres, a pre-constructed hepatic artery model is used as the geometric boundary, and the calculated hemodynamic field is used as the driving force. Simulated particles are released at preset injection positions in the hepatic artery model. Each particle moves along the hepatic artery under the drive of the blood flow field, while the physical properties of the particles are superimposed. Finally, the deposition positions and numbers of all particles are recorded to obtain the predicted distribution of simulated particles in the hepatic artery model.

[0032] Simultaneously, simulated particles are released at predetermined injection sites in the target subject's actual liver. SPECT / CT scans are then used to obtain the actual deposition distribution of these simulated particles in various branches of the hepatic artery, the tumor region, normal liver tissue, and the lungs, serving as the true distribution of the simulated particles. The predetermined injection sites in the hepatic artery model are the same as those in the target subject's actual liver.

[0033] Furthermore, a fluid-simulated particle mapping model is pre-established based on classical physical theories in fluid mechanics and particle dynamics. Here, the fluid refers to the contrast agent, which is a dissolved fluid that completely follows the blood flow without inertial sedimentation. The fluid-simulated particle mapping model maps the flow information of the blood flow field to the trajectory of particles, reflecting the behavior of particles deviating from streamlines in regions of vascular bifurcation, bending, and velocity variation due to their own mass and volume. Its mathematical framework consists of formulas describing physical effects such as particle inertia, sedimentation, and size exclusion. The specific values ​​of the parameters in the model are determined through iterative adjustment by comparing the predicted distribution with the actual distribution.

[0034] For example, the predicted distribution map of simulated particles is spatially registered with the actual distribution map to align the two maps in three-dimensional space. After alignment, the proportion of simulated particle deposition or activity in each region of the predicted and actual distributions is statistically analyzed in both the tumor and normal liver tissue regions. The deviation between the two is calculated for each region to obtain the deviation amount for comparison. Further, parameter optimization algorithms (such as gradient descent, Bayesian optimization, etc.) can be used to automatically search for the optimal values ​​of model parameters with the goal of minimizing the deviation between the predicted and actual distributions. After each iteration, the model parameters are updated, and the simulation is re-enhanced and the deviation is recalculated until the deviation converges, thus obtaining the model parameters of the fluid-simulated particle mapping model.

[0035] To further illustrate the above embodiments, in this application embodiment, the model parameters of the fluid-simulated particle mapping model include inertia coefficient, sedimentation coefficient and size exclusion coefficient. The inertia coefficient is used to characterize the degree to which the simulated particles deviate from the streamline due to inertia at the bifurcation or bend of the blood vessel. The sedimentation coefficient is used to characterize the deposition tendency of the simulated particles in the direction of gravity. The size exclusion coefficient is used to characterize the probability that the simulated particles are stuck and embolized by small blood vessels.

[0036] For example, the predicted distribution is compared with the actual distribution in the tumor area and normal liver tissue area on a region-by-region basis to obtain the particle deposition ratio deviation in each region. The inertia coefficient, sedimentation coefficient and size exclusion coefficient are adjusted according to the deviation amount and deviation direction.

[0037] Regarding the inertia coefficient: when the proportion of particle deposition in the lung region in the predicted distribution is higher than that in the actual distribution, it indicates that the inertia coefficient is too large. The inertia coefficient is reduced based on the preset step size to reduce the degree to which particles deviate from the streamline at the vascular bifurcation point and rush into the lung shunt path; when the proportion of particle deposition in the lung region in the predicted distribution is lower than that in the actual distribution, it indicates that the inertia coefficient is too small. The inertia coefficient is increased based on the preset step size to increase the degree to which particles deviate from the streamline at the vascular bifurcation point.

[0038] Regarding the sedimentation coefficient: when the proportion of particle deposition in a normal liver tissue region in the predicted distribution is higher than that in the actual distribution, and the blood vessels in that region are thicker and the flow velocity is slower, it indicates that the sedimentation coefficient is too large. Based on the preset step size, the sedimentation coefficient is reduced to weaken the tendency of particles to deposit in the direction of gravity. When the proportion of particle deposition in that region in the predicted distribution is lower than that in the actual distribution, and the blood vessels in that region are thicker and the flow velocity is slower, it indicates that the sedimentation coefficient is too small. Based on the preset step size, the sedimentation coefficient is increased to enhance the tendency of particles to deposit in the direction of gravity.

[0039] Regarding the size exclusion coefficient: when the proportion of particles deposited in proximal vessels in the predicted distribution is higher than that in the actual distribution, while there is insufficient deposition in distal vessels, it indicates that the size exclusion coefficient is too large. The size exclusion coefficient is reduced based on the preset step size to reduce the probability of particles being trapped and embolized by small vessels. When the proportion of particles deposited in proximal vessels in the predicted distribution is lower than that in the actual distribution, while there is excessive deposition in distal vessels, it indicates that the size exclusion coefficient is too small. The size exclusion coefficient is increased based on the preset step size to increase the probability of particles being trapped and embolized by small vessels.

[0040] After each adjustment, the simulated particle transport is performed again to obtain the updated predicted distribution, which is then compared with the actual distribution. This process is repeated iteratively until the deviations between the predicted and actual particle deposition ratios in each region meet the preset consistency conditions, thus completing the parameter correction of the fluid-simulated particle mapping model.

