CT guided liver puncture scheme optimization method and system based on big data analysis

Abstract

The invention provides a CT guided liver puncture scheme optimization method and system based on big data analysis, and relates to the technical field of intelligent medical treatment. According to the method, liver CT images under multiple breathing phases are collected, a liver motion database is constructed, and a three-dimensional simulation human body model is driven to achieve real breathing response; puncture operation is executed on the model, real-time track data are collected, a virtual CT image system is constructed by fusing structure images, and dynamic visualization of a puncture path is achieved; the space error of the track end point and the target point is further analyzed, a big data puncture path library is established, feature extraction and clustering modeling are combined, an optimal puncture path scheme matched with an individual structure is output, and the operation precision and the training intelligence level are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart medical technology, and in particular to a CT-guided liver puncture plan optimization method and system based on big data analysis. Background Art

[0002] CT-guided liver puncture has been widely used in tumor biopsy and ablation therapy, offering advantages such as accurate positioning and minimal invasiveness. However, due to limitations such as CT's non-real-time imaging, significant liver movement changes with respiration, and complex puncture paths, the accuracy of the procedure relies heavily on the physician's experience and spatial visualization, leading to the continued risk of complications such as mispuncture and damage to surrounding tissues.

[0003] With the development of new-generation information technologies such as virtual reality, 3D printing, posture tracking, magnetic navigation, and image fusion, the construction of high-fidelity simulation training platforms for physicians' preoperative rehearsal and procedural learning has become a trend. While some VR platforms for surgical training have been developed overseas, systems for CT-guided puncture scenarios, including respiratory position simulation, soft tissue deformation feedback, and real-time display of puncture trajectories, remain in the experimental or undeveloped stages.

[0004] Currently, training systems generally lack the ability to realistically reproduce the dynamic displacement and tissue deformation of the liver driven by the patient's breathing, and are unable to achieve synchronous integration of the puncture needle trajectory and virtual CT imaging during the puncture process. This results in a large difference between training and actual operations, making it difficult to effectively improve the safety and accuracy of clinical punctures, and limiting their application and promotion before minimally invasive tumor treatment. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide a CT-guided liver puncture plan optimization method and system based on big data analysis, which improves the accuracy and visualization level of liver puncture operation and realizes individualized path optimization and intelligent training support based on big data.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A CT-guided liver puncture plan optimization method based on big data analysis, comprising:

[0008] Acquiring liver CT image data under multiple respiratory states, and extracting three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database;

[0009] constructing a three-dimensional simulated human body model with dynamic liver response capability driven by breathing based on the liver motion database, and performing CT scanning on the three-dimensional simulated human body model to obtain real structural image data for virtual training;

[0010] Performing a puncture operation on the simulated human body model and acquiring real-time puncture trajectory data using a puncture needle with an integrated posture sensor;

[0011] registering and fusing the real-time puncture trajectory data with the real structural image data, and inputting the fused data into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states;

[0012] Based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a large data puncture path library containing multi-user operation records;

[0013] The big data puncture path library is used to perform statistical analysis and cluster modeling, and an optimal puncture path plan is output for clinical or training reference.

[0014] Preferably, the method further comprises: acquiring liver CT image data under multiple respiratory states, and extracting three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database, including:

[0015] Instructing the subject to perform multiple sets of liver CT volume scans in deep inhalation, deep exhalation, and calm breathing breath-holding states, respectively, to obtain the liver CT image data;

[0016] performing image segmentation of the liver region based on the liver CT image data, and extracting the upper and lower diameters, left and right diameters, anterior and posterior diameters, and volume of the liver at each respiratory phase to characterize morphological parameters of the liver;

[0017] Calculate the position of the liver's center of mass in the three-dimensional coordinate system at each respiratory phase to obtain the liver's three-dimensional position;

[0018] Based on the coordinates of the liver's center of mass at multiple respiratory phases, the spatial displacement vector of the liver's center of mass is calculated to describe the overall movement of the liver caused by respiratory motion.

[0019] The morphological parameters, the three-dimensional position and the spatial displacement vector are associated and stored to obtain the liver motion database.

[0020] Preferably, the calculation formula of the spatial displacement vector is:

[0021]

[0022] in, is the average spatial displacement vector of the liver mass center in a complete respiratory cycle, N is the number of respiratory phases collected, represents the three-dimensional coordinates of the liver center of mass in the i-th respiratory phase, is the displacement smoothing adjustment factor, which is used to suppress the disturbance effect of abnormal phase jump, where σ is the empirically determined respiratory deformation scale parameter, which is determined by the sample root mean square displacement.

