Vascular exposure prediction model construction method based on liver resection anatomy

Abstract

The invention relates to a method for constructing a blood vessel exposure prediction model during anatomy based on hepatic resection, which comprises the following steps of: constructing geometric structures of liver and blood vessels based on CT (Computed Tomography) and MRI (Magnetic Resonance Imaging) data, and introducing biomechanical characteristics for modeling; the model is combined with a biomechanical principle, and stress and strain behaviors of blood vessels under different physiological conditions and reactions when a surgical instrument applies pressure are simulated through finite element analysis FEA; a static anatomical structure is provided, and the prediction model provides basic data about vascular elasticity and toughness dynamic factors; a hierarchical branch complexity model is introduced, a blood vessel branch network is classified according to the branch number, angle and diameter change features, and the difficulty of blood vessel exposure is analyzed on different levels; establishing blood vessel exposure situations under different operation paths in an operation through operation dynamic simulation based on a physical engine; the deformation or infinitesimal displacement is predicted when the instrument is in contact with tissues around the blood vessel in the operation process.

Description

Technical Field

[0001] The present invention relates to a method for constructing a vascular exposure prediction model, and in particular to a method for constructing a vascular exposure prediction model based on liver resection anatomy. Background Art

[0002] At present, the related technologies for constructing the prediction model of vascular exposure during liver resection dissection, especially the Chinese invention patent 2023115266473, have some drawbacks, which are mainly reflected in the dynamic prediction ability, spatiotemporal data fusion, real-time feedback and adaptability, which are specifically described as follows:

[0003] First of all, the patent uses static 3D modeling technology. Although it can model the vascular information of the liver area and segment the anatomical structure of the liver and blood vessels, such as the left and right branches of the hepatic artery, through the 3D modeling module, this type of static model can only provide preoperative planning and lacks the ability to predict the real-time changes in vascular exposure during the operation. During the operation, the patient's liver and blood vessels may undergo dynamic deformation with the operation of the surgical instruments, and this static modeling technology cannot capture these changes, resulting in the doctor's inability to timely predict the risk of vascular exposure during the operation. In the prior art, although semantic segmentation and instance segmentation models are used to segment anatomical areas such as the liver, gallbladder, and round ligament of the liver, these segmentation results cannot reflect the dynamic stress-strain response of liver tissue and blood vessels during the operation. Since the operation of surgical instruments will cause different degrees of compression and stretching on the blood vessels, the exposure of the blood vessels is not fixed, but may change at any time. The existing scheme lacks the ability to predict and respond to such dynamic changes, and only relies on static planning before surgery, which is prone to errors in actual operation. Secondly, although the prior art can obtain the area range of the target blood vessel in the surgical screen and distinguish the individual branches of the blood vessel through the instance segmentation model, these operations are also based on the segmentation of static images or surgical screens, and lack the dynamic fusion analysis of spatiotemporal data. The key data in the operation includes not only spatial information (such as blood vessel position, structure, etc.), but also temporal information (such as instrument movement, surgical progress, etc.). In liver resection surgery, the movement trajectory and pressure distribution of surgical instruments will affect the deformation and exposure risk of blood vessels, so the real-time fusion analysis of these spatiotemporal data is crucial. However, the analysis module, semantic segmentation module and instance segmentation module in the patent do not fully consider the dynamic interaction process between surgical instruments and blood vessels. Although a vascular branch recognition model based on Transformer encoder is introduced to identify the category and branch structure of blood vessels, these technologies are more from the perspective of image segmentation and static classification, and lack the description and processing capabilities of the dynamic stress-strain relationship generated when the instrument interacts with the blood vessel. Third, the prior art lacks real-time feedback and adaptability during surgery. Although the patent provides an acquisition module for real-time acquisition of surgical images, no adaptive adjustment is made based on these real-time data. During surgery, the risk of vascular exposure may change with the movement of surgical instruments, pressure changes, and fluctuations in the patient's physiological state. Therefore, real-time feedback mechanisms and adaptive models are extremely important. The prediction model can be updated based on real-time collected data to ensure a more accurate prediction of vascular exposure risks during surgery. However, the analysis module in this patent can only divide the area of ​​vascular exposure based on the predetermined surgical plan, and lacks adaptive adjustment of vascular exposure during surgery.For example, during surgery, when the pressure applied by the instrument to the blood vessels exceeds a certain threshold, the blood vessels may be significantly deformed or exposed, but existing technologies are unable to update the model or adjust the surgical path in a timely manner based on these dynamic changes. There is a certain lag, which increases the risk of blood vessel rupture or damage during surgery.

[0004] Finally, the existing technology has limitations in the application of multimodal data fusion and deep learning technology. Although the patent introduces a multi-layer feature extraction network and a classification head model, which can perform image segmentation and vascular recognition through certain machine learning techniques, this segmentation and recognition cannot completely solve the complex dynamic problems in liver resection surgery. The data in modern surgical procedures are complex and diverse, involving multiple information sources such as imaging data (such as CT, MRI), real-time position of surgical instruments, and pressure sensor feedback data, and the existing technology does not use these multimodal data for comprehensive analysis. Especially in dynamic surgical scenarios, the movement trajectory and time series changes of instruments are closely related to the risk of vascular exposure. If the deep learning model can combine these spatiotemporal data and capture the complex interactive relationship between instruments and blood vessels during surgery, it can better predict the risk of vascular exposure in the future. However, this patent does not make full use of these deep learning technologies to process and analyze multimodal and time series data, resulting in its limitations in handling complex surgical dynamic scenes. Summary of the invention

[0005] The purpose of the present invention is to provide a method for constructing a vascular exposure prediction model based on anatomy during liver resection, thereby solving some of the drawbacks pointed out in the background technology.

[0006] The present invention solves the above-mentioned technical problems by adopting the following technical solution, including the steps of:

[0007] S1. Deep anatomical modeling of liver vascular distribution:

[0008] S1.1. Construct the geometric structure of the liver and blood vessels based on CT and MRI data, and introduce biomechanical characteristics modeling; the model combines biomechanical principles, and simulates the stress and strain behavior of blood vessels under different physiological conditions through finite element analysis (FEA), and the response when surgical instruments apply pressure; provide static anatomical structures, and the prediction model provides basic data on dynamic factors of vascular elasticity and toughness;

[0009] S1.2. Introduce a hierarchical branch complexity model to classify the vascular branch network according to the number, angle, and diameter variation characteristics of the branches, and analyze the difficulty of vascular exposure at different levels;

[0010] S2. Dynamic surgical simulation and vascular exposure scenario construction:

[0011] S2.1. Through dynamic surgical simulation based on a physical engine, establish vascular exposure scenarios under different operation paths during surgery, including predicting deformation or micro-displacement caused when the instrument contacts the tissue around the blood vessel during surgery;

[0012] S2.2. Construct a spatiotemporal vascular exposure prediction model to predict the predictability of vascular exposure at different time points. Based on the spatiotemporal data flow, the temporal relationship between intraoperative instrument movement and vascular exposure is captured through the long short-term memory network (LSTM) or the temporal convolutional network (TCN), and the probability of vascular exposure at a specific time point is predicted.