[0041] S150, the modified fluid-simulated particle mapping model is modified based on the physical properties of the radioactive microspheres to obtain the fluid-radioactive microsphere mapping model.

[0042] Specifically, the distribution and calculation of radioactive microspheres are two closely linked steps. First, the modified fluid-simulated particle mapping model is used to calculate where and how much of the microspheres are deposited. Then, based on the activity of the deposits at each location, the energy released by the microspheres is quickly calculated. Therefore, the distribution of radioactive microspheres needs to be determined based on the fluid-simulated particle mapping model that has been validated using MAA. Since the physical properties of MAA-simulated particles and radioactive microspheres are different, the model parameters of the modified fluid-simulated particle mapping model cannot be directly used. Therefore, while keeping the framework of the modified fluid-simulated particle mapping model unchanged, the physical parameters of the MAA-simulated particles are replaced with the physical parameters of the radioactive microspheres, and the model parameters in the model are recalculated to obtain coefficient values ​​suitable for the radioactive microspheres.

[0043] Specifically, the physical properties of radioactive microspheres include particle size, density, and shape. The particle size of radioactive microspheres can be either glass microspheres or resin microspheres. Glass microspheres have a particle size of approximately 20-30 μm, resin microspheres approximately 20-60 μm, while MAA simulated particles have a particle size of approximately 10-90 μm, which affects the size exclusion coefficient. The density of glass microspheres is approximately 3.3 g / cm³, resin microspheres approximately 1.6 g / cm³, while MAA simulated particles have a density of approximately 1.0 g / cm³, which significantly alters the inertia coefficient and sedimentation coefficient. Radioactive microspheres have regular shapes, while MAA simulated particles have irregular shapes, which also affects the size exclusion coefficient.

[0044] For example, for the size exclusion coefficient, the particle size distribution parameters of the MAA simulated particles are replaced with the particle size distribution parameters of the radioactive microspheres. When the radioactive microspheres are glass microspheres, a particle size parameter of about 20-30 μm is used, and when the radioactive microspheres are resin microspheres, a particle size parameter of about 20-60 μm is used. At the same time, the irregular shape factor of the MAA simulated particles is replaced with the regular sphericity shape factor of the radioactive microspheres. These are substituted into the calculation formula of the size exclusion coefficient to recalculate the size exclusion coefficient of the radioactive microspheres.

[0045] For the inertia coefficient, the density parameter of the MAA simulated particles, approximately 1.0 g / cm³, is replaced with the density parameter of the radioactive microspheres. When the radioactive microspheres are glass microspheres, a density parameter of approximately 3.3 g / cm³ is used, and when the radioactive microspheres are resin microspheres, a density parameter of approximately 1.6 g / cm³ is used. Substitute these values ​​into the formula for calculating the inertia coefficient, and the inertia coefficient of the radioactive microspheres is recalculated. For the sedimentation coefficient, the density parameters of the MAA simulated particles were replaced with the density parameters of the radioactive microspheres, and then substituted into the sedimentation coefficient calculation formula to recalculate the sedimentation coefficient of the radioactive microspheres. The recalculated size exclusion coefficient, inertia coefficient, and sedimentation coefficient of the radioactive microspheres are substituted into the model framework of the modified fluid-simulated particle mapping model to obtain the fluid-radioactive microsphere mapping model.

[0046] S160 uses a hemodynamic field as the transport environment to release radioactive microspheres into a hepatic artery model, and uses a fluid-radioactive microsphere mapping model to calculate the trajectory and deposition location of the radioactive microspheres driven by the hemodynamic field, thus obtaining the distribution of the radioactive microspheres.

[0047] Specifically, using a pre-calculated hemodynamic field as the transport environment, virtual particles representing radioactive microspheres are released at preset injection sites in the hepatic artery model according to actual injection parameters. Each particle moves along the blood vessel under the drive of the blood flow velocity field and pressure field. At the same time, the behavior of the particles in the regions of blood vessel bifurcation, bending and flow velocity change is corrected using a fluid-radioactive microsphere mapping model. The inertial deviation, gravitational sedimentation and size exclusion effect caused by the particles' own mass, particle size and shape are simulated. Finally, the deposition position and number of virtual particles of all radioactive microspheres in the hepatic artery model are recorded to obtain the distribution of radioactive microspheres in the hepatic artery model.

[0048] S170, generating a voxel-level radioactive microsphere dose distribution map based on the distribution of radioactive microspheres.

[0049] Specifically, after obtaining the distribution of radioactive microspheres (the deposition location and number of virtual particles of radioactive microspheres in the hepatic artery model), the radioactivity in each three-dimensional voxel unit is determined based on the deposition location and number of radioactive microspheres in the hepatic artery model. Then, using a preset dose calculation method, the radiation energy deposited in each voxel is calculated, and the total energy deposited in each voxel is divided by the mass of the corresponding voxel to obtain the absorbed dose of each voxel. Finally, the absorbed doses of all voxels are summarized to generate a voxel-level radioactive microsphere dose distribution map displayed as a three-dimensional color image.

