Bionic liver model design method and system based on deep learning, medium and computer program product

Abstract

The invention relates to a bionic liver model design method and system based on deep learning, a medium and a computer program product, and the design method comprises the following steps: obtaining medical image data of the liver of a patient, preprocessing the medical image data, and generating a standardized medical image data set; performing image segmentation to generate a unified three-dimensional model containing liver parenchyma, a vascular network, lesion tissues and bile ducts; adding a cuttable marking layer in the three-dimensional space coordinates of the lesion tissue area, defining an operation boundary, and generating an optimized blood vessel model with circulatory system communication; dynamically configuring a blood flow velocity range and a flow resistance threshold value of hemangioma, a bile duct fluid injection parameter and a dynamic adjustment logic parameter of bleeding simulation; and outputting the bionic liver model design file. Compared with the prior art, the method has the advantages that accurate recognition and three-dimensional reconstruction of liver parenchyma, a vascular network, a bile duct and lesion tissue are achieved, the spatial relevance and trueness of a model anatomical structure are remarkably improved, and the model has the real hemodynamic response capacity.

Description

Technical Field

[0001] The present invention relates to a design method for a bionic liver model, and in particular to a design method, system, medium, and computer program product for a bionic liver model based on deep learning. Background Art

[0002] In the planning and simulation of hepatobiliary surgery, 3D-printed bionic liver models have become an important auxiliary tool due to their intuitiveness. Traditional models are mostly based on simple three-dimensional reconstruction of medical imaging data and rely on manual annotation to segment the liver parenchyma, blood vessels and diseased tissues. They have problems such as low segmentation accuracy and distortion of the vascular network topology. Especially when dealing with highly complex lesions (such as hemangioma infiltrating the bile duct), manual segmentation is difficult to accurately restore the spatial interaction between the lesion and the surrounding tissue, resulting in large anatomical errors in the model and an inability to provide a reliable basis for surgical boundary definition and hemodynamic simulation. In addition, the existing models lack the integration of dynamic physiological parameters (such as blood flow velocity and bile duct fluid pressure), which limits their application value in surgical risk assessment and simulation training.

[0003] In recent years, deep learning technology has shown significant advantages in medical image segmentation, and algorithms such as U-Net and 3D CNN have been used to identify anatomical structures in the liver region. However, existing methods mostly focus on the segmentation of a single tissue (such as liver parenchyma or tumor), ignoring the coordinated reconstruction of vascular networks, bile ducts and lesion areas, resulting in the generated three-dimensional models lacking anatomical relevance and mechanical consistency. For example, the angles of vascular branches and the proportion of tube diameters have not been optimized for fluid dynamics, making it difficult to simulate the actual blood flow distribution; at the same time, the model lacks interactive surgical cutting guide markers and dynamic bleeding response mechanisms, and cannot meet the preoperative simulation requirements of complex hepatobiliary surgeries. The singleness of existing 3D printed models in the functional simulation dimension has become a core bottleneck restricting their clinical application.

[0004] For example, reference patent CN118967927A proposes a combined hepatobiliary medical image processing method that can acquire and preprocess medical image data, perform segmentation, analysis, and three-dimensional visualization, and generate auxiliary diagnosis reports, but it does not involve the construction of personalized pathology models based on deep learning and the topological optimization of vascular networks.

[0005] In addition, reference patent CN119581042A constructs a geometric structure model of the liver and blood vessels based on CT and MRI data, and introduces biomechanical characteristics to simulate the pressure applied by surgical instruments. However, its main focus is on the stress and strain analysis of blood vessels, and no in-depth research is conducted on the personalized configuration of the liver model and the definition of surgical boundaries.

[0006] Based on the above analysis, existing technologies still have shortcomings in constructing personalized, high-precision bionic liver models, especially in combining deep learning algorithms for accurate identification of diseased tissues, fine optimization of vascular networks, and generation of personalized pathological models with rich dynamic parameters.