[0013] S3. Vascular exposure prediction model based on fusion of global and local factors:

[0014] S3.1. Design a fusion model to jointly analyze local geometric features including vessel diameter and branching angle with global features including vessel distribution pattern and density of adjacent tissues;

[0015] S3.2. Construct a multimodal data fusion model by integrating multimodal physiological data; perform weighted fusion of data from different information sources through ensemble learning methods;

[0016] S4. Real-time feedback during surgery and model adaptive optimization:

[0017] S4.1. Introduce ultrasound and endoscopic image data collected in real time during surgery, and combine them with a deep learning model for real-time correction. The model adaptively adjusts the prediction results based on the new input information; through the convolution operation of the real-time image, the regional distribution of the expected exposed blood vessels is recalculated.

[0018] Furthermore, the deep anatomical modeling of the liver vascular distribution includes:

[0019] Based on the patient's CT and MRI images, the liver and blood vessel structures are first reconstructed in three dimensions; the morphology of the blood vessels is captured, and the deformation behavior of the blood vessels under different surgical pressure conditions is perceived; the geometric model is constructed through functions Describe, where is a function that defines the three-dimensional geometry of the liver and blood vessels, where x, y, and z represent the three-dimensional coordinates of the space; capturing the complex geometry and distribution characteristics of the blood vessels, the equation combines the local biomechanical properties, including the local deformation velocity field v(x, y, z), and is described by the following formula:

[0020]

[0021] Among them, γ i is the weight coefficient in the geometric model, which is used to control the influence of different anatomical regions; is the Laplace operator It is used to describe the degree of change between points in the geometric structure; v(x, y, z) is the local deformation velocity field, which represents the velocity distribution of blood vessels and surrounding tissues when subjected to force, and reflects the dynamic response of blood vessels; κ is the coupling coefficient, which represents the degree of coupling between the velocity field and the geometric shape, and is used to control the degree of deformation of blood vessels under different conditions.

[0022] Furthermore, the deep anatomical modeling of the liver vascular distribution includes:

[0023] Finite element analysis (FEA) simulates the dynamic behavior of blood vessels through stress-strain relationships and predicts the risk of rupture or exposure during surgery. The core formula for dynamic stress-strain analysis is:

[0024]

[0025] in, It is the local pressure that changes with time, indicating the force acting on the blood vessel at a certain time t and position x; it is used to simulate the pressure exerted on the blood vessel by the surgical instrument during the operation; λ1 and λ2 are the stress-strain influencing factors of the system, λ1 mainly controls the influence of stress on the system, while λ2 controls the influence of the change of contact force on the blood vessel; the two parameters affect the response of the blood vessel under different physiological conditions by adjusting the relationship between pressure and strain; It is the rate of change of strain with time, which describes the strain rate of the blood vessel at time t, that is, the deformation speed of the blood vessel at a certain moment; it represents the instantaneous deformation behavior of the blood vessel and captures the rapid deformation during surgical operations; Δ(x, t) is the spatial distribution function of the contact force between the surgical instrument and the blood vessel, which represents the force exerted by the surgical instrument on the blood vessel at spatial position x and time t.

[0026] Furthermore, the deep anatomical modeling of the liver vascular distribution includes:

[0027] By integrating static geometric information with dynamic stress-strain relationships, a spatiotemporal coupling model is created to predict the dynamic deformation and exposure risk of blood vessels during surgery. The spatiotemporal coupling model is described by the following formula:

[0028]

[0029] Among them, R(x, t) is the output of the fusion model, which represents the comprehensive result of static geometric information and dynamic stress response at position x and time t, and is ultimately used to assess the risk of vascular exposure during surgery;

[0030] η(x) is the weight coefficient, which controls the influence of static geometric features on the overall response; represents the gradient of geometric morphology, reflecting the shape change in static geometric features; μ(t) is a time-dependent dynamic deformation function, describing the dynamic response characteristics of blood vessels at different times; ξ(x, t) is a spatiotemporal coupling factor, representing the combined effect of spatial and temporal changes of stress on blood vessel deformation; This second-order derivative captures the rate of change of pressure in space and reflects the effect of local pressure on the overall vessel.

[0031] Furthermore, the dynamic surgery simulation and vascular exposure scenario construction includes:

[0032] Various data are collected in real time during the operation, including real-time images from ultrasound and endoscopes, instrument position information, and pressure sensor feedback data; multi-dimensional data forms a dynamic spatiotemporal data stream to describe the interactive relationship between instrument movement, pressure application, and blood vessels; the spatiotemporal data stream contains spatial dimensions x, y, z, and is combined with the time dimension t to create a dynamic scene containing spatiotemporal information; through this scene, the trajectory of the surgical instrument is recorded in real time, and the pressure changes of the instrument on the blood vessels at different time points are captured, and then the spatiotemporal model is established:

[0033]

[0034] in represents the risk level of vascular exposure at spatial coordinates x, y, z and time t; the function describes the predictability of vascular exposure at a specific time and spatial point during surgery; represents the pressure gradient applied by the surgical instrument, describing the pressure change applied by the surgical instrument to the blood vessel at position x, y, z and time t′; Represents the instrument movement speed at position x, y, z and time t′; this speed term is used to capture the motion trajectory of the surgical instrument, t′: is the integral time variable, integrated from the initial time to the current time t, describing the comprehensive impact of the instrument on the blood vessel over the past period of time.

