An abdominal cavity organ digital twin model construction system
By acquiring multimodal medical images and generating biomechanical property fields using cross-modal mapping, combined with hierarchical heterogeneous modeling, a simulation model of abdominal organs is constructed, solving the problem of the lack of biomechanical properties in existing models and realizing high-precision surgical simulation and efficient computation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HARBIN SIZHERUI INTELLIGENT MEDICAL EQUIP CO LTD
- Filing Date
- 2026-03-16
- Publication Date
- 2026-06-12
Smart Images
Figure CN122199868A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical image simulation technology, and more specifically, to a system for constructing digital twin models of abdominal organs. Background Technology
[0002] In the technological development of laparoscopic surgical robot-assisted minimally invasive surgery, physics is gradually being applied to surgical planning scenarios through physical simulation technology. By using basic geometric models to carry out simple surgical operation simulations, it provides certain technical support for reducing surgical trauma and improving operational accuracy, becoming an important direction for the development of medical digitalization and intelligence.
[0003] However, most existing 3D reconstruction models are static models that only have a visual appearance and lack the characterization of inherent biomechanical properties such as tissue elasticity and viscosity. This leads to deviations in the physical simulation results of surgical operations and affects the accuracy of surgical simulation. Summary of the Invention
[0004] The problem addressed by this invention is how to improve the efficiency of constructing digital twin models of abdominal organs.
[0005] To address the above problems, this invention provides a system for constructing digital twin models of abdominal organs.
[0006] In a first aspect, the present invention provides a digital twin model construction system for abdominal organs, comprising: The acquisition module is used to acquire multimodal medical images of the target area of the patient's abdominal cavity; The mapping module is used to obtain the biomechanical property field corresponding to the organ structure based on the multimodal medical image data through a cross-modal mapping method; The modeling module is used to construct a simulation model of the abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures.
[0007] Optionally, obtaining the biomechanical property field corresponding to the organ structure based on the multimodal medical image data through a cross-modal mapping method includes: Based on the multimodal medical image data, a three-dimensional segmentation mask corresponding to the organ structure is generated using a trained segmentation model; Based on the regions corresponding to the organ structures in the multimodal medical imaging data, radiomics features corresponding to the organ structures are extracted. Based on the three-dimensional segmentation mask corresponding to the organ structure and the radiomics features, personalized correction data corresponding to the organ structure is obtained through a trained mapping model. Based on the personalized correction data of the organ structure, the biomechanical property field corresponding to the abdominal cavity structure is obtained through exponential correction.
[0008] Optionally, the step of obtaining the biomechanical property field corresponding to the abdominal cavity structure by exponential correction based on the personalized correction data of the organ structure includes: The correction amount for each voxel in the corresponding region of the abdominal cavity structure is obtained based on the personalized correction amount data; The personalized biomechanical parameters corresponding to the voxel are obtained through a preset exponential correction relationship based on the correction amount corresponding to the voxel and the preset initial biomechanical parameters. The biomechanical property field corresponding to the organ structure is generated based on all the personalized biomechanical parameters.
[0009] Optionally, the exponential correction relationship satisfies: E=E0×10 E ; Wherein, E represents the personalized biomechanical parameter, and E0 represents the initial biomechanical parameter. E is the correction amount.
[0010] Optionally, the step of constructing a simulation model of the abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures includes: The three-dimensional segmentation mask corresponding to all the organ structures is fused and partitioned to obtain the tetrahedral model corresponding to the abdominal organs; The tetrahedral model is divided into a core area model and a non-core area model using a preset partitioning rule; The non-core region model is topologically simplified to generate a non-core region mass point spring model; The core region model and the non-core region mass spring model are spliced and fused to obtain the abdominal organ simulation model.
[0011] Optionally, dividing the tetrahedral model into a core region model and a non-core region model using a preset partitioning rule includes: Obtain the lesion region in the tetrahedral model; Based on a preset expansion distance, the lesion region is subjected to three-dimensional morphological expansion to obtain the core area model; The regions in the tetrahedral model other than the core region model are defined as non-core region models.
[0012] Optionally, the step of splicing and fusing the core region model and the non-core region mass spring model to obtain the abdominal organ simulation model includes: The core area boundary nodes are obtained based on the core area model. Based on the non-core region mass point spring model, obtain the non-core region boundary mass points corresponding to the core region model; Based on all the core area boundary nodes and the non-core area boundary particles, a splicing and fusion rule is constructed by distance retrieval, wherein the splicing and fusion rule includes a one-to-one correspondence between the core area boundary nodes and the non-core area boundary particles; The core region model and the non-core region mass spring model are spliced and fused according to the splicing and fusion rules to obtain the abdominal organ simulation model.
[0013] Optionally, obtaining the tetrahedral model corresponding to the abdominal organs by fusing and partitioning the three-dimensional segmentation masks corresponding to all the organ structures includes: A joint segmentation mask is obtained by Boolean fusion based on the three-dimensional segmentation masks corresponding to all the organ structures described. The tetrahedral model is obtained by partitioning the joint segmentation mask into a three-dimensional mesh.
[0014] Optionally, obtaining the tetrahedral model by three-dimensional meshing based on the joint segmentation mask includes: Based on the three-dimensional surface contour of the abdominal organ extracted from the joint segmentation mask, a corresponding closed triangular patch surface mesh is generated. Using the closed triangular facet surface mesh as boundary constraints, the three-dimensional space of the abdominal organ covered by the joint segmentation mask is divided into volume meshes to obtain the tetrahedral model corresponding to the abdominal organ.