[0007] To address the above problems, there is an urgent need for a liver model design method that integrates multimodal medical image segmentation, fluid dynamics simulation, and topological optimization of vascular networks. Summary of the Invention

[0008] The purpose of the present invention is to overcome the defects of the above-mentioned existing technologies and provide a bionic liver model design method, system, medium, and computer program product based on deep learning. Through deep learning algorithms, multimodal medical images are collaboratively segmented and the topology of the vascular network is optimized, which realizes the accurate identification and three-dimensional reconstruction of liver parenchyma, vascular network, bile duct and diseased tissue, significantly improves the spatial correlation and realism of the model's anatomical structure, and the model has real hemodynamic response capabilities.

[0009] The purpose of the present invention can be achieved by the following technical solutions:

[0010] A first aspect of the present invention provides a method for designing a bionic liver model based on deep learning, comprising the following steps:

[0011] S1. Obtain medical imaging data of the patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset;

[0012] S2. Use deep learning algorithms to perform image segmentation on the preprocessed medical image data, identifying and separating the liver parenchyma, vascular network, lesion tissue, and bile ducts, and generating a unified 3D model containing the liver parenchyma, vascular network, lesion tissue, and bile ducts.

[0013] S3. Based on the unified 3D model, a cuttable marker layer is added to the 3D spatial coordinates of the diseased tissue area to define the surgical boundary. Simultaneously, based on the fluid dynamics simulation results, the 3D geometric morphology of the vascular network is topologically optimized to generate an optimized vascular model with circulatory system connectivity.

[0014] S4. Based on the optimized 3D model in S3, dynamically configure the blood flow velocity range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters for bleeding simulation to generate a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters;

[0015] S5. Outputting a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

[0016] Furthermore, S1 specifically includes the following processes:

[0017] Obtain CT or MRI medical imaging data of the patient's liver;

[0018] The medical image data is preprocessed, specifically including: eliminating image noise interference through denoising, normalizing and adjusting the data grayscale range, and generating a standardized medical image dataset that can be directly used for deep learning segmentation.

[0019] Furthermore, S2 specifically includes the following processes:

[0020] Identify and separate the 3D boundaries of liver parenchyma, vascular network, lesion tissue, and bile duct using pre-trained U-Net or 3D CNN models;

[0021] The segmented liver parenchyma, vascular network, lesion tissue, and bile duct are integrated into a unified three-dimensional model with spatial correlation according to anatomical topology, so that the position, morphology, and interaction relationship of each structure conform to the real hepatobiliary anatomical characteristics.

[0022] Furthermore, S3 specifically includes the following processes:

[0023] Embed a cuttable marker layer in the 3D spatial coordinates of the lesion tissue in the unified 3D model, define the surgical boundary range through an algorithm, and provide visual guidance for subsequent surgical simulation;

[0024] Based on the results of fluid dynamics simulation, the three-dimensional geometric morphology of the vascular network is structurally optimized to eliminate abnormal areas of fluid resistance in the circulatory system and generate an optimized vascular model with functional connectivity.

[0025] Furthermore, the fluid dynamics simulation results include fluid dynamics simulation results of blood flow velocity and pressure distribution;

[0026] The means of structural optimization of the three-dimensional geometric morphology of the vascular network include adjusting the branching angle, tube diameter ratio, and connection path.

[0027] Furthermore, S4 specifically includes the following processes:

[0028] According to the pathological characteristics of hemangiomas, the blood flow rate of hemangiomas is dynamically configured to be in the range of 20–100 mL / min and the flow resistance threshold is in the range of 500–2000 Pa·s;

[0029] Set bile duct fluid injection parameters based on the anatomical structure and lesion conditions of the bile duct;

[0030] According to the pathological characteristics of the diseased tissue and the surgical planning requirements, the dynamic adjustment logic parameters of the bleeding simulation are set to generate a personalized pathology model configuration file containing anatomical structure, marking data, and dynamic parameters.