[0023] Preferably, a three-dimensional simulated human body model with dynamic liver response capability driven by breathing is constructed based on the liver motion database, and a CT scan is performed on the three-dimensional simulated human body model to obtain real structural image data for virtual training, including:

[0024] Determining the three-dimensional modeling dimensions of the simulated liver and thoracoabdominal structures based on the average size parameters and center of mass displacement range extracted from the liver motion database;

[0025] 3D printing technology is used to manufacture the human skeleton, soft tissue layer, and simulated liver capsule. The simulated liver capsule is filled with a particle-elastomer hybrid material with elastic deformation capabilities and connected to an external airbag structure to simulate the deformation response under respiratory actuation.

[0026] Constructing a closed abdominal cavity structure outside the simulated liver capsule, and driving the abdominal cavity airbag to change periodically by an air pump device to reproduce the center of mass displacement amplitude recorded in the liver motion database;

[0027] The constructed three-dimensional simulated human body model is placed on a CT scanning platform, and three-dimensional image data under different respiratory drive states are collected as the real structure image data for virtual training.

[0028] Preferably, performing a puncture operation on the simulated human body model and obtaining real-time puncture trajectory data using a puncture needle with an integrated posture sensor includes:

[0029] A puncture needle with a hollow handle structure is provided, and a nine-axis attitude sensor module is embedded in the hollow handle structure to detect the spatial attitude and position changes of the puncture needle in real time;

[0030] Inserting the puncture needle into the abdominal wall of the three-dimensional simulated human body model at a specified incident angle to simulate the liver puncture path under clinical CT guidance;

[0031] During the puncture process, the angle change and displacement information of the puncture needle in the three-dimensional space along the time series is continuously collected and output in the form of attitude quaternion to obtain the real-time puncture trajectory data.

[0032] Preferably, the real-time puncture trajectory data is registered and fused with the real structural image data, and the fused data is input into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states, including:

[0033] Establishing a three-dimensional reconstruction model based on the CT scan image of the three-dimensional simulated human body model, and constructing a spatial reference coordinate system consistent with the posture sensor of the puncture needle;

[0034] Converting each data point in the real-time puncture trajectory data into the reference coordinate system of the three-dimensional reconstructed model to form a corresponding position vector;

[0035] A rigid registration algorithm is used to spatially register the puncture path with the liver structure, generating fused three-dimensional image data containing the puncture needle trajectory and liver structure.

[0036] The three-dimensional image data is input into a virtual CT image system to enable puncture path tracing and analysis under different respiratory states; the virtual CT image system has a built-in dynamic visualization interface that supports multi-phase switching, arbitrary angle rotation and path playback.

[0037] Preferably, based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a large data puncture path library containing multi-user operation records, including:

[0038] Setting the target area in the virtual image system, wherein the target area is set as a preset spherical or ellipsoidal space voxel according to the voxel coordinates of the lesion center;

[0039] Calculating a spatial error vector of a single puncture based on a Euclidean distance between an end point of the real-time puncture trajectory data and a center of the target point;

[0040] Associating and annotating the error vector with the complete trajectory, starting position, puncture angle, and needle insertion depth parameters of the real-time puncture trajectory data to generate a structured puncture record;

[0041] The puncture records of multiple users in multiple case scenarios are numbered and archived, and the big data puncture path library containing multiple users and multiple scenarios is constructed, and a unified standard format is set.

[0042] Preferably, the calculation formula of the spatial error vector is:

[0043]

[0044] in, is the spatial error vector of a single puncture operation, indicating the spatial deviation between the puncture endpoint and the target point; The coordinates of the end point of the puncture trajectory data in three-dimensional space; is the target center coordinate; θ devθ is the angle between the puncture needle endpoint direction vector and the target center line, which is used to reflect the directional deviation of the puncture path; ref is a preset reference angle constant, which indicates the reference value of the allowable direction error; ∈ is a minimum positive real number that prevents the denominator from being zero.

[0045] Preferably, the big data puncture path library is used to perform statistical analysis and cluster modeling to output an optimal puncture path plan for clinical or training reference, including:

[0046] Extracting features from the historical trajectory data in the big data puncture path library to obtain feature data; the feature data includes the starting point position, needle insertion angle, path length, end point error vector modulus, and direction deviation angle;

[0047] Constructing a multidimensional path feature vector set based on the feature data, and clustering similar paths using a density-based clustering algorithm to obtain multiple clusters;

[0048] Performing mean path calculation on the path trajectory within each cluster to obtain a representative path within the cluster;

[0049] Based on the real structure image data generated by the three-dimensional simulated human body model corresponding to the object to be trained, the representative path within the cluster adapted to the spatial structure is matched and output as the optimal puncture path solution.