[0015] Optionally, before acquiring multimodal medical images of the patient's target region, the process includes: Acquire computed tomography (CT) images and magnetic resonance images of the target region; The multimodal medical image is obtained by standardizing the computed tomography image and the magnetic resonance image, wherein the standardization preprocessing includes at least spatial alignment, resolution unification and noise filtering.
[0016] The beneficial effects of the abdominal organ digital twin model construction system of the present invention are as follows: It acquires multimodal medical images of the target area of the patient's abdominal cavity, overcoming the information limitations of single images. Through the complementary imaging of different imaging principles in multimodal images, it can present multi-dimensional and accurate information such as organ anatomical structure, tissue texture, and diffusion characteristics, laying a high-quality raw data foundation for subsequent accurate modeling and avoiding modeling bias caused by single data from the outset. Based on the multimodal medical image data, it obtains the biomechanical property field corresponding to the organ structure through a cross-modal mapping method, breaking through the technical bottleneck of conventional clinical images being unable to non-invasively obtain tissue mechanical parameters. It assigns patient-specific biomechanical parameters to each voxel or mesh vertex of the organ, constructing a model that closely resembles the actual living tissue. The unique biomechanical property field compensates for the core deficiency of existing models that lack biomechanical properties, allowing the model to reproduce the inherent properties of tissues such as elasticity and viscosity. Based on the biomechanical property fields corresponding to the structures of all organs, a hierarchical heterogeneous construction of the abdominal organ simulation model is carried out. The personalized biomechanical property field is combined with the hierarchical heterogeneous construction modeling strategy. A high-precision finite element model is used in the core surgical area and directly associated with exclusive biomechanical parameters. At the same time, a dynamic constraint coupling mechanism between high-precision and low-precision models is established to ensure the continuity of force and displacement transmission. This not only ensures the accuracy of the simulation of key surgical areas, but also avoids simulation distortion caused by unreasonable model construction. The model can realistically simulate the physical deformation and structural displacement of tissues during surgical operations, which greatly improves the overall simulation accuracy. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of a digital twin model construction system for abdominal organs according to an embodiment of the present invention. Detailed Implementation
[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.
[0019] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.
[0020] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.
[0021] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".
[0022] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.
[0023] In related technologies, existing 3D reconstruction models are mostly static models that only possess visual appearance, lacking the characterization of inherent biomechanical properties such as tissue elasticity and viscosity. This leads to deviations in the physical simulation results of surgical operations, affecting the accuracy of surgical simulation. These models can only present the geometric shape of abdominal organs and cannot reproduce the real mechanical response of living tissue under surgical traction, cutting, and suturing operations. They cannot accurately simulate the amplitude and trend of tissue deformation, nor can they predict the displacement trajectory of delicate structures such as blood vessels and bile ducts. Surgical simulations based on such models differ significantly from the physical interaction process in actual surgery, rendering the simulation results worthless. They not only fail to provide doctors with accurate predictions of surgical operations but also make it difficult to conduct effective surgical risk assessments. Ultimately, this significantly reduces the scientific validity and reliability of preoperative planning, making it difficult to meet the clinical needs of laparoscopic minimally invasive surgery for high-precision simulation.
[0024] To address the problems existing in the aforementioned related technologies, this invention provides a system for constructing digital twin models of abdominal organs.
[0025] like Figure 1 As shown, an embodiment of the present invention provides a digital twin model construction system 100 for abdominal organs, comprising: The acquisition module 110 is used to acquire multimodal medical images of the target area of the patient's abdominal cavity.
[0026] Specifically, this involves acquiring multimodal medical images of the target abdominal region. For the abdominal operating area required for laparoscopic surgery planning, at least two different types of preoperative clinical medical images using varying imaging principles are collected. Each image presents the characteristic information of abdominal tissues and organs from different dimensions, forming a complementary image data foundation. This provides comprehensive and accurate raw data for subsequent modeling, with the acquisition scope fully covering the core anatomical structures involved in the surgery, such as the liver, gallbladder, target lesion, major arteries and veins, and bile ducts. For example, for the preoperative planning needs of partial hepatectomy, enhanced computed tomography (CT) images and T2-weighted magnetic resonance imaging (MRI) images of the patient's abdomen are acquired. Enhanced CT images clearly present the anatomical structure, tissue density distribution, and vascular course of abdominal organs, while T2-weighted MRI images accurately reflect details such as the texture features of soft tissues and water molecule diffusion characteristics. The combination of these two images achieves complete acquisition of multidimensional information about the target abdominal region.
[0027] The mapping module 120 is used to obtain the biomechanical property field corresponding to the organ structure based on the multimodal medical image data through a cross-modal mapping method.
[0028] Specifically, based on multimodal medical imaging data, the biomechanical property field corresponding to the organ structure is obtained through cross-modal mapping. First, radiomics features that reflect tissue texture and diffusion characteristics are extracted from the magnetic resonance images in the multimodal medical imaging. Then, based on the three-dimensional geometric segmentation results of the organ, nominal biomechanical parameters matching the tissue type are assigned to each voxel or mesh vertex. Subsequently, the extracted radiomics features are input into a trained generative adversarial network mapper, which learns the nonlinear mapping relationship from the image feature space to the biomechanical parameter increment space and outputs personalized correction values. Finally, the nominal biomechanical parameters and personalized correction values are superimposed to assign exclusive biomechanical parameters to each voxel or mesh vertex of the organ structure, forming a biomechanical property field that covers the entire organ and has patient specificity. Taking the modeling scenario of partial hepatectomy as an example, a total of 500 radiomics features, such as gray-level co-occurrence matrix features and wavelet features, are first extracted from the liver region of the patient's T2-weighted MRI images. Initial Young's modulus nominal values of 10 kPa, 30 kPa, and 100 kPa are assigned to the liver parenchyma, intrahepatic tumor, and blood vessel wall, respectively. Then, the radiomics features are input into a conditional generative adversarial network mapper to obtain a personalized correction amount of the logarithm of Young's modulus. Finally, a Young's modulus attribute field specific to the patient's liver tissue is generated through calculation, which fully presents the differences in biomechanical properties of different parts of the liver.