[0031] Furthermore, the pathological characteristics of the hemangioma include tumor size, location, and hemodynamic characteristics;

[0032] The bile duct fluid injection parameters include fluid injection rate and fluid injection pressure, the fluid injection rate ranges from 0.5 to 5 mL / min, and the fluid injection pressure ranges from 10 to 50 mmHg;

[0033] The dynamic adjustment logic parameters of the bleeding simulation include a bleeding rate adjustment range of 1 to 20 mL / min and a pressure change response time of 0.1 to 2 seconds.

[0034] A second aspect of the present invention provides a bionic liver model design system based on deep learning, comprising:

[0035] a data acquisition module, configured to acquire medical imaging data of a patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset;

[0036] The deep learning segmentation module uses deep learning algorithms to perform image segmentation on pre-processed medical image data, identifying and separating the liver parenchyma, vascular network, and lesion tissue, generating a unified 3D model that includes the liver parenchyma, vascular network, lesion tissue, and bile duct.

[0037] a model optimization module, based on the unified 3D model, adding a cuttable marker layer to the 3D spatial coordinates of the diseased tissue area to define the surgical boundary, and simultaneously topologically optimizing the 3D geometric morphology of the vascular network based on fluid dynamics simulation results to generate an optimized vascular model with circulatory system connectivity;

[0038] The dynamic simulation configuration module dynamically configures the blood flow rate range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters for bleeding simulation based on the 3D model optimized by the model optimization module, generating a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters;

[0039] The output module outputs a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

[0040] A third aspect of the present invention provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, is used to execute the above-mentioned deep learning-based bionic liver model design method.

[0041] A fourth aspect of the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the above-mentioned deep learning-based bionic liver model design method.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1) The present invention uses a deep learning algorithm to collaboratively segment multimodal medical images, achieving accurate identification and three-dimensional reconstruction of liver parenchyma, vascular networks, bile ducts, and diseased tissues, significantly improving the spatial correlation and realism of the model's anatomical structure. The vascular network is topologically optimized based on fluid dynamics simulation to ensure that the vascular branch angles and tube diameter ratios conform to physiological blood flow characteristics, effectively eliminating areas of abnormal circulatory resistance, and enabling the model to have real hemodynamic response capabilities. At the same time, a cuttable marker layer is embedded in the lesion area and the blood flow rate, flow resistance threshold, and bleeding simulation parameters are dynamically configured, so that the model has both surgical boundary visualization guidance and dynamic physiological simulation functions, providing high-precision, multi-dimensional data support for preoperative planning and risk assessment of complex hepatobiliary surgeries.

[0044] 2) The personalized pathology model configuration file of the present invention can directly drive 3D printing to generate a bionic liver entity model, which integrates anatomical structure, fluid mechanics parameters and dynamic bleeding logic, breaking through the functional limitations of traditional static models. Doctors can simulate real surgical operations on the entity model, observe blood flow changes and bleeding responses in real time, optimize the surgical path and reduce the risk of intraoperative complications. In addition, modular design (such as dynamic simulation configuration module and optimized vascular model generation module) supports rapid adaptation to different case needs, greatly shortens the model customization cycle, and promotes the clinical transformation efficiency of precision medical technology in hepatobiliary surgery, with significant social benefits and clinical application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of the process of designing a bionic liver model based on deep learning in the present invention. DETAILED DESCRIPTION

[0046] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Any features such as component models, material names, connection structures, circuit structures, control methods, algorithms, etc. not explicitly described in this technical solution are considered to be common technical features disclosed in the prior art.

[0047] Example 1

[0048] The bionic liver model design method based on deep learning in this invention can be found in Figure 1 , including the following steps:

[0049] S1. Obtain medical imaging data of a patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset.

[0050] S1 specifically includes the following processes:

[0051] Obtain CT or MRI medical imaging data of the patient's liver;

[0052] The medical image data is preprocessed, specifically including: eliminating image noise interference through denoising, normalizing and adjusting the data grayscale range, and generating a standardized medical image dataset that can be directly used for deep learning segmentation.