[0035] Furthermore, the dynamic surgery simulation and vascular exposure scenario construction includes:

[0036] By introducing the long short-term memory network LSTM and the temporal convolutional network TCN, the long-term and short-term dependencies in the time series are learned and predicted; according to the action sequence of surgical instruments in different time periods, the vascular exposure at future moments is predicted; TCN efficiently processes the short-term patterns in the time series through convolution operations and captures the rapid changes in a short period of time during surgery:

[0037] ε t =γ1·σ(W x ·x t +W h ·h t-1 +b)

[0038] where ε t represents the probability of vascular exposure at the current time t; is the output of the system prediction, reflecting the risk of vascular exposure caused by the surgical instrument operation prediction at the current moment; x t represents the input data at time t, including the position, speed, and contact information of the device with the blood vessel; W x Represents the weight matrix of the input data, which is used to adjust the influence weight of the input features; h t-1 Represents the implicit state of the previous moment, retaining the historical information of the operation at the previous moment. Through this state, LSTM captures the long-term dependency of the operation; W h The weight matrix representing the implicit state controls the influence of the state at the previous moment on the current prediction; σ is the activation function, which is a nonlinear function used to map the network output to a probability value range; b is the bias term used to balance the output results.

[0039] Furthermore, the dynamic surgery simulation and vascular exposure scenario construction includes:

[0040] By dynamically combining the spatial interaction and time series of surgical instruments and blood vessels, a comprehensive vascular exposure probability prediction model is generated, and the probability value of vascular exposure is output to guide surgical operation optimization:

[0041]

[0042] in, It represents the probability value of vascular exposure at the future time t+Δt; the final result predicted by the model is used to reflect the vascular exposure risk at a specific future moment; α is the adjustment factor used to control the confidence of the model output; Indicates the exposure risk level of the blood vessel at the current time t, based on the spatiotemporal movement and pressure of the surgical instrument;

[0043] It represents the derivative of the instrument movement speed with respect to time, capturing the acceleration and deceleration of the surgical instrument movement; it is used to analyze the impact of the rapid movement of the instrument on the risk of vascular exposure; Ω is the integration area, representing the three-dimensional space of the entire surgical operation scene. The integration process is to calculate all spatial points in the surgical scene to generate a comprehensive vascular exposure probability.

[0044] The present invention has the following beneficial effects:

[0045] By introducing dynamic surgical simulation and spatiotemporal data flow analysis, the system can monitor the interaction between surgical instruments and blood vessels in real time, and combine data such as pressure and speed to accurately predict the risk of blood vessels being exposed during surgery. This real-time prediction function helps doctors predict high-risk areas in advance, thereby optimizing the surgical path and reducing the possibility of vascular exposure and damage. By combining the spatial operation of surgical instruments with time series, the present invention can provide a comprehensive analysis of vascular deformation and exposure during surgery. During critical operations, the system can automatically adjust the prediction results and provide real-time feedback to doctors, reminding them to take necessary measures to avoid vascular rupture or severe exposure, significantly improving the safety of surgery.

[0046] The present invention combines CT, MRI imaging data, ultrasound and endoscope real-time data and dynamic information of surgical instruments to build a multimodal data fusion model. By integrating these multidimensional data, the system can analyze the dynamic behavior of blood vessels from multiple angles, making the prediction results more accurate and comprehensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The present invention is a flow chart of the method for constructing a vascular exposure prediction model during liver resection anatomy.

[0048] Figure 2 The figure is a flowchart of the deep anatomical modeling of the liver vascular distribution according to the present invention.

[0049] Figure 3 This is a surgical picture of a 45-year-old patient undergoing right liver lobectomy in Example 1 and Example 2 of the present invention. DETAILED DESCRIPTION

[0050] The specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0051] Combination Figure 1One of the core steps of the method for constructing a prediction model for vascular exposure during liver resection is the deep anatomical modeling of the liver vascular distribution. This step uses medical imaging data such as CT and MRI to obtain the three-dimensional geometric structure of the patient's liver and blood vessels and reconstructs them in detail. In order to simulate the pressure exerted by the instrument on the blood vessels during surgery and the resulting vascular deformation, biomechanical property modeling must be introduced. This step combines the mechanical properties of liver tissue and blood vessels, such as elastic modulus, toughness, stress and strain, and other parameters. Through finite element analysis (FEA), the model can simulate the stress (pressure) and strain (deformation) behavior of blood vessels under different physiological conditions. Finite element analysis is a numerical method that divides complex geometric structures into smaller units to calculate the stress and strain distribution of blood vessels under different pressure conditions, especially how they react when surgical instruments apply pressure. This process can provide doctors with detailed data on static anatomical structures and predict the dynamic changes that will occur in blood vessels during surgery, such as biomechanical properties such as elastic deformation and tensile strength of blood vessels. These data provide a vital reference for surgical planning, helping doctors predict possible vascular exposure risks, thereby optimizing surgical paths and preventing vascular injuries.

[0052] In order to better capture the complex distribution characteristics of blood vessels, a hierarchical branch complexity model was further introduced. This model analyzes the complexity of the vascular network in layers, mainly considering geometric characteristics such as the number of vascular branches, branch angles, and diameter changes. Vascular areas with a large number of branches, steep angles, and significant diameter changes usually have a higher risk of exposure, because the blood vessels in these areas are more likely to be inadvertently exposed or damaged during surgical operations. Through this hierarchical branch complexity analysis, the difficulty of vascular exposure during surgery can be accurately predicted at different levels, helping doctors to make more detailed surgical plans for high-risk areas.

[0053] The second step is dynamic surgical simulation and vascular exposure scenario construction. First, the system uses the physics engine to dynamically simulate the surgical process, simulating the interaction between surgical instruments and blood vessels and surrounding tissues under different operation paths, especially when the instruments contact the tissues around the blood vessels, which may cause tissue deformation and micro-displacement. The physics engine can accurately simulate the physical interaction between instruments and soft tissues during surgery, helping to predict the force, deformation response and displacement changes of blood vessels when the instruments are pressed or moved. This simulation not only provides a visual surgical scene, but also provides quantitative reference data for the risks of vascular exposure that may occur during surgery.

[0054] On this basis, the system constructs a spatiotemporal vascular exposure prediction model, which combines the dynamic movement trajectory of the instrument during the operation with the time series data, and can predict the probability of vascular exposure at different time points. The core of this model is to capture the continuity of surgical operations and the causal relationship between different operation steps through spatiotemporal data streams. The spatiotemporal data stream records the location information, operation speed, and impact of surgical instruments on blood vessels and surrounding tissues. To improve the prediction ability of the model, the system introduces two deep learning algorithms: long short-term memory network (LSTM) and temporal convolutional network (TCN). The LSTM network can capture the long-term dependency of instrument operations during surgery, especially in terms of the delayed impact of a certain operation on subsequent vascular exposure. It retains the key operation history during surgery through its memory unit structure and predicts the risk of vascular exposure at future time points. On the other hand, TCN processes time series data through a convolutional network, which can effectively capture the rapid changes and local patterns of surgical instruments in a short time window, and is particularly suitable for analyzing the impact of continuous movement of instruments in a short period of time on vascular exposure during surgery. By combining the advantages of LSTM and TCN, the system can capture the temporal relationship between surgical instrument movement and vascular exposure in real time during surgery, thereby accurately predicting the probability of vascular exposure at a specific time point.