[0050] This application constructs a hepatic artery model through multimodal image fusion, providing a precise geometric basis for dose calculation. Secondly, it extracts blood flow information from two-dimensional hepatic artery angiography images to calculate the hemodynamic field, accurately reflecting the blood flow state of the target object. Then, it uses the blood flow field to drive simulated particle transport and compares it with the actual distribution, iteratively correcting the particle mapping model parameters to eliminate prediction bias caused by differences in the physical properties of contrast agents and particles. After mapping the physical properties of radioactive microspheres to the corrected model, it simulates the movement and deposition of radioactive microspheres in an individualized blood flow environment, ultimately generating a voxel-level dose distribution map. This achieves an improvement from zonal estimation to voxel-level precise calculation, providing a reliable basis for the accurate formulation of treatment plans.

[0051] In some embodiments, constructing a hepatic artery model based on multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images includes: registering the multimodal three-dimensional liver images based on a preset registration method to obtain spatially aligned multimodal three-dimensional liver images; inputting the spatially aligned multimodal three-dimensional liver images into a preset segmentation model to output an initial hepatic artery model; and fusing the initial hepatic artery model with the two-dimensional hepatic artery angiography images to obtain a hepatic artery model.

[0052] Specifically, 3D liver images acquired by different devices, such as CT and MRI, are often taken when the subject's position and respiratory status are different, leading to a mismatch in the spatial location of the same anatomical structure in the two images. Therefore, acquiring multimodal 3D liver images requires registration processing based on a preset registration method to ensure accurate spatial alignment of the multimodal 3D liver images.

[0053] In some embodiments, the preset registration method includes rigid registration and non-rigid registration.

[0054] Rigid registration is used to characterize the elimination of positional differences between multimodal 3D liver images by finding rotation and translation matrices. For example, by maximizing similarity measures such as mutual information or normalized cross-correlation between 3D liver images, a set of globally optimal rotation and translation matrices can be found to eliminate positional differences in liver images of the same target object acquired by different devices or at different times, so that multiple liver images of the target object achieve overall spatial alignment.

[0055] Non-rigid registration is used to characterize the correction of physiological deformations of the liver caused by respiration or changes in body position by calculating a deformable deformation field. For example, a deformable deformation field that describes the displacement of each pixel or voxel can be calculated by using methods such as B-spline-based free deformation, the Demons algorithm based on physical models, or deep learning models, with image similarity (such as mean square error) as the optimization objective. This deformation field is then applied to the image to be registered, aligning it with the reference image in terms of local anatomical structure, thereby correcting local morphological differences in the liver caused by respiratory movements or changes in body position.

[0056] Furthermore, after spatially aligning the multimodal 3D liver images, the liver, tumor, and blood vessels on the images all have clear positions in the same coordinate system. Then, the spatially aligned multimodal 3D liver images can be fed into a pre-trained deep learning segmentation model, such as U-Net, to output the topology of the hepatic artery (initial hepatic artery model), the tumor region, and the normal liver region. The hepatic artery includes the common hepatic artery, proper hepatic artery, left hepatic artery, right hepatic artery, and small branches such as the tumor-feeding artery.

[0057] Furthermore, the preoperative 3D liver model is static, while the position and shape of the liver may change due to factors such as body position and respiration when the target patient is lying on the operating table. Therefore, it is necessary to fuse the preoperative 3D liver model with the intraoperative 2D hepatic artery angiography image to adapt the 3D liver model to the actual anatomical position of the target patient during the operation, thereby obtaining an accurate hepatic artery model.

[0058] This application employs a multimodal image registration method combining rigid and non-rigid registration. First, it eliminates positional differences between 3D liver images acquired by different devices using a globally optimal rotation and translation matrix. Then, it corrects local physiological deformations of the liver caused by respiratory movements or changes in body position using a deformable deformation field, ensuring precise spatial alignment of the multimodal images and providing a unified coordinate reference for subsequent model construction. A deep learning segmentation model is used to automatically extract the hepatic artery topology, tumor region, and normal liver region from the spatially aligned multimodal images, achieving automated reconstruction from 2D images to 3D anatomical structures. Furthermore, by fusing the preoperative static 3D model with real-time intraoperative 2D hepatic artery angiography images, the 3D model is adapted to the patient's actual anatomical position during surgery, thus constructing a hepatic artery model that considers both detailed preoperative anatomical information and real-time intraoperative spatial position, improving the model's accuracy and intraoperative applicability.

[0059] In some embodiments, fusing an initial hepatic artery model with a two-dimensional hepatic artery angiography image to obtain a hepatic artery model includes: establishing an initial projection relationship between the initial hepatic artery model and the two-dimensional hepatic artery angiography image based on the C-arm rotation angle parameters of the hepatic artery angiography device; performing virtual X-ray projection on the initial hepatic artery model based on the initial projection relationship to generate a simulated two-dimensional hepatic artery angiography image; performing a similarity measurement between the simulated two-dimensional hepatic artery angiography image and the real two-dimensional hepatic artery angiography image to obtain a similarity measurement result; and adjusting the position and angle of the initial hepatic artery model based on the similarity measurement result until the simulated two-dimensional hepatic artery angiography image matches the real two-dimensional hepatic artery angiography image to obtain a hepatic artery model.

[0060] Specifically, when using a DSA device to acquire two-dimensional hepatic artery angiography images of a target object, the C-arm of the DSA device can rotate around the patient to capture two-dimensional images from different angles. Geometric parameters such as the rotation angle, tilt angle of the C-arm, and the distance from the X-ray source to the detector can be used to establish an initial projection relationship between an initial hepatic artery model and the two-dimensional hepatic artery angiography image in computer virtual space.