[0053] The core of step S1 is to construct a standardized medical imaging dataset, and its specific implementation includes two key links: first, obtain two-dimensional / three-dimensional tomographic imaging data of the patient's liver through CT or MRI equipment. This type of data has high spatial resolution and can clearly present the anatomical differences of liver parenchyma, blood vessels and diseased tissues. Secondly, preprocessing operations are performed on the original medical images: Gaussian filtering, non-local mean algorithm or noise suppression model based on deep learning (such as DnCNN) are used to improve the image signal-to-noise ratio by eliminating interference signals such as equipment noise and motion artifacts; through linear or nonlinear transformation (such as Z-score normalization, histogram matching), the image grayscale values under different devices and scanning protocols are mapped to a unified interval (such as 0-255) to eliminate the interference of data distribution differences on subsequent deep learning model training. The preprocessed dataset has the characteristics of spatial alignment, controllable noise and consistent grayscale, providing high-fidelity input for deep learning segmentation.

[0054] S2. Use deep learning algorithms to perform image segmentation on the preprocessed medical imaging data to identify and separate the liver parenchyma, vascular network, diseased tissue, and bile ducts, and generate a unified three-dimensional model that includes the liver parenchyma, vascular network, diseased tissue, and bile ducts.

[0055] S2 specifically includes the following processes:

[0056] Identify and separate the 3D boundaries of liver parenchyma, vascular network, lesion tissue, and bile duct using pre-trained U-Net or 3D CNN models;

[0057] The segmented liver parenchyma, vascular network, lesion tissue, and bile duct are integrated into a unified three-dimensional model with spatial correlation according to anatomical topology, so that the position, morphology, and interaction relationship of each structure conform to the real hepatobiliary anatomical characteristics.

[0058] In specific implementation, a pre-trained U-Net or 3D CNN model is used to analyze the pre-processed medical imaging data, and the feature information of the image is extracted layer by layer through the model's convolutional layer, pooling layer and other structures. During the pre-training process, the model has learned a large amount of characteristic patterns of medical imaging data and can accurately identify and segment the three-dimensional boundaries of the liver parenchyma, vascular network, diseased tissue, and bile duct. After obtaining the boundaries of these tissues, the segmented parts are integrated according to the anatomical topological relationship, and the corresponding three-dimensional reconstruction algorithm is used to combine the segmentation results of these two-dimensional slices into a unified three-dimensional model with spatial correlation. In this process, the position, morphology and mutual interaction of each tissue need to be corrected and adjusted to ensure that the final generated three-dimensional model conforms to the anatomical characteristics of the real liver and gallbladder.

[0059] In specific implementations, U-Net, through its unique encoder-decoder structure, is able to effectively capture detailed image features while maintaining spatial information. 3D CNN can directly process three-dimensional data, better capturing the continuity and correlation of tissue structures in three-dimensional space. These models are trained using large amounts of labeled data, learning features such as grayscale and texture of different tissues in images, thereby achieving automated image segmentation. The model training process will not be detailed here. When generating a unified three-dimensional model, integration based on anatomical topological relationships is key. This involves understanding and applying knowledge of liver anatomy to ensure that the position and morphology of each tissue in three-dimensional space conform to the actual anatomical structure, making the model highly realistic and usable.

[0060] S3. Based on the unified three-dimensional model, a cuttable marker layer is added to the three-dimensional spatial coordinates of the diseased tissue area to define the surgical boundary. At the same time, based on the fluid dynamics simulation results, the three-dimensional geometric morphology of the vascular network is topologically optimized to generate an optimized vascular model with circulatory system connectivity.