[0055] The third step is a vascular exposure prediction model based on the fusion of global and local factors. The core of the design of this model is to integrate the joint analysis of local and global features to achieve more accurate vascular exposure prediction. Local features mainly include geometric parameters directly related to the blood vessels themselves, such as the diameter and branching angle of the blood vessels. These features can directly reflect the specific morphological structure of the blood vessels and the difficulty of exposure during surgery. The global features focus more on the distribution pattern of blood vessels in the entire liver and their relationship with surrounding tissues, such as the density of adjacent tissues and the distribution pattern of blood vessels in the liver. This global information is crucial for vascular exposure prediction, because the liver has a complex tissue structure and the risk of vascular exposure in different parts will be affected by the characteristics of adjacent tissues. For example, higher density tissues may provide more support for blood vessels and reduce the risk of exposure.

[0056] On this basis, a multimodal data fusion model is constructed by integrating multimodal physiological data. Multimodal data includes different information sources such as imaging data from CT, MRI, physiological parameters collected in real time during surgery (such as blood flow, blood pressure), and historical surgical data. These data come from different sources and in different forms, but each type of data can reflect the possibility of vascular exposure from different angles. By integrating these data, a more comprehensive prediction model can be formed.

[0057] In order to optimize the data fusion of different information sources, the model adopts the method of ensemble learning. Ensemble learning is a method to improve prediction performance by combining multiple models. Here, local features, global features and multimodal data are jointly analyzed through weighted fusion. The data of each information source are given different weights in the overall model according to their impact on vascular exposure risk. These weights are determined through data training, which can ensure that the contribution of each data in the model is reasonably evaluated and utilized.

[0058] The fourth step is real-time feedback during surgery and adaptive model optimization. The core of this step is to use the data collected in real time during surgery, especially ultrasound and endoscopic imaging data, combined with a deep learning model to dynamically correct the vascular exposure prediction. During surgery, as the instrument moves and the tissue deforms, the pre-established static model may no longer be accurate. Therefore, the introduction of real-time data enables the model to be adaptively adjusted based on the latest anatomical conditions. By inputting real-time imaging data into the deep learning network, the model can analyze this new data and correct the predicted results according to changes that occur during surgery. By learning a large amount of historical surgical data and real-time image patterns, the deep learning model can accurately identify the vascular and tissue features in the image to ensure the accuracy and flexibility of the prediction.

[0059] To perform this real-time correction, the model uses a convolutional neural network (CNN) to perform convolution operations on real-time image data. The convolution layer can extract spatial features in the image, such as the edges and shapes of blood vessels and changes in the density of surrounding tissues. This operation can dynamically update the spatial distribution information of blood vessels in a real-time environment, helping the model to recalculate the regional distribution of expected exposed blood vessels. If instrument contact in a certain area during surgery causes local tissue displacement or pressure changes, the convolution operation can quickly identify these changes and feed them back to the model for adjustment. After receiving the new data, the model adaptively recalculates the risk of vascular exposure in different areas and presents these results to the surgeon in real time, helping the surgeon make more accurate operational decisions at critical moments.

[0060] Embodiment 1:

[0061] Combination Figure 2 , Figure 3 A 45-year-old patient is undergoing right hepatic lobectomy. Before the operation, imaging data of the patient's liver and blood vessels were collected through CT and MRI data. The doctor hopes to use this model to predict intraoperative vascular exposure to ensure the safety of the surgical path and reduce the risk of possible vascular injury.

[0062] First, based on the patient's CT and MRI images, the system reconstructs the patient's liver and its vascular structure in three dimensions. To describe the complex geometry of the liver and blood vessels, x, y, and z represent the three-dimensional coordinates of the space. This geometric function can capture the morphology of the liver and blood vessels, including geometric features such as the length, diameter, and branch angle of the blood vessels. In order to ensure that the vascular exposure model can reflect the actual situation during surgery, the deformation behavior of the blood vessels under different surgical pressure conditions must also be considered. Therefore, the model combines local biomechanical properties to capture the local deformation velocity field v (x, y, z), which is used to represent the velocity distribution and deformation response of blood vessels and their surrounding tissues when subjected to instrument pressure.

[0063] The geometric model is described by the following formula:

[0064]

[0065] In this formula, γ i is the weight coefficient in the geometric model, which is used to control the influence of different anatomical regions. Due to the different tissue properties and vascular density in different regions of the liver, γ i The value of may be different in different regions. It can be set in the main blood vessel area, i The value range of γ is 0.8-1.0, indicating that the geometric characteristics of this area have a greater impact on the surgical process. In the smaller branch vessel area, γ i The value range of is 0.4-0.6, which means that the geometric features of these areas have little effect on the overall deformation.

[0066] In the equation is the Laplace operator, which is used to describe the degree of change between points in a geometric structure. This operator can reflect the deformation trend of perivascular tissue under different surgical pressures, especially those areas with more severe deformation. Doctors can identify areas of blood vessels that may cause significant deformation during surgery and predict risks in advance.

[0067] The local deformation velocity field v(x, y, z) represents the velocity distribution of blood vessels and surrounding tissues when they are subjected to force. It can reflect the dynamic response of blood vessels and surrounding tissues when surgical instruments apply pressure to blood vessels. Through this velocity field, doctors can foresee whether the blood vessels will be displaced or deformed at the moment when the surgical instrument applies pressure, thereby affecting the surgical path. In the right lobe area of ​​the patient's liver, the maximum value of the velocity field v(x, y, z) is set to 5mm / s, indicating that when the instrument applies pressure, the maximum deformation speed of the blood vessels and surrounding tissues is 5 millimeters per second. In smaller branch blood vessel areas, the value of the velocity field may be lower, about 2-3mm / s, indicating that the deformation of these areas is small and relatively stable.

[0068] Finally, κ is the coupling coefficient, which indicates the degree of coupling between the velocity field and the geometric shape. By adjusting the value of K, the system can control the degree of deformation of the blood vessel under different pressure conditions. In the trunk vessel area, the value of κ may range from 0.7-0.9, indicating that the pressure of the device in this area will cause a more significant deformation response, while in the branch vessel area, the value of K ranges from 0.3-0.5, indicating that these areas are less deformed and insensitive to force.