[0061] Then, based on the initial projection mapping relationship, the initial hepatic artery model is projected through a perspective view along the direction from the virtual X-ray source to the virtual detector. This simulates the attenuation process of X-rays passing through the initial three-dimensional hepatic artery model, generating a simulated two-dimensional hepatic artery angiography image. This simulated image reflects the two-dimensional projection shape that the initial three-dimensional hepatic artery model should present under the current position and angle assumptions.

[0062] Subsequently, the similarity between the simulated two-dimensional hepatic artery angiography image and the real two-dimensional hepatic artery angiography image acquired by the hepatic artery angiography device is measured. The difference in gray-level gradient or mutual information between the two is calculated to obtain the similarity measurement result that quantitatively represents the degree of matching between the two.

[0063] Finally, using the similarity measurement result as the objective function, a preset 2D-3D registration algorithm was employed to iteratively optimize the six-degree-of-freedom pose parameters of the initial hepatic artery model in three-dimensional space. In each iteration, the model's translation vector and rotation matrix were adjusted, the simulation image was regenerated, and the similarity measurement was calculated until the similarity measurement result converged to the preset matching conditions. At this point, a precise spatial mapping relationship was established between the three-dimensional model and the patient's intraoperative anatomical position. The three-dimensional model after registration is the hepatic artery model.

[0064] In some embodiments, calculating the hemodynamic field within a hepatic artery model based on two-dimensional hepatic artery angiography images includes: dividing the hepatic artery model into a computational grid and assigning blood physical parameters to the computational grid; extracting the time-density curve of the contrast agent's first passage through the hepatic artery vascular tree from the two-dimensional hepatic artery angiography images, and calculating the waveform of blood flow velocity or volumetric flow rate over time based on the time-density curve as the inlet boundary condition of the hepatic artery vascular tree; setting the outlet resistance or outlet flow rate ratio of the hepatic artery vascular tree outlet based on the diameter of the outlet branch of the hepatic artery vascular tree and tissue perfusion requirements, using Murray's law or preset perfusion parameters as the outlet boundary condition of the hepatic artery vascular tree; and numerically solving the computational grid with assigned blood physical parameters based on the inlet boundary condition, outlet boundary condition, Navier-Stokes equations, and mass conservation equations to obtain the hemodynamic field within the hepatic artery model.

[0065] Specifically, firstly, the vascular tree in the hepatic artery model is spatially discretized, and a volume mesh composed of tetrahedral or hexahedral elements is generated inside its surface model to obtain a computational mesh. Each element in the computational mesh is then assigned physical parameters of blood, including setting blood density and blood viscosity, while assuming that blood is an incompressible Newtonian fluid.

[0066] Secondly, from the two-dimensional hepatic artery angiography images, the time-density curve of the contrast agent first passing through the hepatic artery vascular tree is extracted at the entrance position such as the common hepatic artery or the proper hepatic artery. Based on this time-density curve, the waveform of blood flow velocity or volumetric flow rate changes with time is estimated by applying dilution theory or optical flow method, and it is used as the entrance boundary condition at the entrance of the hepatic artery vascular tree.

[0067] Meanwhile, based on the vessel diameter of each outlet branch of the hepatic artery vascular tree and the perfusion requirements of the tissues supplied, Murray's law or preset perfusion parameters are used to set outlet resistance or outlet flow ratio for each outlet as outlet boundary conditions. Among them, preset perfusion parameters include hepatic artery perfusion volume, portal vein perfusion volume, and total perfusion volume of the whole liver, etc. The preset perfusion parameters can be determined according to the actual situation, and no specific restrictions are imposed here.

[0068] Finally, substituting the above inlet and outlet boundary conditions into the governing equations consisting of the Navier-Stokes equations and the mass conservation equations, the governing equations are as follows:

[0069]

[0070] Where v represents the velocity vector field; p represents the pressure field; μ represents the dynamic viscosity; and f represents the volume force term (such as gravity).

[0071] Then, on the divided grid, the finite volume method is used to spatially discretize the governing equations, transforming the continuous partial differential equations into a system of linear or nonlinear algebraic equations on discrete grid cells. A SIMPLE (Semi-Implicit Method for Pressure Linked Equations) algorithm is used to iteratively solve the coupled velocity and pressure equations, updating the velocity and pressure values ​​at each grid node in each iteration until the normalized residuals at all grid nodes are less than a preset convergence criterion (e.g., residual ≤ 10). -4 Once the iteration converges, the velocity vector distribution and pressure scalar distribution of each grid node in the entire hepatic artery model are output, thereby obtaining the hemodynamic field (velocity and pressure values ​​at every point in the blood vessel at every moment).

[0072] In some embodiments, generating a voxel-level radioactive microsphere dose distribution map based on the distribution of radioactive microspheres includes: determining the radioactivity within each voxel of the hepatic artery model based on the distribution of radioactive microspheres to obtain an activity distribution map; calculating the absorbed radiation dose using a graphics processor-accelerated Monte Carlo simulation method for a first region in the activity distribution map to obtain the dose distribution of the tumor region, wherein the first region is used to characterize the tumor region; calculating the absorbed radiation dose using a dose convolution kernel method for a second region in the activity distribution map to obtain the dose distribution of the normal liver tissue region, wherein the second region is used to characterize the normal liver tissue region; and merging the dose distribution of the tumor region with the dose distribution of the normal liver tissue region to obtain a voxel-level radioactive microsphere dose distribution map.