[0050] A CT-guided liver puncture plan optimization system based on big data analysis, including:

[0051] a liver motion data construction unit, configured to obtain liver CT image data under multiple respiratory states, and extract the three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database;

[0052] a dynamic simulation model generation unit, configured to construct a three-dimensional simulation human body model having a dynamic liver response capability driven by breathing based on the liver motion database, and perform a CT scan on the three-dimensional simulation human body model to obtain real structural image data for virtual training;

[0053] a puncture trajectory acquisition unit, configured to perform a puncture operation on the simulated human body model and acquire real-time puncture trajectory data using a puncture needle with an integrated posture sensor;

[0054] an image fusion and playback unit, configured to register and fuse the real-time puncture trajectory data with the real structural image data, and input the fused data into a virtual CT image system for playback and analysis, so as to achieve dynamic visualization of the puncture path under different respiratory states;

[0055] A spatial error analysis and path recording unit, configured to analyze the spatial error between the real-time puncture trajectory data and the target point based on the virtual image system, so as to construct a large data puncture path library containing multi-user operation records;

[0056] The path cluster analysis and recommendation unit is used to perform statistical analysis and cluster modeling using the big data puncture path library, and output the optimal puncture path plan for clinical or training reference.

[0057] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0058] By constructing a respiratory-driven simulated human model, capturing puncture trajectories in real time, and integrating structural imaging data, this method enables dynamic visualization and error quantification analysis of puncture paths under different respiratory conditions. Furthermore, a large-scale path library is constructed based on multi-user operation data, and the optimal path solution is output. This effectively improves the accuracy of puncture training, individual adaptability, and scientific nature of path decision-making, and has excellent clinical practical value and training promotion prospects. This invention improves the accuracy and visualization of liver puncture operations, and realizes individualized path optimization and intelligent training support based on big data. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0060] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0061] Figure 2 A schematic diagram of the technical route provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0063] The purpose of the present invention is to provide a CT-guided liver puncture plan optimization method and system based on big data analysis, which improves the accuracy and visualization level of liver puncture operation and realizes individualized path optimization and intelligent training support based on big data.

[0064] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0065] Figure 1 A flow chart of the method provided in the embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a CT-guided liver puncture plan optimization method based on big data analysis, comprising:

[0066] Step 100: Acquire liver CT image data under multiple respiratory states, and extract the three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database;

[0067] Step 200: constructing a three-dimensional simulated human body model with a dynamic liver response capability driven by breathing based on the liver motion database, and performing a CT scan on the three-dimensional simulated human body model to obtain real structural image data for virtual training;

[0068] Step 300: performing a puncture operation on a simulated human body model, and obtaining real-time puncture trajectory data using a puncture needle with an integrated posture sensor;

[0069] Step 400: registering and fusing the real-time puncture trajectory data with the real structural image data, and inputting the fused data into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states;

[0070] Step 500: Based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a big data puncture path library containing multi-user operation records;

[0071] Step 600: Utilize the big data puncture path library to perform statistical analysis and cluster modeling, and output the optimal puncture path plan for clinical or training reference.

[0072] Preferably, the method further comprises: acquiring liver CT image data under multiple respiratory states, and extracting three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database, including:

[0073] Instructing the subject to perform multiple sets of liver CT volume scans in deep inhalation, deep exhalation, and calm breathing breath-holding states, respectively, to obtain the liver CT image data;

[0074] performing image segmentation of the liver region based on the liver CT image data, and extracting the upper and lower diameters, left and right diameters, anterior and posterior diameters, and volume of the liver at each respiratory phase to characterize morphological parameters of the liver;

[0075] Calculate the position of the liver's center of mass in the three-dimensional coordinate system at each respiratory phase to obtain the liver's three-dimensional position;

[0076] Based on the coordinates of the liver's center of mass at multiple respiratory phases, the spatial displacement vector of the liver's center of mass is calculated to describe the overall movement of the liver caused by respiratory motion.

[0077] The morphological parameters, the three-dimensional position and the spatial displacement vector are associated and stored to obtain the liver motion database.

[0078] Specifically, to enable the subsequent virtual training platform to faithfully reproduce the dynamic behavior of the liver during the respiratory cycle, it is necessary to model the spatial position changes and structural deformations of the liver under representative respiratory states and construct a structured liver motion database. This step must meet the following core data requirements: first, the three-dimensional center position of the liver at each phase; second, the trajectory of the liver's center of mass as it changes with respiration; and third, the scale information of its volume and morphology. Without any of these data, subsequent respiratory-driven simulation or dynamic alignment of the puncture path will not be possible.

[0079] Furthermore, this step does not rely on a real-time respiratory monitoring system. Instead, it utilizes a phased breath-hold acquisition method to obtain liver CT data during three phases: deep inspiration, deep expiration, and quiet breathing. All three states must be acquired continuously from the same body position to ensure structural stability relative to surrounding tissues. Data is collected only once for each state, without involving time-series imaging, facilitating standardization and data comparison.