[0061] Specifically, generating an optimized vascular model with circulatory system connectivity means analyzing and adjusting the original three-dimensional vascular network model during the design of a bionic liver model based on deep learning, so that it becomes a system that is internally connected and can simulate real blood flow. Specifically, the vascular network is first evaluated using fluid dynamics simulation technology to identify areas that may cause poor blood flow or abnormal resistance, such as vascular stenosis, unreasonable branching angles, or inharmonious tube diameter ratios. Then, by adjusting the geometric morphology of the blood vessels, including optimizing branching angles, adjusting tube diameters, and correcting connection paths, these abnormal areas are eliminated to ensure that blood can flow smoothly throughout the entire vascular network, thereby generating an optimized vascular model that conforms to the anatomical structure and has functional connectivity. This process is crucial for accurately simulating the blood circulation and related physiological functions of the liver, and can provide more realistic and reliable model support for medical research, surgical planning, and teaching.

[0062] In S3, the specific process includes the following:

[0063] Embed a cuttable marker layer in the 3D spatial coordinates of the lesion tissue in the unified 3D model, define the surgical boundary range through an algorithm, and provide visual guidance for subsequent surgical simulation;

[0064] Based on the results of fluid dynamics simulation, the three-dimensional geometric morphology of the vascular network is structurally optimized to eliminate abnormal areas of fluid resistance in the circulatory system and generate an optimized vascular model with functional connectivity.

[0065] The fluid dynamics simulation results include fluid dynamics simulation results of blood flow velocity and pressure distribution;

[0066] The means of structural optimization of the three-dimensional geometric morphology of the vascular network include adjusting the branching angle, tube diameter ratio, and connection path.

[0067] In specific implementation, the 3D spatial coordinates of the diseased tissue are located within a unified 3D model. A cuttable marker layer is then embedded, and a specialized, known algorithm is used to precisely define the surgical boundary, providing visual guidance for surgical simulation. Next, based on the fluid dynamics simulation results, the blood flow velocity and pressure distribution within the vascular network are analyzed to identify areas of abnormal fluid resistance. Finally, the 3D geometry of the vascular network is optimized by adjusting the branching angles, diameter ratios, and connectivity paths, generating an optimized vascular model with good functional connectivity.

[0068] In specific implementations, fluid dynamics simulation can accurately simulate the flow of blood within blood vessels, including parameters such as flow rate and pressure distribution. By analyzing these parameters, areas of abnormal fluid resistance in the vascular network can be discovered, which may affect the normal circulation of blood. Adjusting the branching angles, diameter ratios, and connection paths of blood vessels can change the flow state of blood, reduce fluid resistance, and thus optimize the functional connectivity of the vascular network. Embedding a cuttable marker layer and defining the surgical boundary range utilizes computer graphics and image processing technology to intuitively display surgical-related information on the three-dimensional model, providing precise guidance for surgical simulation.

[0069] S4. Based on the optimized three-dimensional model in S3, dynamically configure the blood flow rate range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters of the bleeding simulation to generate a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters.

[0070] S4 specifically includes the following processes:

[0071] According to the pathological characteristics of hemangiomas, the blood flow rate of hemangiomas is dynamically configured to be in the range of 20–100 mL / min and the flow resistance threshold is in the range of 500–2000 Pa·s;

[0072] Set bile duct fluid injection parameters based on the anatomical structure and lesion conditions of the bile duct;

[0073] According to the pathological characteristics of the diseased tissue and the surgical planning requirements, the dynamic adjustment logic parameters of the bleeding simulation are set to generate a personalized pathology model configuration file containing anatomical structure, marking data, and dynamic parameters.

[0074] In a specific implementation, the pathological characteristics of the hemangioma include tumor size, location, and hemodynamic characteristics;

[0075] The bile duct fluid injection parameters include fluid injection rate and fluid injection pressure, the fluid injection rate ranges from 0.5 to 5 mL / min, and the fluid injection pressure ranges from 10 to 50 mmHg;

[0076] The dynamic adjustment logic parameters of the bleeding simulation include a bleeding rate adjustment range of 1 to 20 mL / min and a pressure change response time of 0.1 to 2 seconds.