[0029] Modeling module 130 is used to construct a simulation model of abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures.
[0030] Specifically, based on the biomechanical property fields corresponding to all organ structures, a hierarchical heterogeneous simulation model of the abdominal organs is constructed. This involves relying on the patient-specific biomechanical property fields of different organ parts and employing a hierarchical heterogeneous modeling strategy to build a physical computational model adapted to surgical simulation. Differentiated modeling is used to address the simulation accuracy and computational efficiency requirements of different surgical regions. Simultaneously, a dynamic constraint coupling mechanism is established at the interface between models of different accuracy levels to ensure continuous transmission of force and displacement at the interface, achieving a balance between simulation accuracy and computational efficiency. In specific modeling, the core surgical region can be represented using a high-precision finite element model, with its mesh vertices directly linked to the personalized parameters of the biomechanical property field. The surrounding tissues outside the core region are represented using a computationally more efficient point mass spring model, whose elasticity... The spring stiffness coefficient is derived from the average value of the biomechanical properties of the corresponding region. Taking the modeling of partial liver resection as an example, the surgical core area is defined by extending 2cm outward from the tumor boundary. The liver mesh in this area is kept in its original state to construct a high-precision finite element model. Each mesh vertex is matched with a specific biomechanical parameter such as Young's modulus. For liver tissue outside the core area, the mesh is topologically simplified, and the mesh vertices are regarded as mass points and the mesh edges are regarded as springs. The spring stiffness is calculated based on the average Young's modulus of the corresponding region to construct a mass spring model. At the same time, at the interface between the core area and the non-core area, a linear constraint equation is established between each interface mass point and its nearest node on the finite element model to ensure that the displacement of the two is coordinated when under force. Finally, a liver organ simulation model that balances accuracy and efficiency is formed.
[0031] In this embodiment, multimodal medical images of the target area of the patient's abdominal cavity are acquired, overcoming the information limitations of single images. By complementing images with different imaging principles in the multimodal images, precise information such as organ anatomy, tissue texture, and diffusion characteristics can be presented, laying a high-quality raw data foundation for subsequent accurate modeling and avoiding modeling bias caused by single data from the source. Based on the multimodal medical image data, the biomechanical property field corresponding to the organ structure is obtained through cross-modal mapping, breaking through the technical bottleneck that conventional clinical images cannot non-invasively obtain tissue mechanical parameters. Patient-specific biomechanical parameters are assigned to each voxel or mesh vertex of the organ, constructing a biomechanical property field that closely matches the real characteristics of living tissue. This approach addresses the core deficiency of existing models that lack biomechanical properties, enabling models to reproduce the inherent properties of tissues such as elasticity and viscosity. Based on the biomechanical property fields corresponding to all organ structures, a hierarchical heterogeneous simulation model of the abdominal organs is constructed. This combines personalized biomechanical property fields with a hierarchical heterogeneous modeling strategy. A high-precision finite element model is used in the core surgical area and directly associated with specific biomechanical parameters. Simultaneously, a dynamic constraint coupling mechanism between high- and low-precision models is established to ensure the continuity of force and displacement transmission. This not only guarantees the accuracy of the simulation in key surgical areas but also avoids simulation distortion caused by unreasonable model construction. The model can realistically simulate the physical deformation and structural displacement of tissues during surgical operations, significantly improving the overall simulation accuracy.
[0032] Optionally, obtaining the biomechanical property field corresponding to the organ structure based on the multimodal medical image data through a cross-modal mapping method includes: Based on the multimodal medical image data, a three-dimensional segmentation mask corresponding to the organ structure is generated using a trained segmentation model; Based on the regions corresponding to the organ structures in the multimodal medical imaging data, radiomics features corresponding to the organ structures are extracted. Based on the three-dimensional segmentation mask corresponding to the organ structure and the radiomics features, personalized correction data corresponding to the organ structure is obtained through a trained mapping model. Based on the personalized correction data of the organ structure, the biomechanical property field corresponding to the abdominal cavity structure is obtained through exponential correction.
[0033] Specifically, based on multimodal medical image data, a trained segmentation model generates 3D segmentation masks corresponding to organ structures. This involves using a deep convolutional neural network segmentation model pre-trained on a publicly available dataset and fine-tuned with a local medical dataset to automatically segment pre-processed multimodal medical images using multiple labels. This accurately identifies and extracts 3D binary masks of target organ structures such as the liver, lesions, blood vessels, and bile ducts within the abdominal cavity. Manual verification and correction of segmentation deviations ensure the accuracy of organ geometry. Furthermore, radiomics features corresponding to organ structures are extracted from the regions corresponding to organ structures in the multimodal medical image data. This involves extracting various radiomics features, such as gray-level co-occurrence matrix features and wavelet features, from the target organ regions of magnetic resonance imaging (MRI) images in multimodal images, reflecting tissue texture and diffusion characteristics, thereby mining microscopic information about the tissue. Based on the 3D segmentation mask and radiomics features corresponding to organ structures, personalized correction data corresponding to organ structures are obtained through a trained mapping model. This refers to assigning nominal biomechanical parameters matching the tissue type to each voxel or mesh vertex of the organ based on the 3D segmentation mask. The extracted radiomics features are then input into the trained mapping model (generative adversarial network mapper), which learns the nonlinear mapping relationship between image features and biomechanical parameter increments, and outputs personalized corrections for the nominal parameters. Based on the personalized correction data of organ structures, the biomechanical attribute field corresponding to the abdominal cavity structure is obtained through exponential correction. That is, the nominal biomechanical parameters and personalized corrections are superimposed and calculated through exponential operation, assigning patient-specific biomechanical parameters to each voxel or mesh vertex of the organ, forming a biomechanical attribute field covering the entire organ.