[0069] Through the calculation of the above formula, the system can predict in real time the deformation behavior and exposure risk of different vascular regions when subjected to pressure from surgical instruments during surgery. For example, during the surgical simulation of this patient, the system calculated the main trunk vessels of the right lobe of the liver. The function value changes rapidly after pressure is applied, and combined with the local biomechanical properties and velocity field distribution, it predicts that the area may be exposed during surgery and reminds the doctor to avoid the area or reduce pressure during the operation. In addition, in the area of ​​small blood vessel branches, the system calculates The smaller changes in function values ​​indicate that these areas are relatively safe during surgery and have a lower risk of vascular exposure.

[0070] The 45-year-old patient was undergoing right hepatic lobectomy. Doctors had used CT and MRI imaging data to reconstruct a three-dimensional structural model of the patient's liver and blood vessels before the operation. In order to further predict the risk of vascular exposure and rupture during surgery, the system will use finite element analysis (FEA) to simulate the stress-strain relationship of the blood vessels, thereby predicting the dynamic behavior of the blood vessels that may occur when the surgical instrument applies pressure.

[0071] Finite element analysis captures the deformation behavior of blood vessels under different pressure conditions through stress-strain analysis. Stress refers to the force applied by the device to the blood vessel, while strain describes the deformation of the blood vessel due to external forces. In order to dynamically simulate these phenomena, the system introduces the following stress-strain analysis formula:

[0072]

[0073] In this formula, is the local pressure, which represents the force applied by the surgical instrument to the blood vessel at time t and position x. Set in the main blood vessel area of ​​the right lobe of the liver, the pressure applied by the doctor when operating the instrument is about 1000-1200 Pa (Pascal). After this pressure value is entered into the formula, The temporal trend of local pressure can be reflected through the integration process.

[0074] The stress-strain influencing factors λ1 and λ2 control the effects of stress and contact force on blood vessels, respectively. In the right lobe of the liver of this patient, the λ1 value range is set to 0.8-1.2, which is mainly used to control the influence of stress, because the blood vessels in this area have a larger diameter and are subjected to relatively uniform external forces. λ2 is used to control the influence of changes in contact force on blood vessels. In the branch vessel area, the value of λ2 is set to 0.5-0.7, indicating that the blood vessels in the branch area are more sensitive to changes in external forces, especially when the instrument contacts a small displacement, which may cause the strain behavior of the blood vessels to be more obvious.

[0075] The rate of change of strain over time represents the strain rate of the blood vessel at time t. In the actual calculation process, it is set at the initial stage of the operation, and the applied pressure causes the instantaneous deformation rate of the main blood vessel to be 0.02mm / s, while the deformation rate of the branch blood vessel is higher, about 0.05mm / s. These rates reflect the instantaneous deformation behavior of blood vessels in different regions under the pressure of surgical instruments, especially the deformation caused by excessive pressure or improper operation of the instrument during the operation.

[0076] Δ(x, t) is the spatial distribution function of the contact force between surgical instruments and blood vessels. It is assumed that during the operation, the contact force between the instrument and the main blood vessels of the right lobe of the liver is distributed relatively evenly, with a contact force of 100Pa per unit area. For branch vessels, the spatial distribution of contact force is relatively uneven and may be concentrated in certain specific areas, such as a local area where the contact force reaches 200Pa. By introducing this contact force function, the system can accurately capture the force applied by the instrument to different positions of the blood vessels during surgery, thereby better predicting the deformation and exposure risk of blood vessels in different areas.

[0077] Assuming the operation lasted for 5 seconds, the system calculated each time segment using a formula. For example, in the first second, the doctor applied a lower pressure, and the system calculated the local pressure of the main blood vessels of the right lobe of the liver. At the 3rd second, due to the rapid movement of the instrument and the increase in pressure, the system calculated that the local pressure increased to 1200Pa, and the strain rate of the branch blood vessels reached 0.05mm / s. At this time, the local deformation was large, and the system prompted the doctor that the risk of blood vessel exposure increased, and it was recommended to slow down the operation or adjust the pressure point.

[0078] From the above examples, we can see that the system can accurately predict the exposure and rupture risks of blood vessels in different areas during surgery by combining stress-strain formulas with actual data for dynamic calculations. For example, by reasonably setting the values ​​of λ1 and λ2, the system can perform personalized control of the stress and contact force of the main and branch vessels. By analyzing the dynamic response of Δ(x, t) and Δ(x, t), the system can not only predict the instantaneous deformation behavior of blood vessels when surgical instruments apply pressure, but also help doctors adjust operations based on real-time feedback to avoid potential risks of vascular damage.

[0079] The 45-year-old patient's right lobectomy has entered a critical stage. In order to ensure the safety of the surgical instrument operation, the doctor decided to further use the spatiotemporal coupling model to accurately predict the risk of vascular exposure during the operation. By integrating static geometric information with dynamic stress-strain relationships, the model can capture the potential impact of surgical operations on blood vessels in both time and space dimensions, especially predicting the dynamic deformation and exposure risk of blood vessels under pressure changes.

[0080] First, the core formula of the space-time coupling model is:

[0081]

[0082] in, is the output of the spatiotemporal coupling model, which represents the comprehensive result of the static geometric information and dynamic stress response of the blood vessel at position x and time t. This value is used to help doctors assess the exposure risk of blood vessels during surgery. In order to better explain this formula and its practical application, it will be explained in detail through specific examples.

[0083] At the current stage of the operation, the doctor needs to evaluate the exposure of the main and branch vessels of the right lobe of the liver. First, η(x) is the weight coefficient that controls the impact of static geometric features on the overall model response. Set in the main vessel area of ​​the right lobe of the liver, the value range of η(x) is 0.9-1.0, because the static geometric features of the main vessels have a very important impact on the overall surgical exposure risk. In the branch vessel area, the value range of η(x) is smaller, possibly 0.4-0.6, reflecting that these areas have a weaker impact in the static state. This weight coefficient can ensure that the model makes accurate assessments on local static geometric features by reflecting the importance of blood vessels in different regions.

[0084] then, The gradient of the geometric morphology reflects the shape change in the static geometric features. In the main vessels of the right lobe of the liver, the gradient is small and the shape change is not large, indicating that the vascular structure is relatively regular and stable. In the branch vessel area, the gradient changes greatly and the shape change is more complex. The gradient range is set to 0.01-0.05 in the main area, while the gradient in the branch area may be as high as 0.08-0.12. This change indicates that deformation is more likely to occur in the branch vessel area, so special attention should be paid to these areas during surgery.