[0077] During the specific implementation, it is first necessary to collect and analyze the pathological characteristics of the hemangioma, such as the size, specific location and hemodynamic characteristics of the tumor. Based on this information, the blood flow rate range of the hemangioma is dynamically set to between 20 and 100 mL / min, and the flow resistance threshold is set to between 500 and 2000 Pa·s. Then, based on the anatomical structure characteristics and pathological conditions of the bile duct, the speed and pressure parameters of the bile duct fluid injection are determined. The fluid injection speed is generally between 0.5 and 5 mL / min, and the fluid injection pressure is between 10 and 50 mmHg. Finally, based on the pathological characteristics of the lesion tissue and the specific requirements of the surgical plan, the relevant parameters of the bleeding simulation are set, including a bleeding rate adjustment range of 1 to 20 mL / min and a pressure change response time of 0.1 to 2 seconds, thereby generating a personalized pathology model configuration file containing anatomical structure information, marker data and dynamic parameters.

[0078] In specific implementation, the setting of the blood flow rate and flow resistance of the hemangioma is based on the principles of fluid mechanics. The appropriate flow rate and flow resistance range are determined according to the size, location and hemodynamic characteristics of the hemangioma to simulate the real blood flow state. When setting the bile duct fluid injection parameters, the anatomical structure and pathological conditions of the bile duct are taken into consideration. By adjusting the injection speed and pressure, the flow of bile in the bile duct is simulated. The setting of the bleeding simulation parameters is based on the characteristics of the diseased tissue and the needs of the surgical planning. By adjusting the bleeding rate and pressure change response time, the bleeding that may occur during the operation is simulated, providing a more realistic simulation environment for the operation.

[0079] S5. Outputting a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

[0080] During implementation, the personalized pathology model configuration file must first be parsed to extract key information, including anatomical structure, marker data, and dynamic parameters. This information is then converted into formats recognizable by 3D printing equipment, such as STL and OBJ, using professional 3D modeling software or specialized conversion tools. During this conversion process, the model's geometric accuracy and detailed integrity must be ensured to accurately reproduce the liver's anatomical structure and pathological features. Furthermore, the technical requirements of 3D printing, such as material properties, print resolution, and layer thickness, must be considered, and the model must be appropriately optimized and adjusted to ensure that the printed bionic liver model meets both design requirements and actual medical application needs.

[0081] Example 2

[0082] The bionic liver model design system based on deep learning in this embodiment includes:

[0083] a data acquisition module, configured to acquire medical imaging data of a patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset;

[0084] The deep learning segmentation module uses deep learning algorithms to perform image segmentation on pre-processed medical image data, identifying and separating the liver parenchyma, vascular network, and lesion tissue, generating a unified 3D model that includes the liver parenchyma, vascular network, lesion tissue, and bile duct.

[0085] a model optimization module, based on the unified 3D model, adding a cuttable marker layer to the 3D spatial coordinates of the diseased tissue area to define the surgical boundary, and simultaneously topologically optimizing the 3D geometric morphology of the vascular network based on fluid dynamics simulation results to generate an optimized vascular model with circulatory system connectivity;

[0086] The dynamic simulation configuration module dynamically configures the blood flow rate range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters for bleeding simulation based on the 3D model optimized by the model optimization module, generating a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters;

[0087] The output module outputs a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

[0088] When the system starts, the data acquisition module first obtains the CT or MRI medical imaging data of the patient's liver and performs preprocessing operations such as denoising and normalization on this data to generate a standardized medical imaging data set. Next, the deep learning segmentation module uses the pre-trained U-Net model to perform image segmentation on the pre-processed data, accurately identifying and separating the three-dimensional boundaries of the liver parenchyma, vascular network, diseased tissue, and bile duct, and integrating these structures into a unified three-dimensional model. Subsequently, the model optimization module embeds a cuttable marker layer in the three-dimensional spatial coordinates of the diseased tissue area to define the surgical boundary, and based on the fluid dynamics simulation results, it performs topological optimization on the geometric morphology of the vascular network to generate an optimized vascular model with circulatory system connectivity. Based on the optimized 3D model, the dynamic simulation configuration module dynamically configures the hemangioma's blood flow rate to a range of 20 to 100 mL / min, with a flow resistance threshold of 500 to 2000 Pa·s. It also sets the bile duct fluid injection rate to 0.5 to 5 mL / min and the injection pressure to 10 to 50 mmHg. Based on the pathological characteristics of the lesion and surgical planning requirements, it also sets the dynamic adjustment logic parameters for the bleeding simulation, generating a personalized pathology model configuration file containing anatomical structures, marker data, and dynamic parameters. Finally, the output module converts this configuration file into an STL-formatted 3D printing file, which guides the 3D printer to produce a high-precision, personalized bionic liver model.