[0080] Furthermore, among the three sets of image data obtained, only the liver body related areas are retained for library construction, and the whole abdomen data is not used. This step does not use general image segmentation tools, but directly performs three-dimensional spatial screening based on the liver shape template established by the present invention, retaining only the main liver body voxel area and excluding the lobe edges and adjacent blood vessels and gallbladder adhesion areas. This screening template is established through actual scanning data, does not rely on public atlases, and has a clear boundary control threshold. The maximum up and down, height and width, and front-to-back axial dimensions of the liver body are automatically calculated from the extreme points of the area, and manual estimation and input are not allowed.

[0081] Furthermore, the three-dimensional center position of the liver in each respiratory state is automatically extracted using a voxel-weighted center point method. This must be automatically generated by the system within the intact liver tissue area, and manual annotation is not allowed. To obtain the center of mass trajectory, the center points of the three states must be arranged in sequence, and the system automatically calculates the displacement path between adjacent states. All center of mass data must be uniformly converted into millimeter-level positioning coordinates in the Cartesian coordinate system used by the simulation system to support subsequent model driving and coordinate alignment.

[0082] Exemplarily, the resulting database structure includes: unique identification, scan phase label, liver volume, maximum dimensions in three dimensions, spatial center position, and continuous displacement path. This structure is the sole data source for simulation model construction and path error dynamic simulation in this invention. It must be calculated using real image data and cannot be replaced by simulated data.

[0083] Preferably, the calculation formula of the spatial displacement vector is:

[0084]

[0085] in, is the average spatial displacement vector of the liver mass center in a complete respiratory cycle, N is the number of respiratory phases collected, represents the three-dimensional coordinates of the liver center of mass in the i-th respiratory phase, is the displacement smoothing adjustment factor, which is used to suppress the disturbance effect of abnormal phase jump, where σ is the empirically determined respiratory deformation scale parameter, which is determined by the sample root mean square displacement.

[0086] In the present invention, in order to accurately depict the overall spatial displacement trend of the liver during natural breathing, a spatial displacement vector calculation method based on a discrete centroid point sequence is used to construct an average motion expression model of the liver's respiratory response. The calculation of the spatial displacement vector is not based on the maximum / minimum value difference or endpoint connection method, but rather adopts a weighted average method of the centroid difference between consecutive respiratory phases, and introduces a disturbance suppression mechanism to enhance the robustness to abnormal phase changes. Specifically, the position of the liver centroid under multiple calibrated respiratory phases is used as a sequence input in the present invention, and the three-dimensional displacement vectors between adjacent phases are extracted in sequence, and then all adjacent displacement vectors are weighted summed. Unlike simple equal-weighted averaging, this scheme introduces a smoothing adjustment factor to implement an exponential penalty on the amplitude of adjacent displacement vectors. If the centroid difference between two phases is abnormally large, the impact of this factor on the overall average vector will be automatically reduced, avoiding the problem of model error expansion due to unstable individual respiratory control. The setting of the adjustment factor is based on the root mean square displacement amplitude of the entire group of centroid sequences as a scale reference, without the need for user intervention or manual setting of hyperparameters, ensuring the engineering portability and clinical applicability of the algorithm. Furthermore, this weighted approach exhibits greater physical consistency than the traditional mean vector method, preserving the liver's tendency to slowly shift with respiration while simultaneously shielding data from irregular disturbances. This significantly improves the accuracy of dynamic liver morphology restoration driven by simulation. This spatial displacement vector is used to control the subsequent periodic movement of the simulation model in three-dimensional space and serves as a key coordinate correction benchmark in the dynamic matching mechanism for the puncture path.

[0087] Preferably, a three-dimensional simulated human body model with dynamic liver response capability driven by breathing is constructed based on the liver motion database, and a CT scan is performed on the three-dimensional simulated human body model to obtain real structural image data for virtual training, including:

[0088] Determining the three-dimensional modeling dimensions of the simulated liver and thoracoabdominal structures based on the average size parameters and center of mass displacement range extracted from the liver motion database;

[0089] 3D printing technology is used to manufacture the human skeleton, soft tissue layer, and simulated liver capsule. The simulated liver capsule is filled with a particle-elastomer hybrid material with elastic deformation capabilities and connected to an external airbag structure to simulate the deformation response under respiratory actuation.

[0090] Constructing a closed abdominal cavity structure outside the simulated liver capsule, and driving the abdominal cavity airbag to change periodically by an air pump device to reproduce the center of mass displacement amplitude recorded in the liver motion database;

[0091] The constructed three-dimensional simulated human body model is placed on a CT scanning platform, and three-dimensional image data under different respiratory drive states are collected as the real structure image data for virtual training.