[0034] For example, taking partial hepatectomy modeling as an example, a 3DU-Net model, pre-trained on a pre-defined dataset and fine-tuned on a local liver tumor dataset, is first used to segment enhanced CT images, generating three-dimensional segmentation masks for structures such as liver parenchyma, intrahepatic tumors, and hepatic vessels. These masks are then manually corrected by the physician to correct any deviations. Next, 500 radiomics features are extracted from the liver region of the patient's T2-weighted MRI images, and nominal Young's modulus values of 10 kPa, 30 kPa, and 100 kPa are assigned to the liver parenchyma, tumor, and vessel walls, respectively. The three-dimensional segmentation masks and radiomics features are then input into a pre-trained conditional generative adversarial network mapper to obtain personalized correction data for the logarithm of Young's modulus. Finally, an exponential correction is used to generate a Young's modulus attribute field specific to the patient's liver. In this optional embodiment, personalized mechanical property field data is obtained by performing cross-membrane mapping on multimodal medical images. This not only ensures the accuracy of organ geometric structure segmentation, but also breaks through the technical bottleneck that conventional clinical images cannot non-invasively obtain personalized biomechanical parameters. This provides a biomechanical property basis that fits the real characteristics of the patient's living tissue for the construction of abdominal organ simulation models. It effectively makes up for the lack of biomechanical properties in existing static models and improves the accuracy and realism of subsequent surgical simulations from the core property level.
[0035] Optionally, the step of obtaining the biomechanical property field corresponding to the abdominal cavity structure by exponential correction based on the personalized correction data of the organ structure includes: The correction amount for each voxel in the corresponding region of the abdominal cavity structure is obtained based on the personalized correction amount data; The personalized biomechanical parameters corresponding to the voxel are obtained through a preset exponential correction relationship based on the correction amount corresponding to the voxel and the preset initial biomechanical parameters. The biomechanical property field corresponding to the organ structure is generated based on all the personalized biomechanical parameters.
[0036] Optionally, the exponential correction relationship satisfies: E=E0×10 E ; Wherein, E represents the personalized biomechanical parameter, and E0 represents the initial biomechanical parameter. E is the correction amount.
[0037] Specifically, the correction amount for each voxel in the corresponding region of the abdominal structure is obtained based on personalized correction data. That is, the personalized correction data output by the trained mapping model is precisely allocated to each voxel in the corresponding region of the abdominal organ structure according to the spatial location of the voxel, and each voxel is assigned a unique corresponding biomechanical parameter correction value, realizing voxel-level precise mapping of the correction amount. Based on the correction amount corresponding to the voxel and the preset initial biomechanical parameters, the personalized biomechanical parameters corresponding to the voxel are obtained through a preset exponential correction relationship. This is to match the pre-set nominal biomechanical initial parameters to voxels of different tissue types, and then combine the initial parameters with the voxel-specific correction amount through the exponential correction relationship, completing the transformation from general nominal parameters to patient-specific individual parameters. Based on all personalized biomechanical parameters, the biomechanical attribute field corresponding to the organ structure is generated. That is, the personalized biomechanical parameters of all voxels in the organ structure are integrated to form a biomechanical attribute field covering the entire organ, with spatial continuity and patient specificity, fully presenting the differences in mechanical properties of different parts of the organ.
[0038] For example, taking the modeling of partial hepatectomy as an example, the personalized correction data of the logarithm of Young's modulus output by the mapping model is assigned to each voxel in the liver region according to its spatial location. For example, the voxels of liver parenchyma, intrahepatic tumor, and vascular wall are matched with the initial Young's modulus parameters of 10 kPa, 30 kPa, and 100 kPa, respectively. Then, through the preset exponential correction relationship formula, the exclusive personalized Young's modulus parameter, i.e., personalized biomechanical parameter, is calculated for each voxel. Finally, the personalized Young's modulus parameters of all liver voxels are integrated to generate the exclusive Young's modulus attribute field of the patient's liver, i.e., the biomechanical attribute field. The biomechanical attribute field can clearly reflect the differences in mechanical properties of liver parenchyma, tumor, vascular wall and surrounding tissues.
[0039] In this optional embodiment, the precise conversion of biomechanical parameters from the overall nominal to the voxel-level personalized model is realized. The scientific and rational nature of parameter correction is ensured through the exponential correction relationship. The generated biomechanical property field has both patient-specific and spatially refined characteristics, accurately restoring the real mechanical properties of different tissues in the living abdominal cavity. This makes up for the defects of the existing technology in the generalization of biomechanical parameters and the inability to reflect individual differences and regional characteristics. It lays the core mechanical property foundation for the subsequent construction of high-fidelity abdominal organ simulation models, and greatly improves the accuracy and realism of tissue deformation and structural displacement simulation in surgical simulation.
[0040] Optionally, the step of constructing a simulation model of the abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures includes: The three-dimensional segmentation mask corresponding to all the organ structures is fused and partitioned to obtain the tetrahedral model corresponding to the abdominal organs; The tetrahedral model is divided into a core area model and a non-core area model using a preset partitioning rule; The non-core region model is topologically simplified to generate a non-core region mass point spring model; The core region model and the non-core region mass spring model are spliced and fused to obtain the abdominal organ simulation model.