[0085] μ(t) is a time-dependent dynamic deformation function that describes the dynamic response characteristics of blood vessels at different times. At the first second of the operation, due to the small pressure applied by the instrument, the dynamic deformation function μ(t) value of the trunk blood vessel is 0.02. At the third second, as the pressure applied by the instrument increases, the value of μ(t) increases to 0.05, indicating that the deformation of the blood vessel gradually intensifies. In the branch vessel area, the dynamic response is more significant. Set within the same time period, the value of μ(t) increases from 0.05 to 0.1, which indicates that these blood vessels are more likely to undergo larger deformations when under pressure, and the exposure risk increases.

[0086] Finally, ξ(x, t) is a spatiotemporal coupling factor, which represents the combined effect of spatial and temporal stress changes on vascular deformation. Through this factor, the system can capture the pressure changes of blood vessels at different spatial points and different time periods. It is assumed that during surgery, the pressure changes of the instrument in the main blood vessel area of ​​the right lobe of the liver are relatively stable, and the value range of ξ(x, t) is 0.6-0.8. In the branch blood vessel area, due to the complexity of the instrument operation and the unevenness of the contact force, the value range of ξ(x, t) may be 0.3-0.5, reflecting that the stress changes in these areas are more significant.

[0087] It is the second-order derivative of pressure in space, which can capture the rate of change of pressure applied by surgical instruments on blood vessels at different locations. In the trunk blood vessel area, the set pressure changes more evenly, and the value of the second-order derivative is small, about 0.02, which means that the pressure applied by the instrument to this area is relatively stable. In the branch blood vessel area, the pressure changes more dramatically, and the second-order derivative value may be 0.1, indicating that the pressure changes in these areas are more complex, and there may be aggravated vascular deformation caused by local pressure, which increases the risk of exposure.

[0088] By substituting these data into the formula, the exposure risk of blood vessels at a specific time point and spatial position can be calculated. For example, at the 3rd second of the operation, the system calculates that at a certain position x of the main trunk blood vessel of the right lobe of the liver, The value of is 0.7, indicating that the risk of vascular exposure in this area is relatively low. In a high-risk area of ​​a branch vessel, the system calculates The value is 1.2, which indicates to doctors that there is a high risk of vascular exposure. It is recommended to reduce pressure or adjust the surgical path during the operation.

[0089] This spatiotemporal coupling model can comprehensively assess the risk of vascular exposure during surgery by combining static geometric features with dynamic stress-strain analysis, and help doctors adjust operations in real time during surgery to minimize the risk of vascular exposure. The calculation results show the feasibility of the model, which can not only provide real-time risk assessment, but also effectively reduce uncertainty during surgery by dynamically adjusting prediction results.

[0090] Embodiment 2:

[0091] Combination Figure 3 A 45-year-old patient is undergoing right hepatic lobectomy. The doctor wants to assess the risk of surgical instruments exposing the main and branch vessels of the right hepatic lobe at a specific moment. The system forms a spatiotemporal data stream through real-time collected ultrasound, endoscopic image data, instrument position information, and pressure sensor feedback data. The system uses the following spatiotemporal model formula:

[0092]

[0093] Step 1: Get real-time data

[0094] Set the pressure applied by the surgical instrument to the main blood vessels of the right lobe of the liver at a certain moment to 800 Pa , while the pressure applied in the branch vessel area is 1000 Pa. Pressure gradient The velocity of the device is 0.05 Pa / mm in the trunk area and 0.08 Pa / mm in the branch area. It is 2 mm / s in the main vessel area and 1 mm / s in the branch vessel area.

[0095] Step 2: Substitute into the formula

[0096] Substituting these data into the formula, the exposure risks of the main vessels and branch vessels were calculated separately.

[0097] Calculation of exposure risk in major vascular territories

[0098] In the main vascular area, the pressure gradient 0.05Pa / mm, instrument speed The speed is 2mm / s. If the operation starts at the 0th second and the doctor continues the operation until the 3rd second, the time period from 0 to 3 seconds needs to be integrated:

[0099]

[0100] Compute this integral:

[0101]

[0102] Therefore, at 3 seconds, the vascular exposure risk level in the trunk vessel area is 0.3, which indicates that the exposure risk in this area is low and the doctor can continue to operate according to the current path.

[0103] Calculation of exposure risk for branch vessel territories

[0104] In the branch vessel area, the pressure gradient 0.08Pa / mm, the speed of the device is 1mm / s. Similarly, calculate the time period from 0 seconds to 3 seconds:

[0105]

[0106] Compute this integral:

[0107]

[0108] Therefore, at 3 seconds, the risk of vascular exposure in the branch vessel area is 0.24. Although slightly higher than that in the trunk area, it is still within an acceptable range. However, if the pressure or speed increases, this value may increase rapidly, suggesting that the doctor should make appropriate adjustments to the operation.

[0109] Step 3: Interpret the calculation results:

[0110] From the above calculations, it can be seen that at the 3rd second, the exposure risk of the main vessels of the right lobe of the liver is 0.3, which is relatively low, while the exposure risk of the branch vessel area is 0.24, which is relatively high but still within the controllable range. Based on these calculation results, the system can provide real-time feedback to doctors, indicating whether the current operation is safe or whether it is necessary to reduce the pressure of the instrument or adjust the surgical path.

[0111] A 45-year-old patient is undergoing a right lobectomy of the liver, and the doctor hopes to use more advanced dynamic surgical simulation tools to predict vascular exposure during future operations. To this end, the system introduces long short-term memory networks (LSTM) and temporal convolutional networks (TCN) to process time series data, helping doctors better control the impact of instruments on blood vessels during surgery. Through the combination of LSTM and TCN, the system can learn the operation history of the instrument and predict the risk of vascular exposure in the future, thereby improving the safety and accuracy of the operation.

[0112] During surgery, LSTM is responsible for capturing long-term dependencies, that is, the delayed impact of a surgical operation on vascular exposure in the future. TCN is good at handling short-term operational changes and capturing rapid changes in a short period of time during surgery. By combining these two networks, the system can effectively cope with the complexity of long-term and short-term dependencies during surgery. The system's vascular exposure prediction formula is as follows:

[0113] ε t =γ1·σ(W x ·x t +W h ·h t-1 +b)

[0114] In this formula, ε tIt represents the probability of vascular exposure at the current time t and is the final output of the system. This value reflects the risk of vascular exposure that may be caused by the operation of surgical instruments at the current moment, and doctors can adjust their surgical strategies based on this prediction result.