[0092] Alternatively, in the present invention, a three-dimensional simulated human body model with respiratory-driven responsiveness is constructed. The key is to achieve the passive deformation capability and controlled spatial displacement behavior of the simulated liver in the physical structure, and the displacement range of the structure under controlled drive strictly corresponds to the center of mass movement amplitude recorded in the database. The simulated liver adopts a double-layer structure. Its outer shell is a closed capsule structure, the membrane material has compressibility and stable resilience, and the interior is filled with a medium-sized particle-elastomer composite medium. On the one hand, the medium can undergo reversible flow deformation under the action of air pressure, and on the other hand, it can maintain the consistency and self-support of the overall tissue under the constraints of the outer shell. The liver capsule is not a closed unit, but is connected to an external closed air capsule cavity through a catheter system. The external air capsule cavity is placed below the abdominal cavity of the model to form a complete cavity around the bottom of the liver capsule, and the interior of the cavity is driven by an air pump to form periodic pressure changes.

[0093] Specifically, the pressure change is not set by programming, but the pressure control parameters are inferred based on the displacement amplitude of the liver's center of mass between respiratory phases in the database. The system calculates the gas filling and releasing ratio required for the pressure change per unit volume of the air cavity based on the maximum and minimum range of the liver's displacement in the Z axis (vertical direction), and generates a set of control instructions that limit the filling and releasing rate based on the actual volume conduction relationship of the sac. The air pump device uses this parameter as a target to perform periodic alternation of positive and negative pressures, controlling the outer cavity to achieve synchronous expansion and retraction within the simulated respiratory frequency, thereby driving the liver sac to produce continuous and predictable physical displacement behavior in the vertical direction. The abdominal cavity structure is a closed and stable shell, and the internal space is precisely aligned with the contour of the simulated liver to prevent the simulated liver from undergoing lateral displacement or rotation in an undesirable direction under the action of air pressure, ensuring that the movement path is consistent with actual physiological changes. The closed abdominal cavity and the skeleton structure are positioned using an interlocking assembly method, without the need for any metal rigid brackets, thereby ensuring that the CT imaging process does not produce artifacts.

[0094] Furthermore, after completing the above structure, this embodiment places the model on a CT platform, sets the static pressure state according to the pressure instructions corresponding to the three typical respiratory phases in the database, and maintains a fixed time in each state to ensure that the tissue is fully shaped. A full-body CT scan is performed each time the pressure is static, and a total of three sets of image data are obtained, corresponding to the simulated liver structure in the inspiratory phase, expiratory phase and mid-phase state respectively. The obtained image is directly input as "real structure image data" for the virtual path playback system, which is unified with the spatial coordinate system of the subsequent path trajectory superposition analysis to form the image space reference system in the training process. Since the image acquisition process is derived from the true restoration of displacement driven by the database, the image has a high degree of physiological consistency in the mapping of spatial structure and dynamic trajectory.

[0095] Preferably, performing a puncture operation on the simulated human body model and obtaining real-time puncture trajectory data using a puncture needle with an integrated posture sensor includes:

[0096] A puncture needle with a hollow handle structure is provided, and a nine-axis attitude sensor module is embedded in the hollow handle structure to detect the spatial attitude and position changes of the puncture needle in real time;

[0097] Inserting the puncture needle into the abdominal wall of the three-dimensional simulated human body model at a specified incident angle to simulate the liver puncture path under clinical CT guidance;

[0098] During the puncture process, the angle change and displacement information of the puncture needle in the three-dimensional space along the time series is continuously collected and output in the form of attitude quaternion to obtain the real-time puncture trajectory data.

[0099] Specifically, in order to obtain realistic puncture trajectory data with temporal continuity and spatial resolution, the present invention integrates a nine-axis attitude sensor into the puncture needle, enabling real-time capture of the needle's three-dimensional attitude and position changes during the puncture process. The puncture needle utilizes a hollow handle structure, encapsulating a miniature nine-axis sensor module. This module, comprising a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, outputs standardized attitude quaternion data after hardware filtering and coordinate transformation. In a simulated clinical procedure, the needle is inserted at a specified angle through the abdominal wall of a simulated mannequin and traverses soft tissue to the target liver region, continuously capturing and recording the needle's spatial trajectory. The sensor outputs real-time needle attitude change data at a millisecond frequency. Combined with initial position calibration information, a continuous displacement trajectory is generated through inertial accumulation and magnetic correction algorithms, generating a complete time-series puncture path. This path data provides the sole input for subsequent trajectory registration, spatial error analysis, and path clustering modeling.

[0100] Preferably, the real-time puncture trajectory data is registered and fused with the real structural image data, and the fused data is input into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states, including:

[0101] Establishing a three-dimensional reconstruction model based on the CT scan image of the three-dimensional simulated human body model, and constructing a spatial reference coordinate system consistent with the posture sensor of the puncture needle;

[0102] Converting each data point in the real-time puncture trajectory data into the reference coordinate system of the three-dimensional reconstructed model to form a corresponding position vector;

[0103] A rigid registration algorithm is used to spatially register the puncture path with the liver structure, generating fused three-dimensional image data containing the puncture needle trajectory and liver structure.

[0104] The three-dimensional image data is input into a virtual CT image system to enable puncture path tracing and analysis under different respiratory states; the virtual CT image system has a built-in dynamic visualization interface that supports multi-phase switching, arbitrary angle rotation and path playback.