[0041] Specifically, the three-dimensional segmentation masks corresponding to all organ structures are fused and partitioned to obtain tetrahedral models corresponding to the abdominal organs. This involves spatially fusing the three-dimensional segmentation masks of various organs within the abdominal cavity, such as the liver, lesions, blood vessels, and bile ducts. Then, a professional meshing algorithm is used to divide the fused overall organ geometry into tetrahedral meshes, generating a tetrahedral model with both geometric accuracy and topological rationality, thus establishing the basic geometric framework for physical simulation. The tetrahedral model is divided into core and non-core areas using preset partitioning rules. These rules are based on surgical operation requirements and typically include the lesion and surrounding area... A defined area of tissue is designated as the surgical core region, while the remaining tissues are designated as non-core regions, enabling differentiated construction of the model across different regions. The non-core region model is topologically simplified to generate a non-core region mass-spring model. This involves simplifying the tetrahedral mesh of the non-core region, treating mesh vertices as mass points and mesh edges as springs, thus transforming it into a computationally more efficient mass-spring model. The core region model and the non-core region mass-spring model are then combined to obtain the abdominal organ simulation model. A dynamic constraint coupling mechanism is established at the interface between the two types of models to ensure continuous transmission of force and displacement at the interface, achieving seamless fusion of high- and low-precision models.
[0042] For example, taking the modeling of partial hepatectomy as an example, the three-dimensional segmentation masks of liver parenchyma, intrahepatic tumor, and hepatic blood vessels are first fused. The overall geometric structure of the fused liver is then divided into tetrahedral meshes to generate a liver tetrahedral model. Then, according to the preset rules for the outward expansion of the tumor boundary, the tetrahedral mesh within this range is defined as the core area model, and the remaining areas are non-core area models. The non-core area models are topologically simplified, with mesh vertices set as mass points and mesh edges as springs. The spring stiffness is calculated based on the average Young's modulus of the corresponding area to generate a mass point spring model. Finally, the core area tetrahedral model and the non-core area mass point spring model are spliced together, and a linear constraint equation is established at the interface between each mass point and its nearest node on the core area model to complete the construction of the liver organ simulation model.
[0043] In this optional embodiment, the integrity and accuracy of the abdominal organ geometric model are ensured by fusion subdivision. The regional modeling strategy based on surgical needs takes into account both the simulation accuracy of the core surgical area and the computational efficiency of the non-core area. The combination of topology simplification and dynamic constraint coupling realizes the seamless integration of high and low accuracy models. The constructed simulation model can accurately reproduce the physical response of the tissues in the key surgical area, and significantly reduce the computational complexity of the overall simulation. It effectively balances the high fidelity and real-time performance of the simulation, and solves the technical problem of difficulty in balancing accurate modeling and efficient computation in the prior art. It provides reliable model support for high-fidelity pre-operative simulation and accurate preoperative planning of laparoscopic surgery.
[0044] Optionally, dividing the tetrahedral model into a core region model and a non-core region model using a preset partitioning rule includes: Obtain the lesion region in the tetrahedral model; Based on a preset expansion distance, the lesion region is subjected to three-dimensional morphological expansion to obtain the core area model; The regions in the tetrahedral model other than the core region model are defined as non-core region models.
[0045] Specifically, the lesion region in the tetrahedral model is obtained by precisely extracting the mesh region that perfectly matches the location and shape of the lesion from the tetrahedral mesh model of the abdominal organs, based on the lesion identification information of the previous three-dimensional segmentation mask of the organ structure. This achieves accurate positioning and division of the lesion region in the tetrahedral model. Based on a preset expansion distance, the lesion region is subjected to three-dimensional morphological expansion to obtain the core area model. Through a three-dimensional morphological expansion algorithm, the lesion region is uniformly expanded to the surrounding space according to the expansion distance set in the preoperative plan, incorporating the lesion and the surrounding tissues that need high-precision simulation into the core area, forming a complete core area tetrahedral model. The areas in the tetrahedral model other than the core area model are defined as non-core area models. Using the core area model as the boundary, the entire tetrahedral model is divided into regions. All organ mesh regions outside the core area are defined as non-core area models, achieving accurate division of the core and non-core areas of the abdominal organ simulation model.
[0046] For example, taking the modeling of partial liver resection as an example, the mesh lesion area corresponding to the tumor is accurately extracted from the liver tetrahedral model according to the segmentation mask information of the intrahepatic tumor. The tumor lesion area is then subjected to three-dimensional morphological expansion according to a preset expansion distance of 2cm, so that the tumor and the surrounding liver tissue mesh within a 2cm range are all included to form the core area model of the liver surgery. Then, the remaining liver tissue mesh areas in the liver tetrahedral model other than the core area are determined as non-core area models.
[0047] In this optional embodiment, the lesion area is rapidly located through a precise segmentation mask, and the core area is delineated through a standardized operation of three-dimensional morphological expansion. This ensures the accuracy, consistency, and scientific nature of the core area delineation, allowing the core area to completely cover the key areas of the surgical operation. At the same time, it achieves a clear and non-overlapping division between the core and non-core areas, laying a precise regional division foundation for the subsequent hierarchical and heterogeneous construction of high- and low-precision simulation models. This ensures that high-precision models are used to simulate key surgical areas to guarantee simulation accuracy, while lightweight models are used to improve computational efficiency for non-critical areas, effectively balancing the accuracy and computational speed of the abdominal organ simulation model.