[0115] When the operation is in the 5th second, the system needs to predict the risk of blood vessel exposure in the next 3 seconds (6th to 8th seconds). t Including the position, speed, and contact information of the surgical instrument with the blood vessels. At the 5th second, the instrument is located at the main blood vessel of the right lobe of the liver, the applied pressure is 900Pa, the instrument speed is 2.5mm / s, and the contact area with the main blood vessel is 20mm 2 This data is fed into the network and processed by LSTM and TCN to generate predictions.

[0116] First, x t Represents the input data at the current time point t=5, which includes pressure, velocity and contact area. Set the weight matrix W of the input data x The value range of is 0.5-0.7, which is used to control the contribution of each input feature to the model prediction. The current weight is 0.6. The implicit state h at the previous moment t-1 The operation history information of the 4th second is retained. It is set that at the 4th second, the pressure applied by the instrument is 850Pa, the instrument speed is 2.2mm / s, and the contact area is 18mm 2 This historical information is passed to the current moment through the hidden state and is transmitted through the weight matrix W h Adjust its influence, W h The value range is 0.6-0.8, and the value is 0.7 at this time.

[0117] Next, the activation function σ is used as a nonlinear function to map the output of the model to a range of probability values. σ is set to the sigmoid function, which can map the output result to between 0 and 1. Finally, the bias term b is used to balance the output result of the model. The value range of b is set to 0.1-0.3, and is currently set to 0.2.

[0118] Substitute these data into the formula to calculate:

[0119] ε t =0.6·σ((0.6·900+0.7·850)+0.2)

[0120] First, calculate the weighted sum of the input data and the hidden state:

[0121] 0.6·900=540.0.7·850=595

[0122] After summing:

[0123] 540+595+0.2=1135.2

[0124] Next, the sigmoid activation function is used to calculate:

[0125] σ(1135.2)≈1

[0126] Therefore, the vascular exposure risk probability ε at the current time t = 5 seconds is t Close to 1, this means that the risk of blood vessel exposure caused by the current surgical operation is very high. Based on this result, the system will prompt the doctor to reduce pressure or adjust the position of the instrument in real time to reduce the risk of exposure.

[0127] Through the combination of LSTM and TCN, the system can not only calculate the risk of vascular exposure in real time based on the current operation, but also predict the risk at future moments based on historical data. Assuming that the doctor continues to operate at the same speed and pressure, the system can predict the risk from the 6th to the 8th second. Based on the data from the previous few seconds, the system predicts that at the 6th second, the risk of vascular exposure is still high, about 0.95, while at the 7th and 8th seconds, the doctor reduces the pressure, and the system predicts that the risk drops to 0.7 and 0.5 respectively. Doctors can adjust operations based on these predicted values ​​to ensure the safety of the operation.

[0128] By introducing LSTM and TCN, the system can accurately predict the risk of vascular exposure based on the operation history of surgical instruments and current input data. The data and calculation process in the example show that this method can capture the complexity of long-term and short-term dependencies during surgery and provide real-time decision support for doctors.

[0129] The right lobectomy of a 45-year-old patient has entered a critical stage, and the doctor hopes to use more accurate prediction tools to determine the risk of vascular exposure in the future. To this end, the system introduces a comprehensive vascular exposure probability prediction model, which can combine the spatial interaction and time series dynamics of surgical instruments to generate a probability value for vascular exposure, helping doctors optimize surgical operations and reduce potential risks.

[0130] The core formula of this model is:

[0131]

[0132] The output of this formula It represents the probability of blood vessel exposure at the future time t+Δt, and predicts future risks based on the current operation of surgical instruments. In order to understand this formula, the application of the model will be explained in detail through specific data and calculation process, and the role of each variable in the operation will be explained.

[0133] Assume that at the current time t = 5 seconds, the doctor is operating the instrument to apply pressure to the main blood vessels of the right lobe of the patient's liver. The speed is 2.5 mm / s, the pressure is 900 Pa, and the contact area between the device and the blood vessel is 25 mm 2 These data are input into the system in real time, and the system calculates the exposure risk level of the blood vessels at the current time t = 5 seconds. Set the system to calculate the current exposure risk This indicates that there are certain risks in the current surgical operation.

[0134] α is a factor used to adjust the confidence of the model output, which indicates the doctor's confidence in the prediction results. Its value is usually adjusted according to the complexity of the surgical scene and the doctor's experience. In this operation, the doctor chooses α = 0.9, indicating a high degree of confidence in the model's prediction results. This value will play a weighted role in the final predicted probability of vascular exposure.

[0135] Changes in machine speed

[0136] During surgery, changes in the speed of surgical instruments have a significant impact on the risk of vascular exposure. Capture the rate of change of the moving speed of the device (i.e. acceleration or deceleration). Assume that the speed of the current device increases to 3 mm / s from t = 5 seconds to t = 6 seconds, so the acceleration is This means that accelerated operation of the device during this time period may result in an increased risk of vascular exposure, which will be calculated by the system.

[0137] The integration area Ω represents the three-dimensional space in the surgical scene. The system needs to calculate each spatial point in the entire surgical operation scene. The integration area is set to the main blood vessels of the right lobe of the liver and the nearby branch vessels. The volume of the main blood vessels is about 500mm 3 , and the volume of the branch vessels is 150mm 3 The three-dimensional integral region Ω of the entire surgical scene includes these two parts.

[0138] Substitute these data into the formula to calculate:

[0139]

[0140] First calculate the product part of the integral term:

[0141] 0.7.0.5=0.35

[0142] Next, perform spatial integration on the entire integration area Ω. Set the integration result of the main vessels to 350 mm 3 , while the integral result of the branch vessels is 105mm 3, then the total integral result is:

[0143] ∫ Ω 0.35dx dy dz=0.35·(350+105)=0.35·455=159.25

[0144] Finally, multiply the integral result by the adjustment factor α = 0.9:

[0145]

[0146] Through calculation, the system concluded that the predicted probability of vascular exposure at the future time t+Δt=6 seconds is 143.325. Although the absolute value of this value is high, it represents a relative quantitative risk, suggesting that the doctor's current operation may lead to a higher risk of vascular exposure. Based on this result, the system recommends that the doctor slow down the operation or reduce the pressure applied to reduce the risk of vascular exposure in the future.

Claims

1. A method for constructing a vascular exposure prediction model during liver resection anatomy, characterized by: S1. Deep anatomical modeling of liver vascular distribution: S1.