[0105] In the present invention, in order to achieve the spatial fusion of the three-dimensional puncture path and the simulated liver structure, the first requirement is to establish a spatial reference alignment mechanism between the puncture trajectory data and the model CT image. This step does not rely on an external tracking system or image recognition algorithm, but completes the one-to-one correspondence between the trajectory data and the image data through the construction of a homologous coordinate reference and the mapping of a rigid transformation matrix. Specifically, the simulated human body model determines the correspondence between its spatial origin position, the direction of the axis and the scanning position during the construction phase. This relationship is written into the spatial calibration parameters of the image data during CT scan image reconstruction and is used to generate a three-dimensional reconstructed image coordinate system consistent with the physical model. The posture quaternion and displacement trajectory data collected by the puncture needle sensor have been aligned with the simulation model coordinates through static calibration at the acquisition starting point, so they can be directly written into the image coordinate system through mathematical transformation to ensure that the spatial accuracy of the trajectory point positioning in the image space is not less than 1.5mm.

[0106] The spatial registration of trajectory data is not processed by endpoint or linear fitting, but is achieved through a global rigid registration algorithm. That is, with the goal of minimizing the total Euclidean distance between the trajectory point cloud and the liver structure boundary point cloud, the trajectory is automatically rotated and translated so that its path shape and image structure are globally aligned in three-dimensional space, rather than locally overlapping. This registration method can effectively avoid path misalignment caused by posture starting errors or flexible deformation of the model, ensuring that the spatial direction of the puncture path in the entire liver structure and the error calculation are clinically valid. Any trajectory data directly rendered without spatial registration cannot meet the spatial consistency requirements and cannot be used for error calculation or path scoring.

[0107] After the fusion is completed, the trajectory and image data are uniformly encapsulated into a resolvable three-dimensional structure and input into the virtual CT image system designed by the present invention. The system only retains the liver voxel layer and the puncture path voxel layer, and establishes an index mapping across respiratory phases, supports synchronous switching using respiratory phase labels defined in the database, and replays the relative changes of the path within the simulation structure in a time-series manner. The system allows the operator to freely rotate the view and enter any section, tracking the dynamic offset changes of the end of the path relative to the target point in continuous phases, and performing spatial distance measurement and angle evaluation. Since the path data in the system is derived from real-time acquisition and undergoes coordinate correction, registration and reconstruction processing, the puncture behavior displayed by the system can fully reflect the operational error, forming a traceable, quantifiable and comparable path evaluation mechanism.

[0108] Preferably, based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a large data puncture path library containing multi-user operation records, including:

[0109] Setting the target area in the virtual image system, wherein the target area is set as a preset spherical or ellipsoidal space voxel according to the voxel coordinates of the lesion center;

[0110] Calculating a spatial error vector of a single puncture based on a Euclidean distance between an end point of the real-time puncture trajectory data and a center of the target point;

[0111] Associating and annotating the error vector with the complete trajectory, starting position, puncture angle, and needle insertion depth parameters of the real-time puncture trajectory data to generate a structured puncture record;

[0112] The puncture records of multiple users in multiple case scenarios are numbered and archived, and the big data puncture path library containing multiple users and multiple scenarios is constructed, and a unified standard format is set.

[0113] In the present invention, in order to achieve quantitative evaluation of puncture path accuracy and continuous optimization of training strategies, it is necessary to perform spatial error analysis on the results of each puncture operation in a virtual image system and establish an archivable large data path library in a structured form. This process is based on the spatial voxel coordinates of the target target to construct a spatial target area with clear boundaries. It is preferably set as a spherical or ellipsoidal spatial voxel with a central coordinate corresponding to the center of mass of the lesion and a radius determined by the tumor safety tolerance error, so that the system can automatically determine the hit situation and relative error amplitude of the puncture result. The system calculates the spatial distance between the coordinates of the end point of the puncture trajectory and the center of the target voxel. The error vector is formed by the spatial difference between the two. This error not only records the spatial accuracy of the path end point, but can also be derived for subsequent path fitting and evaluation function modeling. It is the only quantitative source for puncture accuracy evaluation.

[0114] In order to completely convert the single operation data into archivable samples, this embodiment performs structured extraction of the core attributes of each path, including: the spatial coordinates of the starting point on the model surface, the initial angle vector of the path, the maximum depth of the puncture needle and its corresponding timestamp, the entire sequence of path points in the real-time trajectory, and the spatial error vector calculated from the end position and the target position. The system packages these data into a unified data recording unit, and sets the standard field order, data format and naming conventions, and adds the user number, case ID, simulation structure version and operation time information for archiving. All records are structured and organized according to multi-user, multi-case and multi-parameter dimensions to form the big data puncture path library described in the present invention. This library not only supports statistical analysis and path clustering, but can also be used for path recommendation model training. Without error vector normalization or trajectory parameter structured processing, it will not be possible to form a high-quality path sample set with analysis and learning value.