[0048] Optionally, the step of splicing and fusing the core region model and the non-core region mass spring model to obtain the abdominal organ simulation model includes: The core area boundary nodes are obtained based on the core area model. Based on the non-core region mass point spring model, obtain the non-core region boundary mass points corresponding to the core region model; Based on all the core area boundary nodes and the non-core area boundary particles, a splicing and fusion rule is constructed by distance retrieval, wherein the splicing and fusion rule includes a one-to-one correspondence between the core area boundary nodes and the non-core area boundary particles; The core region model and the non-core region mass spring model are spliced and fused according to the splicing and fusion rules to obtain the abdominal organ simulation model.
[0049] Specifically, the core area boundary nodes are obtained based on the core area model. This involves extracting all mesh nodes located at the model edge, serving as the boundary between the core and non-core areas, from the high-precision finite element model of the core area of the abdominal cavity model, thus achieving accurate identification and extraction of the core area boundary nodes. The non-core area boundary particles are obtained based on the non-core area mass spring model, selecting particles adjacent to the core area boundary nodes and located at the edge of the non-core area from the non-core area mass spring model as the non-core area boundary particles matching the core area boundary. Finally, a splicing and fusion rule is constructed based on all core area boundary nodes and non-core area boundary particles through distance retrieval. This splicing and fusion rule includes both core area boundary nodes and non-core area boundary particles. The one-to-one correspondence of boundary mass points is established by calculating the spatial distance between each core area boundary node and the surrounding non-core area boundary mass points using a spatial distance retrieval algorithm. The nearest core area boundary node is paired with the non-core area boundary mass point to establish a one-to-one spatial constraint relationship, forming a standardized splicing and fusion rule. Based on the splicing and fusion rule, the core area model and the non-core area mass point spring model are spliced and fused to obtain the abdominal organ simulation model. According to the constructed one-to-one correspondence rule, linear constraint equations are established for each pair of matched core area boundary nodes and non-core area boundary mass points to ensure that the displacements of the two are coordinated and consistent under force, realizing seamless splicing and fusion of high and low precision models at the boundary to form a complete abdominal organ simulation model.
[0050] For example, taking the modeling of partial liver resection as an example, all boundary nodes of the liver core area are extracted from the finite element model of the core area defined by extending 2cm outward from the liver tumor. Then, the corresponding non-core area boundary mass points are selected from the mass spring model of the non-core area. By using distance retrieval, the nearest non-core area boundary mass point is matched for each core area boundary node, and a one-to-one splicing and fusion rule is established. Then, according to the rule, a linear constraint is established for each pair of matching nodes and mass points, so that the finite element model of the core area and the mass spring model of the non-core area can be seamlessly fused at the boundary, and finally a complete liver organ simulation model is obtained.
[0051] In this optional embodiment, the accuracy of model boundary matching is ensured by accurately extracting the boundary nodes and mass points of the core area and non-core area. The one-to-one correspondence splicing and fusion rules established by distance retrieval make the splicing of high and low precision models more scientific and standardized. The establishment of linear constraints effectively ensures the continuous transmission of force and displacement at the model interface, completely solving the problems of simulation distortion and discontinuous mechanical response that are prone to occur at the splicing of high and low precision models. This significantly improves the overall simulation accuracy and mechanical consistency of the abdominal organ simulation model, while maintaining the structural integrity of the layered heterogeneous simulation model, and taking into account both the high fidelity of the simulation of the surgical core area and the high computational efficiency of the non-core area.
[0052] Optionally, obtaining the tetrahedral model corresponding to the abdominal organs by fusing and partitioning the three-dimensional segmentation mask corresponding to all the organ structures includes: A joint segmentation mask is obtained by Boolean fusion based on the three-dimensional segmentation masks corresponding to all the organ structures described. The tetrahedral model is obtained by partitioning the joint segmentation mask into a three-dimensional mesh.
[0053] Specifically, a joint segmentation mask is obtained by Boolean fusion based on the three-dimensional segmentation masks corresponding to all organ structures. This involves superimposing and fusing the three-dimensional binary segmentation masks of different organ structures in the same spatial coordinate system using Boolean union operations to remove redundant and overlapping areas and fill structural gaps, resulting in a joint segmentation mask with continuous boundaries and complete coverage of the target organ's anatomical structure. Morphological optimization is then performed to repair defects such as burrs and holes at the boundaries, ensuring the regularity of the organ contours. Subsequently, the three-dimensional surface contours of the abdominal organs are extracted from the optimized joint segmentation mask, generating closed triangular facet surface meshes. Using this mesh as a rigid boundary constraint, the Delaunay three-dimensional meshing algorithm is used to perform volume meshing on the three-dimensional space corresponding to the mask, discretizing it into a large number of non-overlapping, topologically compatible tetrahedral elements. At the same time, the mesh size and element quality are controlled according to surgical requirements, and mesh refinement is performed on key surgical areas such as tumors and blood vessels, ultimately generating a high-precision tetrahedral model of the abdominal organs.
[0054] In this optional embodiment, Boolean fusion ensures the spatial integrity and anatomical accuracy of multi-organ structures. Delaunay meshing and mesh refinement in key areas improve the stability and computational efficiency of subsequent physical simulations while maintaining the geometric accuracy of the model. This effectively solves the misalignment and discontinuity problems that easily occur when splicing single-organ models, providing a geometrically accurate and topologically reasonable basic mesh carrier for building hierarchical heterogeneous simulation models. Optionally, obtaining the tetrahedral model by partitioning the joint segmentation mask into a three-dimensional mesh includes: Based on the three-dimensional surface contour of the abdominal organ extracted from the joint segmentation mask, a corresponding closed triangular patch surface mesh is generated. Using the closed triangular facet surface mesh as boundary constraints, the three-dimensional space of the abdominal organ covered by the joint segmentation mask is divided into volume meshes to obtain the tetrahedral model corresponding to the abdominal organ.