1. Construct the geometric structure of the liver and blood vessels based on CT and MRI data, and introduce biomechanical characteristics modeling; the model combines biomechanical principles, and simulates the stress and strain behavior of blood vessels under different physiological conditions through finite element analysis (FEA), and the response when surgical instruments apply pressure; provide static anatomical structures, and the prediction model provides basic data on dynamic factors of vascular elasticity and toughness; S1.

2. Introduce a hierarchical branch complexity model to classify the vascular branch network according to the number, angle, and diameter variation characteristics of the branches, and analyze the difficulty of vascular exposure at different levels; S2. Dynamic surgical simulation and vascular exposure scenario construction: S2.

1. Through dynamic surgical simulation based on a physical engine, establish vascular exposure scenarios under different operation paths during surgery, including predicting deformation or micro-displacement caused when the instrument contacts the tissue around the blood vessel during surgery; S2.

2. Construct a spatiotemporal vascular exposure prediction model to predict the predictability of vascular exposure at different time points. Based on the spatiotemporal data flow, the temporal relationship between intraoperative instrument movement and vascular exposure is captured through the long short-term memory network (LSTM) or the temporal convolutional network (TCN), and the probability of vascular exposure at a specific time point is predicted. S3. Vascular exposure prediction model based on fusion of global and local factors: S3.

1. Design a fusion model to jointly analyze local geometric features including vessel diameter and branching angle with global features including vessel distribution pattern and density of adjacent tissues; S3.

2. Construct a multimodal data fusion model by integrating multimodal physiological data; perform weighted fusion of data from different information sources through ensemble learning methods; S4. Real-time feedback during surgery and model adaptive optimization: S4.

1. Introduce ultrasound and endoscopic image data collected in real time during surgery, and combine them with a deep learning model for real-time correction. The model adaptively adjusts the prediction results based on the new input information; through the convolution operation of the real-time image, the regional distribution of the expected exposed blood vessels is recalculated.

2. The method for constructing a prediction model for vascular exposure during liver resection according to claim 1, characterized in that The in-depth anatomical modeling of liver vascularity includes: Based on the patient's CT and MRI images, the liver and blood vessel structures are first reconstructed in three dimensions; the morphology of the blood vessels is captured, and the deformation behavior of the blood vessels under different surgical pressure conditions is perceived; the geometric model is constructed through functions Describe, where is a function that defines the three-dimensional geometry of the liver and blood vessels, where x, y, and z represent the three-dimensional coordinates of the space; capturing the complex geometry and distribution characteristics of the blood vessels, the equation combines the local biomechanical properties, including the local deformation velocity field v(x, y, z), and is described by the following formula: Among them, γ i is the weight coefficient in the geometric model, which is used to control the influence of different anatomical regions; is the Laplace operator It is used to describe the degree of change between points in the geometric structure; v(x, y, z) is the local deformation velocity field, which represents the velocity distribution of blood vessels and surrounding tissues when subjected to force, and reflects the dynamic response of blood vessels; κ is the coupling coefficient, which represents the degree of coupling between the velocity field and the geometric shape, and is used to control the degree of deformation of blood vessels under different conditions.

3. The method for constructing a prediction model for vascular exposure during liver resection anatomy according to claim 2, characterized in that The in-depth anatomical modeling of the liver vascular distribution includes: finite element analysis (FEA) to simulate the dynamic behavior of blood vessels through stress-strain relationships and predict the risk of rupture or exposure during surgery.

4. The method for constructing a prediction model for vascular exposure during liver resection according to claim 3, characterized in that The deep anatomical modeling of the liver vascular distribution includes: creating a spatiotemporal coupling model by fusing static geometric information with dynamic stress-strain relationships to predict the dynamic deformation and exposure risk of blood vessels during surgery.

5. The method for constructing a prediction model for vascular exposure during liver resection anatomy according to claim 1, characterized in that The dynamic surgical simulation and vascular exposure scenario construction includes: Various data are collected in real time during the operation, including real-time images from ultrasound and endoscopes, instrument position information, and pressure sensor feedback data; multi-dimensional data forms a dynamic spatiotemporal data stream to describe the interaction between instrument movement, pressure application, and blood vessels; the spatiotemporal data stream contains spatial dimensions x, y, z, combined with the time dimension t, to create a dynamic scene containing spatiotemporal information; by recording the trajectory of surgical instruments in real time, the pressure changes of instruments on blood vessels at different time points are captured, and then a spatiotemporal model is established: in represents the risk level of vascular exposure at spatial coordinates x, y, z and time t; the function describes the predictability of vascular exposure at a specific time and spatial point during surgery; Represents the pressure gradient applied by the surgical instrument, describing the surgical instrument at position x, y, z and time t ′ Changes in pressure exerted on blood vessels; Represents the position x, y, z and time t ′ The velocity of the instrument at the position; this velocity term is used to capture the trajectory of the surgical instrument, t ′ : It is the integral time variable, which is integrated from the initial time to the current time t, and describes the comprehensive impact of the device on the blood vessel in the past period of time.

6. The method for constructing a prediction model for vascular exposure during liver resection anatomy according to claim 5, characterized in that The dynamic surgical simulation and vascular exposure scenario construction includes: By introducing the long short-term memory network LSTM and the temporal convolutional network TCN, the long-term and short-term dependencies in the time series are learned and predicted; based on the action sequence of surgical instruments in different time periods, the vascular exposure at future moments is predicted; TCN efficiently processes short-term patterns in the time series through convolution operations and captures rapid changes within a short period of time during surgery.

7. The method for constructing a prediction model for vascular exposure during liver resection anatomy according to claim 6, characterized in that The dynamic surgical simulation and vascular exposure scenario construction includes: By dynamically combining the spatial interaction and time series of surgical instruments and blood vessels, a comprehensive vascular exposure probability prediction model is generated, and the probability value of vascular exposure is output to guide surgical operation optimization: in, It represents the probability value of vascular exposure at the future time t+Δt; the final result predicted by the model is used to reflect the vascular exposure risk at a specific future moment; α is the adjustment factor used to control the confidence of the model output; Indicates the exposure risk level of the blood vessel at the current time t, based on the spatiotemporal movement and pressure of the surgical instrument; It represents the derivative of the instrument movement speed with respect to time, capturing the acceleration and deceleration of the surgical instrument movement; it is used to analyze the impact of the rapid movement of the instrument on the risk of vascular exposure; Ω is the integration area, representing the three-dimensional space of the entire surgical operation scene. The integration process is to calculate all spatial points in the surgical scene to generate a comprehensive vascular exposure probability.

Images