[0115] Preferably, the calculation formula of the spatial error vector is:

[0116]

[0117] in, is the spatial error vector of a single puncture operation, indicating the spatial deviation between the puncture endpoint and the target point; The coordinates of the end point of the puncture trajectory data in three-dimensional space; is the target center coordinate; θ dev θ is the angle between the puncture needle endpoint direction vector and the target center line, which is used to reflect the directional deviation of the puncture path; ref is a preset reference angle constant, which indicates the reference value of the allowable direction error; ∈ is a minimum positive real number that prevents the denominator from being zero.

[0118] In the present invention, in order to achieve a refined assessment of the accuracy of the puncture operation, it is necessary not only to calculate the geometric offset between the puncture needle end point and the target center, but also to introduce a directional error factor to comprehensively reflect the deviation of the puncture path in terms of spatial positioning and incident direction. The spatial error vector calculation method proposed in the present invention does not adopt the conventional Euclidean distance model, but introduces an improved calculation structure with directional angle weights. This method takes the spatial vector from the puncture end point to the target center as the error principal component and multiplies it by a correction factor related to the degree of deviation of the puncture direction. The correction factor is determined by the nonlinear amplification coefficient composed of the square of the angle between the puncture end point posture vector and the target direction vector. The larger the angle, the more the puncture direction deviates from the target center line, and the error vector is exponentially amplified; when the directions are consistent, the correction factor approaches 1. In order to avoid unstable behavior caused by the denominator approaching zero due to an angle that is too small, a constant minimum value is introduced into the formula as a denominator compensation term to ensure numerical stability. The above-mentioned modified structure breaks through the limitations of the traditional path endpoint scoring method in terms of directional considerations, and can effectively identify the operational differences between "deflecting but hitting" and "straight stabbing but missing the target", providing more discriminative original features for path classification and clustering modeling.

[0119] Preferably, the big data puncture path library is used to perform statistical analysis and cluster modeling to output an optimal puncture path plan for clinical or training reference, including:

[0120] Extracting features from the historical trajectory data in the big data puncture path library to obtain feature data; the feature data includes the starting point position, needle insertion angle, path length, end point error vector modulus, and direction deviation angle;

[0121] Constructing a multidimensional path feature vector set based on the feature data, and clustering similar paths using a density-based clustering algorithm to obtain multiple clusters;

[0122] Performing mean path calculation on the path trajectory within each cluster to obtain a representative path within the cluster;

[0123] Based on the real structure image data generated by the three-dimensional simulated human body model corresponding to the object to be trained, the representative path within the cluster adapted to the spatial structure is matched and output as the optimal puncture path solution.

[0124] Optionally, in the present invention, the core of the puncture path cluster analysis is to perform structured modeling of the path behavior and extract discriminative operational features. The system extracts five core features from each structured puncture path record, including the starting point position, initial needle insertion angle, path geometric length, end point error vector modulus, and direction deviation angle, to construct a standardized set of path feature vectors. This set is used to measure operational differences and strategy aggregation and cannot be replaced by a single geometric parameter. After normalization, the feature data is input into the density-driven clustering module. There is no need to preset the number of clusters. The system automatically forms multiple path clusters based on the distribution density of path features in high-dimensional space to ensure that each cluster reflects a type of puncture pattern that is consistent in structure and strategy. For each cluster, the system calculates the point-by-point average trajectory of the intra-cluster trajectory on the spatial path to obtain a representative central path for use as a candidate for subsequent solutions.

[0125] Specifically, candidate paths must undergo a spatial adaptation test with the training object structure before they can be output as the optimal path solution. Based on the structural image of the three-dimensional simulated human body model corresponding to the object, the system extracts the liver boundary and the needle window position and establishes an individual spatial restriction model. The representative path in the cluster is projected into the structure, and the three-dimensional envelope is used to determine whether the path exists completely within the liver boundary, thereby excluding puncture failures or structural conflict paths. Paths that meet the conditions will be further scored according to historical hit rate, end point deviation, and incident angle safety, and the one with the highest score will be output as the recommended path for the individual.