[0073] Specifically, in the above embodiments, the distribution of radioactive microspheres within the hepatic artery model has been obtained, namely, the final deposition location and quantity of each virtual radioactive microsphere. Based on this distribution, using the three-dimensional space of the hepatic artery model as a reference frame, the entire computational domain is divided into a regular three-dimensional voxel grid. The number of radioactive microsphere particles falling into each voxel unit is counted, and the total accumulated radioactive activity within each voxel is calculated based on the radioactivity carried by each particle. The activity values ​​of all voxels together constitute an activity distribution map, which characterizes the overall activity deposition of radioactive microspheres in space.

[0074] The first region, the tumor region, requires high precision in dose calculation; therefore, a Monte Carlo simulation accelerated by a graphics processor (GPU) is employed. This method calculates β particles released by radioactive microspheres deposited within voxels in the tumor region. Specifically, based on the radioactivity of each voxel in the activity distribution map, the number of β particles to be simulated within each voxel is determined. Utilizing the massively parallel computing capabilities of the GPU, numerous β particle simulation threads are simultaneously initiated. Each β particle is emitted from its corresponding voxel's microsphere deposition location, and its step direction and step size are randomly determined according to a physical scattering model. During its interaction with tissue atoms, kinetic energy is gradually deposited. After each interaction, the particle's energy decays and its direction deflects until its energy is exhausted. Throughout this process, the energy deposited by each β particle as it passes through each voxel is recorded in real time. After all β particle simulations are completed, the total energy deposited within each voxel in the tumor region is accumulated and divided by the tissue mass of the corresponding voxel to obtain the absorbed dose of each voxel. The dose values ​​of all voxels in the tumor region collectively constitute the dose distribution of the tumor region.

[0075] The second region, the normal liver tissue area, is quite large. Using Monte Carlo simulation particle-by-particle would be computationally intensive, making it unsuitable for real-time intraoperative calculations. Therefore, the dose convolution kernel method is used for rapid calculation. The dose convolution kernel is a pre-calculated three-dimensional dose distribution function, physically representing the absorbed dose value produced by a point source of unit activity at various locations in an infinitely homogeneous medium. This kernel is pre-obtained and stored through a one-time high-precision Monte Carlo simulation or experimental measurement, ready for direct use in each calculation. Specifically, the activity distribution of the normal liver tissue region in the activity distribution map is considered as a set of point sources, with each voxel corresponding to a point source, and the activity value of that voxel representing its source strength. The goal of the convolution calculation is to calculate the sum of the dose contributions from all point sources at each voxel location. For each target voxel, its absorbed dose is contributed by all source voxels. Specifically, for each source voxel, the spatial distance between it and the target voxel is calculated. The dose contribution value corresponding to this distance is then found in the pre-stored dose convolution kernel. This contribution value is multiplied by the activity value of the source voxel to obtain the dose contribution of the source voxel to the target voxel. This process is repeated for all source voxels, and the dose contributions of each source voxel to the target voxel are summed to obtain the total absorbed dose of the target voxel. This process is equivalent to performing a three-dimensional spatial convolution operation between the three-dimensional activity distribution matrix of the normal liver tissue region and the dose convolution kernel. After the convolution operation, each voxel in the normal liver tissue region obtains an absorbed dose value, and the dose values ​​of all voxels together constitute the dose distribution of the normal liver tissue region.

[0076] The dose distribution in the tumor region and the dose distribution in the normal liver tissue region are merged according to their respective spatial coordinates. Since the two regions do not overlap, the merged dataset covers the entire liver computational domain, and each voxel has a defined absorbed dose value. This dataset is a voxel-level radioactive microsphere dose distribution map, which can be presented in the form of a three-dimensional color image, intuitively showing the distribution of radiation doses in the tumor region and the normal liver tissue region, providing doctors with quantitative evidence to assess the efficacy and safety of treatment plans.

[0077] This application transforms the distribution of radioactive microsphere deposition into a voxel-level activity distribution map and employs differentiated dose calculation strategies for tumor and normal liver tissue regions. Specifically, for tumor regions requiring high precision, a Monte Carlo simulation accelerated by a graphics processor is used to accurately simulate the random transport and energy deposition process of β particles in the tissue. For the larger normal liver tissue regions, a dose convolution kernel method is used for rapid convolution operations. This approach significantly improves overall computational efficiency while ensuring the accuracy of dose calculation in the tumor region. Ultimately, a voxel-level radioactive microsphere dose distribution map covering the entire liver is generated, presenting the distribution of high and low radiation doses in a three-dimensional color image, providing a quantitative basis for evaluating the effectiveness and safety of treatment plans.

[0078] In some embodiments, after determining the dose distribution map of voxel-level radioactive microspheres, intraoperative real-time monitoring and postoperative iterative correction are performed, as follows: Intraoperative real-time monitoring and pulmonary shunt early warning: During the injection of radioactive microspheres, parameters such as injection flow rate, injection volume, and catheter tip position of the microspheres are acquired in real time using two-dimensional hepatic arteriography. These parameters are simultaneously input into a pre-constructed fluid-radioactive microsphere mapping model. Based on this model, the real-time transport status of the microspheres in the current hemodynamic field is dynamically simulated, continuously tracking the distribution ratio of microspheres in various branches of the hepatic artery. When the model detects an abnormal shunt path of microspheres from the hepatic artery to the lungs, and the estimated proportion of microsphere activity entering the lungs exceeds a preset safety threshold, an immediate pulmonary shunt early warning signal is issued, and intervention suggestions are provided to the surgeon, including pausing the injection, adjusting the catheter tip position, or embolizing the abnormal shunt path to avoid excessive radiation damage to the lungs.