[0055] Specifically, firstly, from the optimized joint segmentation mask, surface extraction algorithms, such as the Moving Cubes (MC) algorithm, the Moving Tetrahedra (MT) algorithm, or region growing, are used to accurately extract the 3D surface contours of abdominal organs. These contours completely restore the external morphology of the organ and the surface features of its key internal structures (such as tumors and blood vessels). The Moving Cubes algorithm, by traversing the voxels of the 3D volumetric data and determining isosurfaces based on the voxel grayscale values (distinguishing between the target and background regions), extracts the organ surface contours. This algorithm is highly adaptable and can accurately restore the curved surface morphology of complex organs, making it suitable for processing relatively regular organs such as the liver. The Moving Tetrahedra algorithm, based on tetrahedral voxel data, generates a surface contour composed of triangular facets by determining the position of the isosurfaces within the voxels. Compared to the Moving Cubes algorithm, the Moving Tetrahedra algorithm provides a more precise extraction method. The Cubes algorithm is better suited for handling irregular voxel distributions, especially for organs with complex internal structures such as the liver. The region growing method uses target voxels in the joint segmentation mask as seed points and grows according to preset gray-level similarity or spatial adjacency rules, gradually expanding to form a complete organ surface contour. This method is suitable for handling organ regions with relatively blurred boundaries but coherent structures. Subsequently, triangular facet discretization is used to segment the continuous 3D surface contour into a large number of interconnected, non-overlapping, and gapless triangular facets. Simultaneously, the discretized triangular facets are optimized by removing redundant facets and vertices and repairing irregular facets, ensuring that the final generated triangular facet surface mesh is closed, continuous, and smooth, accurately matching the actual anatomical morphology of the organ. Using the surface mesh of the closed triangular facet as boundary constraints, the three-dimensional space of the abdominal organ covered by the joint segmentation mask is partitioned into a volume mesh. The three-dimensional Delaunay tetrahedral partitioning algorithm is adopted, with the surface mesh as a rigid boundary, to perform filling discretization of the three-dimensional space inside the organ, dividing the entire organ space into a large number of topologically compatible, non-overlapping, and non-void tetrahedral elements. During the partitioning process, the quality of the tetrahedral elements is controlled according to the needs of surgical simulation. The mesh is appropriately densified in key surgical areas such as tumors and blood vessels to improve the local partitioning accuracy, while the element size is appropriately increased in non-critical areas to balance accuracy and computational efficiency, and finally the tetrahedral model corresponding to the abdominal organ is obtained.
[0056] For example, taking partial hepatectomy modeling as an example, the three-dimensional surface contours of the liver and internal tumors and blood vessels are extracted from the liver joint segmentation mask. Closed triangular surface meshes are generated by the Marching Cubes algorithm. After repairing the mesh boundary burrs, the three-dimensional Delaunay tetrahedron subdivision algorithm is used as the boundary constraint to perform volume mesh subdivision of the liver in three-dimensional space. The mesh is densified for the intrahepatic tumor and the surrounding 2cm area, and the mesh is appropriately simplified for the remaining liver parenchyma areas. Finally, a geometrically accurate and simulation-adaptive liver tetrahedron model is generated.
[0057] In this optional embodiment, by accurately extracting the surface contour and generating an optimized closed triangular facet surface mesh, precise boundary constraints are provided for volume mesh generation, avoiding geometric distortion during the generation process. The three-dimensional Delaunay tetrahedral generation algorithm combined with a regional mesh refinement strategy ensures both the topological rationality and geometric accuracy of the tetrahedral model, enabling accurate reproduction of the internal structural features of organs, while also achieving a balance between simulation accuracy and computational efficiency. This provides a high-quality basic mesh carrier for subsequent layered heterogeneous modeling, biomechanical property mapping, and surgical simulation, effectively solving the simulation deviation problems caused by uneven mesh accuracy and boundary distortion in existing generation methods.
[0058] Optionally, before acquiring multimodal medical images of the patient's target region, the process includes: Acquire computed tomography (CT) images and magnetic resonance images of the target region; The multimodal medical image is obtained by standardizing the computed tomography image and the magnetic resonance image, wherein the standardization preprocessing includes at least spatial alignment, resolution unification and noise filtering.
[0059] Specifically, before acquiring multimodal medical images of the patient's target area, computed tomography (CT) and magnetic resonance imaging (MRI) images of the target area need to be acquired in a targeted manner. The two types of images present the characteristics of abdominal organs from the dimensions of tissue density, anatomical structure and soft tissue texture, and diffusion characteristics, respectively, providing complementary raw data for subsequent modeling. Then, the CT and MRI images are subjected to standardized preprocessing. Spatial alignment unifies the spatial coordinate system of different images, resolution unification achieves voxel specification consistency, and noise filtering reduces interference noise in the images. After a series of processing steps, standardized multimodal medical images are obtained, laying a high-quality data foundation for subsequent modeling steps.
[0060] For example, enhanced CT images and T2-weighted MRI images of the patient's abdomen are first acquired. Then, non-rigid registration algorithms such as affine transformation and B-spline transformation are used to align the spatial coordinate system of the MRI images with that of the CT images. The two types of images are then isotropically resampled to 1 mm. 3 Voxels are used to achieve uniform resolution, and nonlocal mean filtering is applied to denoise the images. After standardized preprocessing, usable multimodal medical images of the abdominal target region are obtained.