[0089] Example 3

[0090] In this embodiment, the storage medium of the computer executable instructions can be a CD, a USB flash drive, a mobile hard disk, a solid-state drive, a mechanical hard disk, etc. When the computer executable instructions stored in these storage media are loaded and executed by the computer processor, they can operate according to the preset process of the bionic liver model design method based on deep learning. Specifically, the computer executable instructions will first call the deep learning model to process and analyze the patient's medical imaging data, identify and separate the different tissue structures of the liver, and then dynamically adjust the relevant parameters according to the pathological characteristics and surgical requirements, and finally generate a personalized bionic liver model design file. This design file can be directly used in 3D printing equipment to produce a high-precision, personalized bionic liver model for use in medical teaching, surgical simulation and other scenarios.

[0091] Example 4

[0092] The computer program product in this embodiment includes a computer program stored in a non-volatile storage medium (such as a CD, a USB flash drive, a mobile hard drive, a solid-state drive, or a mechanical hard drive, etc.). When the program is loaded and executed by a computer processor, it can complete all steps from medical imaging data acquisition, preprocessing, image segmentation, three-dimensional reconstruction, to personalized configuration file generation according to the above-mentioned bionic liver model design method based on deep learning. Ultimately, the program can also convert the generated personalized pathology model configuration file into a design file format that can be recognized by a 3D printing device, thereby guiding the 3D printer to produce a high-precision, personalized bionic liver model for use in scenarios such as medical teaching, scientific research experiments, and clinical surgery simulations.

[0093] The above description of the embodiments is intended to facilitate understanding and use of the invention by those skilled in the art. It will be apparent that those skilled in the art can readily make various modifications to these embodiments and apply the general principles described herein to other embodiments without requiring inventive effort. Therefore, the present invention is not limited to the above-described embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention, without departing from the scope of the present invention, should be within the scope of protection of the present invention.

Claims

1. A bionic liver model design method based on deep learning, characterized in that: The following steps are involved: S1. Obtain medical imaging data of the patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset; S2. Use deep learning algorithms to perform image segmentation on the preprocessed medical image data, identifying and separating the liver parenchyma, vascular network, lesion tissue, and bile ducts, and generating a unified 3D model containing the liver parenchyma, vascular network, lesion tissue, and bile ducts. S3. Based on the unified 3D model, a cuttable marker layer is added to the 3D spatial coordinates of the diseased tissue area to define the surgical boundary. Simultaneously, based on the fluid dynamics simulation results, the 3D geometric morphology of the vascular network is topologically optimized to generate an optimized vascular model with circulatory system connectivity. S4. Based on the optimized 3D model in S3, dynamically configure the blood flow velocity range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters for bleeding simulation to generate a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters; S5. Outputting a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

2. The method for designing a bionic liver model based on deep learning according to claim 1, characterized in that: In S1, specific The following processes are included: Obtain CT or MRI medical imaging data of the patient's liver; The medical image data is preprocessed, specifically including: eliminating image noise interference through denoising, normalizing and adjusting the data grayscale range, and generating a standardized medical image dataset that can be directly used for deep learning segmentation.