[0079] Postoperative Iterative Correction: After the injection of radioactive microspheres, PET / CT scans of the target subject are acquired, directly reflecting the actual distribution of the radioactive microspheres in vivo. The actual distribution of radioactive microspheres reconstructed from the PET / CT image is spatially registered with the preoperative predicted distribution, and quantitative comparisons are performed region by region in both the tumor area and normal liver tissue area to calculate the dose deviation in each region. Based on the dose deviation, the inertia coefficient, sedimentation coefficient, and size exclusion coefficient in the fluid-radioactive microsphere mapping model are iteratively adjusted until the deviation between the adjusted predicted distribution and the actual distribution in the postoperative PET / CT scan meets a preset convergence condition, such as the dose deviation being less than a preset threshold. The corrected model parameters are stored and used for prediction of subsequent cases. By continuously accumulating case data, the predictive accuracy and generalization ability of the model are gradually improved.

[0080] This application uses a fluid-radioactive microsphere mapping model to track the microsphere transport status in real time during surgery. When excessive pulmonary shunt is detected, an immediate warning is issued and intervention suggestions are provided to effectively avoid excessive radiation damage to the lungs. After surgery, the actual distribution of the radioactive microspheres reconstructed from PET / CT images is quantitatively compared with the preoperative predicted distribution region by region. Based on the dose deviation, the inertia coefficient, sedimentation coefficient, and size exclusion coefficient of the mapping model are iteratively corrected, and the corrected parameters are fed back to the model, forming a closed-loop optimization mechanism from prediction, validation to correction. This allows the model to continuously evolve with the accumulation of cases, gradually improving prediction accuracy and generalization ability.

[0081] Corresponding to the above embodiments, this application also proposes a device for calculating the dose of radioactive microspheres.

[0082] Reference Figure 2 The radioactive microsphere dose calculation device 200 includes: an acquisition module 210, a construction module 220, a first calculation module 230, a driving module 240, a first correction module 250, a second correction module 260, a second calculation module 270, and a generation module 280.

[0083] The system comprises the following modules: Acquisition module 210 acquires multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object. Construction module 220 constructs a hepatic artery model based on the multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images. First calculation module 230 calculates the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography images. Driving module 240 drives the transport of simulated particles using the hemodynamic field to obtain the predicted distribution of the simulated particles. First correction module 250 acquires the actual distribution of the simulated particles of the target object, compares the predicted distribution with the actual distribution, and corrects the model parameters of the fluid-simulated particle mapping model based on the comparison results to obtain a corrected fluid-simulated particle mapping model. Second correction module 260 corrects the corrected fluid-simulated particle mapping model based on the physical properties of the radioactive microspheres to obtain a fluid-radioactive microsphere mapping model. Second calculation module 270 releases radioactive microspheres into the hepatic artery model using the hemodynamic field as the transport environment, and calculates the trajectory and deposition position of the radioactive microspheres driven by the hemodynamic field using the fluid-radioactive microsphere mapping model to obtain the distribution of the radioactive microspheres. The generation module 280 is used to generate a voxel-level radioactive microsphere dose distribution map based on the distribution of radioactive microspheres.

[0084] According to one embodiment of this application, the construction module 220 is specifically used to: register multimodal three-dimensional liver images based on a preset registration method to obtain spatially aligned multimodal three-dimensional liver images; input the spatially aligned multimodal three-dimensional liver images into a preset segmentation model to output an initial hepatic artery model; and fuse the initial hepatic artery model with two-dimensional hepatic artery angiography images to obtain a hepatic artery model.

[0085] According to one embodiment of this application, the preset registration method includes rigid registration and non-rigid registration. Rigid registration is used to characterize the elimination of positional differences between multimodal three-dimensional liver images by finding rotation and translation matrices. Non-rigid registration is used to characterize the correction of physiological deformation of the liver caused by respiration or changes in body position by calculating deformable deformation fields.

[0086] According to one embodiment of this application, the construction module 220 is specifically used to: establish an initial projection relationship between an initial hepatic artery model and a two-dimensional hepatic artery angiography image based on the C-arm rotation angle parameters of the hepatic artery angiography device; perform virtual X-ray projection on the initial hepatic artery model based on the initial projection relationship to generate a simulated two-dimensional hepatic artery angiography image; perform a similarity measurement between the simulated two-dimensional hepatic artery angiography image and the real two-dimensional hepatic artery angiography image to obtain a similarity measurement result; and adjust the position and angle of the initial hepatic artery model based on the similarity measurement result until the simulated two-dimensional hepatic artery angiography image matches the real two-dimensional hepatic artery angiography image to obtain the hepatic artery model.