[0061] In this optional embodiment, by selectively acquiring two core medical images—computed tomography (CT) images and magnetic resonance imaging (MRI) images—the advantages of different imaging technologies are fully utilized, ensuring the comprehensiveness of the original image data. Standardized preprocessing eliminates spatial and resolution differences between different modal images, reduces noise interference on subsequent segmentation and feature extraction, and effectively improves the data quality and consistency of multimodal medical images. This avoids subsequent modeling deviations caused by inconsistent image specifications and noise interference from the data source, providing reliable standardized data support for the precise implementation of subsequent steps such as organ segmentation and biomechanical property mapping, significantly improving the accuracy and stability of the overall modeling process.
[0062] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.
Claims
1. A system for constructing digital twin models of abdominal organs, characterized in that, include: The acquisition module is used to acquire multimodal medical images of the target area of the patient's abdominal cavity; The mapping module is used to obtain the biomechanical property field corresponding to the organ structure based on the multimodal medical image data through a cross-modal mapping method; The modeling module is used to construct a simulation model of the abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures.
2. The abdominal organ digital twin model construction system according to claim 1, characterized in that, The step of obtaining the biomechanical property field corresponding to the organ structure through a cross-modal mapping method based on the multimodal medical image data includes: Based on the multimodal medical image data, a three-dimensional segmentation mask corresponding to the organ structure is generated using a trained segmentation model; Based on the regions corresponding to the organ structures in the multimodal medical imaging data, radiomics features corresponding to the organ structures are extracted. Based on the three-dimensional segmentation mask corresponding to the organ structure and the radiomics features, personalized correction data corresponding to the organ structure is obtained through a trained mapping model. Based on the personalized correction data of the organ structure, the biomechanical property field corresponding to the abdominal cavity structure is obtained through exponential correction.
3. The abdominal organ digital twin model construction system according to claim 2, characterized in that, The process of obtaining the biomechanical property field corresponding to the abdominal cavity structure by exponentially correcting the personalized correction data of the organ structure includes: The correction amount for each voxel in the corresponding region of the abdominal cavity structure is obtained based on the personalized correction amount data; The personalized biomechanical parameters corresponding to the voxel are obtained through a preset exponential correction relationship based on the correction amount corresponding to the voxel and the preset initial biomechanical parameters. The biomechanical property field corresponding to the organ structure is generated based on all the personalized biomechanical parameters.
4. The abdominal organ digital twin model construction system according to claim 3, characterized in that, The exponential correction relationship satisfies: E=E0×10 E ; Wherein, E represents the personalized biomechanical parameter, and E0 represents the initial biomechanical parameter. E is the correction amount.
5. The abdominal organ digital twin model construction system according to claim 2, characterized in that, The construction of a simulation model of abdominal organs through hierarchical heterogeneity based on the biomechanical property fields corresponding to all the organ structures includes: The three-dimensional segmentation mask corresponding to all the organ structures is fused and partitioned to obtain the tetrahedral model corresponding to the abdominal organs; The tetrahedral model is divided into a core area model and a non-core area model using a preset partitioning rule; The non-core region model is topologically simplified to generate a non-core region mass point spring model; The core region model and the non-core region mass spring model are spliced and fused to obtain the abdominal organ simulation model.
6. The abdominal organ digital twin model construction system according to claim 5, characterized in that, The step of dividing the tetrahedral model into a core region model and a non-core region model according to a preset partitioning rule includes: Obtain the lesion region in the tetrahedral model; Based on a preset expansion distance, the lesion region is subjected to three-dimensional morphological expansion to obtain the core area model; The regions in the tetrahedral model other than the core region model are defined as non-core region models.
7. The abdominal organ digital twin model construction system according to claim 5, characterized in that, The process of splicing and fusing the core region model and the non-core region mass spring model to obtain the abdominal organ simulation model includes: The core area boundary nodes are obtained based on the core area model. Based on the non-core region mass point spring model, obtain the non-core region boundary mass points corresponding to the core region model; Based on all the core area boundary nodes and the non-core area boundary particles, a splicing and fusion rule is constructed by distance retrieval, wherein the splicing and fusion rule includes a one-to-one correspondence between the core area boundary nodes and the non-core area boundary particles; The core region model and the non-core region mass spring model are spliced and fused according to the splicing and fusion rules to obtain the abdominal organ simulation model.
8. The abdominal organ digital twin model construction system according to claim 5, characterized in that, The step of obtaining a tetrahedral model of the abdominal organs by fusing and partitioning the three-dimensional segmentation mask corresponding to all the organ structures includes: A joint segmentation mask is obtained by Boolean fusion based on the three-dimensional segmentation masks corresponding to all the organ structures described. The tetrahedral model is obtained by partitioning the joint segmentation mask into a three-dimensional mesh.
9. The abdominal organ digital twin model construction system according to claim 8, characterized in that, The process of obtaining the tetrahedral model by three-dimensional mesh partitioning based on the joint segmentation mask includes: Based on the three-dimensional surface contour of the abdominal organ extracted from the joint segmentation mask, a corresponding closed triangular patch surface mesh is generated. Using the closed triangular facet surface mesh as boundary constraints, the three-dimensional space of the abdominal organ covered by the joint segmentation mask is divided into volume meshes to obtain the tetrahedral model corresponding to the abdominal organ.
10. The abdominal organ digital twin model construction system according to claim 1, characterized in that, Prior to acquiring multimodal medical images of the patient's target region, the process includes: Acquire computed tomography (CT) images and magnetic resonance images of the target region; The multimodal medical image is obtained by standardizing the computed tomography image and the magnetic resonance image, wherein the standardization preprocessing includes at least spatial alignment, resolution unification and noise filtering.