3. The method for designing a bionic liver model based on deep learning according to claim 1, characterized in that: S2 specifically includes the following processes: Identify and separate the 3D boundaries of liver parenchyma, vascular network, lesion tissue, and bile duct using pre-trained U-Net or 3D CNN models; The segmented liver parenchyma, vascular network, lesion tissue, and bile duct are integrated into a unified three-dimensional model with spatial correlation according to anatomical topology, so that the position, morphology, and interaction relationship of each structure conform to the real hepatobiliary anatomical characteristics.

4. The method for designing a bionic liver model based on deep learning according to claim 1, characterized in that: S3, specifically The following processes are included: Embed a cuttable marker layer in the 3D spatial coordinates of the lesion tissue in the unified 3D model, define the surgical boundary range through an algorithm, and provide visual guidance for subsequent surgical simulation; Based on the results of fluid dynamics simulation, the three-dimensional geometric morphology of the vascular network is structurally optimized to eliminate abnormal areas of fluid resistance in the circulatory system and generate an optimized vascular model with functional connectivity.

5. The method for designing a bionic liver model based on deep learning according to claim 4, characterized in that: The fluid dynamics simulation results include fluid dynamics simulation results of blood flow velocity and pressure distribution; The means of structural optimization of the three-dimensional geometric morphology of the vascular network include adjusting the branching angle, tube diameter ratio, and connection path.

6. The method for designing a bionic liver model based on deep learning according to claim 1, characterized in that: S4 specifically includes the following processes: According to the pathological characteristics of hemangiomas, the blood flow rate of hemangiomas is dynamically configured to be in the range of 20–100 mL / min and the flow resistance threshold is in the range of 500–2000 Pa·s; Set bile duct fluid injection parameters based on the anatomical structure and lesion conditions of the bile duct; According to the pathological characteristics of the diseased tissue and the surgical planning requirements, the dynamic adjustment logic parameters of the bleeding simulation are set to generate a personalized pathology model configuration file containing anatomical structure, marking data, and dynamic parameters.

7. The method for designing a bionic liver model based on deep learning according to claim 6, characterized in that: The pathological characteristics of the hemangioma include tumor size, location, and hemodynamic characteristics; The bile duct fluid injection parameters include fluid injection rate and fluid injection pressure, the fluid injection rate ranges from 0.5 to 5 mL / min, and the fluid injection pressure ranges from 10 to 50 mmHg; The dynamic adjustment logic parameters of the bleeding simulation include a bleeding rate adjustment range of 1 to 20 mL / min and a pressure change response time of 0.1 to 2 seconds.

8. A bionic liver model design system based on deep learning, characterized in that: include: a data acquisition module, configured to acquire medical imaging data of a patient's liver, preprocess the medical imaging data, and generate a standardized medical imaging dataset; The deep learning segmentation module uses deep learning algorithms to perform image segmentation on pre-processed medical image data, identifying and separating the liver parenchyma, vascular network, and lesion tissue, generating a unified 3D model that includes the liver parenchyma, vascular network, lesion tissue, and bile duct. a model optimization module, based on the unified 3D model, adding a cuttable marker layer to the 3D spatial coordinates of the diseased tissue area to define the surgical boundary, and simultaneously topologically optimizing the 3D geometric morphology of the vascular network based on fluid dynamics simulation results to generate an optimized vascular model with circulatory system connectivity; The dynamic simulation configuration module dynamically configures the blood flow rate range and flow resistance threshold of the hemangioma, bile duct fluid injection parameters, and dynamic adjustment logic parameters for bleeding simulation based on the 3D model optimized by the model optimization module, generating a personalized pathology model configuration file containing anatomical structure, labeling data, and dynamic parameters; The output module outputs a bionic liver model design file that can be directly used for 3D printing based on the personalized pathology model configuration file.

9. A storage medium containing computer-executable instructions, characterized in that: The storage medium of the computer-executable instructions, when executed by a computer processor, is used to execute the bionic liver model design method based on deep learning according to any one of claims 1 to 7.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for designing a bionic liver model based on deep learning is implemented.

Images