[0087] According to one embodiment of this application, the first calculation module 230 is specifically used to: divide the hepatic artery model into a grid to obtain a computational grid, and assign blood physical parameters to the computational grid; extract the time-density curve of the contrast agent passing through the hepatic artery vascular tree for the first time from the two-dimensional hepatic artery angiography image, and calculate the waveform of blood flow velocity or volumetric flow rate changing with time based on the time-density curve, as the inlet boundary condition of the hepatic artery vascular tree; set the outlet resistance or outlet flow rate ratio of the hepatic artery vascular tree outlet according to the diameter of the outlet branch of the hepatic artery vascular tree and the tissue perfusion requirements, using Murray's law or preset perfusion parameters, as the outlet boundary condition of the hepatic artery vascular tree; and perform numerical solution on the computational grid assigned with blood physical parameters based on the inlet boundary condition, the outlet boundary condition, the Navier-Stokes equation, and the mass conservation equation to obtain the hemodynamic field within the hepatic artery model.

[0088] According to one embodiment of this application, the model parameters of the fluid-simulated particle mapping model include an inertia coefficient, a sedimentation coefficient, and a size exclusion coefficient. The inertia coefficient is used to characterize the degree to which the simulated particles deviate from the streamline due to inertia at the bifurcation or bend of the blood vessel. The sedimentation coefficient is used to characterize the deposition tendency of the simulated particles in the direction of gravity. The size exclusion coefficient is used to characterize the probability that the simulated particles are stuck and embolized by small blood vessels.

[0089] According to one embodiment of this application, the generation module 280 is specifically used to: determine the radioactivity within each voxel of the hepatic artery model based on the distribution of radioactive microspheres to obtain an activity distribution map; for a first region in the activity distribution map, calculate the absorbed radiation dose using a graphics processor-accelerated Monte Carlo simulation method to obtain the dose distribution of the tumor region, wherein the first region is used to characterize the tumor region; for a second region in the activity distribution map, calculate the absorbed radiation dose using a dose convolution kernel method to obtain the dose distribution of the normal liver tissue region, wherein the second region is used to characterize the normal liver tissue region; and merge the dose distribution of the tumor region with the dose distribution of the normal liver tissue region to obtain a voxel-level radioactive microsphere dose distribution map.

[0090] It should be noted that the above-described embodiments and explanations of the beneficial effects of the method for calculating the dose of radioactive microspheres also apply to the radioactive microsphere dose calculation device in the embodiments of this application. To avoid redundancy, they will not be elaborated in detail here.

[0091] Corresponding to the above embodiments, this application also proposes a computer-readable storage medium.

[0092] The computer-readable storage medium of this application stores a program for calculating the dose of radioactive microspheres, which, when executed by a processor, implements the aforementioned method for calculating the dose of radioactive microspheres.

[0093] It should be noted that the above-described embodiments and explanations of the beneficial effects of the method for calculating the dose of radioactive microspheres are also applicable to the computer-readable storage medium of the embodiments of this application. To avoid redundancy, they will not be elaborated in detail here.

[0094] Corresponding to the above embodiments, this application also proposes an electronic device.

[0095] See Figure 3 As shown, the electronic device 300 of this application includes a memory 310, a processor 320, and a radioactive microsphere dose calculation program stored in the memory 310 and executable on the processor 320. When the processor executes the radioactive microsphere dose calculation program, it implements the aforementioned radioactive microsphere dose calculation method.

[0096] It should be noted that the above-described embodiments and explanations of the beneficial effects of the method for calculating the dose of radioactive microspheres also apply to the electronic devices in the embodiments of this application. To avoid redundancy, they will not be elaborated in detail here.

[0097] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0098] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0099] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example 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.

[0100] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0101] In this application, unless otherwise expressly specified and limited, the terms "installation," "connection," "joining," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0102] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for calculating the dose of radioactive microspheres, characterized in that, include: Acquire multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object, and construct a hepatic artery model based on the multimodal three-dimensional liver images and the two-dimensional hepatic artery angiography images; Calculate the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography images; The hemodynamic field is used to drive simulated particle transport in order to obtain the predicted distribution of the simulated particles; The true distribution of simulated particles of the target object is obtained, and the predicted distribution is compared with the true distribution. Based on the comparison results, the model parameters of the fluid-simulated particle mapping model are corrected to obtain the corrected fluid-simulated particle mapping model. The modified fluid-simulated particle mapping model is modified based on the physical properties of the radioactive microspheres to obtain the fluid-radioactive microsphere mapping model. The radioactive microspheres are released into the hepatic artery model using the hemodynamic field as the transport environment, and the motion trajectory and deposition position of the radioactive microspheres driven by the hemodynamic field are calculated using the fluid-radioactive microsphere mapping model to obtain the distribution of radioactive microspheres. Based on the aforementioned distribution of radioactive microspheres, a voxel-level radioactive microsphere dose distribution map is generated.

2. The method for calculating the dose of radioactive microspheres according to claim 1, characterized in that, The construction of the hepatic artery model based on the multimodal three-dimensional liver image and the two-dimensional hepatic artery angiography image includes: The multimodal three-dimensional liver images are registered based on a preset registration method to obtain spatially aligned multimodal three-dimensional liver images. The spatially aligned multimodal 3D liver image is input into a preset segmentation model to output an initial hepatic artery model; The initial hepatic artery model is fused with the two-dimensional hepatic artery angiography image to obtain the hepatic artery model.

3. The method for calculating the dose of radioactive microspheres according to claim 2, characterized in that, The preset registration method includes rigid registration and non-rigid registration. The rigid registration is used to characterize the elimination of positional differences between the multimodal three-dimensional liver images by finding rotation and translation matrices. The non-rigid registration is used to characterize the correction of physiological deformation of the liver caused by respiration or changes in body position by calculating deformable deformation fields.

4. The method for calculating the dose of radioactive microspheres according to claim 2, characterized in that, The process of fusing the initial hepatic artery model with the two-dimensional hepatic artery angiography image to obtain the hepatic artery model includes: Based on the C-arm rotation angle parameters of the hepatic artery angiography device, an initial projection relationship is established between the initial hepatic artery model and the two-dimensional hepatic artery angiography image. Based on the initial projection relationship, the initial hepatic artery model is subjected to virtual X-ray projection to generate a simulated two-dimensional hepatic artery angiography image. The similarity between the simulated two-dimensional hepatic artery angiography image and the real two-dimensional hepatic artery angiography image is measured to obtain the similarity measurement result. The position and angle of the initial hepatic artery model are adjusted based on the similarity measurement results until the simulated two-dimensional hepatic artery angiography image matches the real two-dimensional hepatic artery angiography image, thus obtaining the hepatic artery model.

5. The method for calculating the dose of radioactive microspheres according to claim 1, characterized in that, The calculation of the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography image includes: The hepatic artery model is divided into a grid to obtain a computational grid, and blood physical parameters are assigned to the computational grid. The time-density curve of the contrast agent passing through the hepatic artery vascular tree for the first time is extracted from the two-dimensional hepatic artery angiography image, and the waveform of blood flow velocity or volumetric flow rate changing with time is calculated based on the time-density curve as the entry boundary condition of the hepatic artery vascular tree; Based on the diameter of the exit branch of the hepatic artery vascular tree and the tissue perfusion requirements, the exit resistance or exit flow rate ratio of the exit of the hepatic artery vascular tree is set using Murray's law or preset perfusion parameters as the exit boundary condition of the hepatic artery vascular tree. Based on the inlet boundary conditions, the outlet boundary conditions, the Navier-Stokes equations, and the mass conservation equations, numerical solutions are performed on a computational grid endowed with blood physical parameters to obtain the hemodynamic field within the hepatic artery model.

6. The method for calculating the dose of radioactive microspheres according to claim 1, characterized in that, The model parameters of the fluid-simulated particle mapping model include inertia coefficient, sedimentation coefficient, and size exclusion coefficient. The inertia coefficient is used to characterize the degree to which the simulated particle deviates from the streamline due to inertia at the bifurcation or bend of the blood vessel. The sedimentation coefficient is used to characterize the deposition tendency of the simulated particle in the direction of gravity. The size exclusion coefficient is used to characterize the probability that the simulated particle is stuck and embolized by a small blood vessel.

7. The method for calculating the dose of radioactive microspheres according to claim 1, characterized in that, The generation of a voxel-level radioactive microsphere dose distribution map based on the distribution of the radioactive microspheres includes: The radioactivity within each voxel of the hepatic artery model was determined based on the distribution of the radioactive microspheres to obtain an activity distribution map; For the first region in the activity distribution map, the radiation absorbed dose is calculated using a graphics processor-accelerated Monte Carlo simulation method to obtain the dose distribution of the tumor region, wherein the first region is used to characterize the tumor region. For the second region in the activity distribution map, the radiation absorbed dose is calculated using the dose convolution kernel method to obtain the dose distribution of the normal liver tissue region, wherein the second region is used to characterize the normal liver tissue region; The dose distribution in the tumor region is combined with the dose distribution in the normal liver tissue region to obtain the dose distribution map of the voxel-level radioactive microspheres.

8. A device for calculating the dose of radioactive microspheres, characterized in that, include: The acquisition module is used to acquire multimodal three-dimensional liver images and two-dimensional hepatic artery angiography images of the target object; The construction module is used to construct a hepatic artery model based on the multimodal three-dimensional liver image and the two-dimensional hepatic artery angiography image; The first calculation module is used to calculate the hemodynamic field within the hepatic artery model based on the two-dimensional hepatic artery angiography image. A driving module is used to drive the simulated particle transport using the hemodynamic field to obtain the predicted distribution of the simulated particles; The first correction module is used to obtain the true distribution of the simulated particles of the target object, compare the predicted distribution with the true distribution, and correct the model parameters of the fluid-simulated particle mapping model based on the comparison results to obtain the corrected fluid-simulated particle mapping model. The second correction module is used to correct the modified fluid-simulated particle mapping model based on the physical properties of the radioactive microspheres to obtain the fluid-radioactive microsphere mapping model. The second calculation module is used to release the radioactive microspheres into the hepatic artery model using the hemodynamic field as the transport environment, and to calculate the motion trajectory and deposition position of the radioactive microspheres under the drive of the hemodynamic field using the fluid-radioactive microsphere mapping model, so as to obtain the distribution of radioactive microspheres. The generation module is used to generate a voxel-level radioactive microsphere dose distribution map based on the distribution of the radioactive microspheres.

9. A computer-readable storage medium, characterized in that, It stores a program for calculating the dose of radioactive microspheres, which, when executed by a processor, implements the method for calculating the dose of radioactive microspheres according to any one of claims 1-7.

10. An electronic device, characterized in that, The method includes a memory, a processor, and a radioactive microsphere dose calculation program stored in the memory and capable of running on the processor. When the processor executes the radioactive microsphere dose calculation program, it implements the radioactive microsphere dose calculation method according to any one of claims 1-7.

Images