[0126] Corresponding to the above method, such as Figure 2 As shown, this embodiment also provides a CT-guided liver puncture plan optimization system based on big data analysis, including:

[0127] a liver motion data construction unit, configured to obtain liver CT image data under multiple respiratory states, and extract the three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database;

[0128] a dynamic simulation model generation unit, configured to construct a three-dimensional simulation human body model having a dynamic liver response capability driven by breathing based on the liver motion database, and perform a CT scan on the three-dimensional simulation human body model to obtain real structural image data for virtual training;

[0129] a puncture trajectory acquisition unit, configured to perform a puncture operation on the simulated human body model and acquire real-time puncture trajectory data using a puncture needle with an integrated posture sensor;

[0130] an image fusion and playback unit, configured to register and fuse the real-time puncture trajectory data with the real structural image data, and input the fused data into a virtual CT image system for playback and analysis, so as to achieve dynamic visualization of the puncture path under different respiratory states;

[0131] A spatial error analysis and path recording unit, configured to analyze the spatial error between the real-time puncture trajectory data and the target point based on the virtual image system, so as to construct a large data puncture path library containing multi-user operation records;

[0132] The path cluster analysis and recommendation unit is used to perform statistical analysis and cluster modeling using the big data puncture path library, and output the optimal puncture path plan for clinical or training reference.

[0133] The beneficial effects of the present invention are as follows:

[0134] (1) The present invention constructs a liver motion database covering multiple respiratory states and drives a three-dimensional simulated human body model based on real anatomical parameters. It realizes the visual reconstruction and spatial registration of the puncture path in a dynamic respiratory scenario for the first time, overcoming the technical bottleneck of the existing static simulation system that cannot reflect the periodic displacement and deformation of the liver.

[0135] (2) The present invention introduces high-precision puncture trajectory data collected by posture sensors, combined with the dynamic rendering and path error calculation model of the virtual CT image system, which can quantitatively evaluate the spatial deviation during the operation and provide a traceable and comparable standardized data basis for individualized training and path optimization.

[0136] (3) The present invention constructs a multi-user, multi-scenario big data puncture path library, adopts feature extraction and clustering modeling methods, and forms a representative set of operation strategies. It can automatically screen the optimal path solution in combination with individual structures, and realize the structural adaptability and experience-driven fusion of path recommendation.

[0137] (4) The present invention significantly improves the intelligence and precision of CT-guided liver puncture training and preoperative planning, and has strong clinical application promotion value and potential for the development of intelligent medical training systems. It breaks through the limitations of existing methods in terms of poor repeatability, rough path evaluation, and lack of data-driven optimization methods.

[0138] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0139] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A CT-guided liver puncture plan optimization method based on big data analysis, characterized in that: include: Acquiring liver CT image data under multiple respiratory states, and extracting three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database; constructing a three-dimensional simulated human body model with dynamic liver response capability driven by breathing based on the liver motion database, and performing CT scanning on the three-dimensional simulated human body model to obtain real structural image data for virtual training; Performing a puncture operation on the simulated human body model and acquiring real-time puncture trajectory data using a puncture needle with an integrated posture sensor; registering and fusing the real-time puncture trajectory data with the real structural image data, and inputting the fused data into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states; Based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a large data puncture path library containing multi-user operation records; The big data puncture path library is used to perform statistical analysis and cluster modeling, and an optimal puncture path plan is output for clinical or training reference.

2. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: Also includes: Acquiring liver CT image data under multiple respiratory states, and extracting three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database, including: Instructing the subject to perform multiple sets of liver CT volume scans in deep inhalation, deep exhalation, and calm breathing breath-holding states, respectively, to obtain the liver CT image data; performing image segmentation of the liver region based on the liver CT image data, and extracting the upper and lower diameters, left and right diameters, and anterior and posterior diameters and volumes of the liver at each respiratory phase to characterize morphological parameters of the liver; Calculate the position of the liver's center of mass in the three-dimensional coordinate system at each respiratory phase to obtain the liver's three-dimensional position; Based on the coordinates of the liver's center of mass at multiple respiratory phases, the spatial displacement vector of the liver's center of mass is calculated to describe the overall movement of the liver caused by respiratory motion. The morphological parameters, the three-dimensional position and the spatial displacement vector are associated and stored to obtain the liver motion database.

3. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 2, characterized in that: The calculation formula of the spatial displacement vector is: in, is the average spatial displacement vector of the liver mass center in a complete respiratory cycle, N is the number of respiratory phases collected, represents the three-dimensional coordinates of the liver mass center in the i-th respiratory phase, is the displacement smoothing adjustment factor, which is used to suppress the disturbance effect of abnormal phase jump, where σ is the empirically determined respiratory deformation scale parameter, which is determined by the sample root mean square displacement.

4. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: A three-dimensional simulated human body model with a dynamic liver response capability driven by breathing is constructed based on the liver motion database, and a CT scan is performed on the three-dimensional simulated human body model to obtain real structural image data for virtual training, including: Determining the three-dimensional modeling dimensions of the simulated liver and thoracoabdominal structures based on the average size parameters and center of mass displacement range extracted from the liver motion database; 3D printing technology is used to manufacture the human skeleton, soft tissue layer, and simulated liver capsule. The simulated liver capsule is filled with a particle-elastomer hybrid material with elastic deformation capabilities and connected to an external airbag structure to simulate the deformation response under respiratory actuation. Constructing a closed abdominal cavity structure outside the simulated liver capsule, and driving the abdominal cavity airbag to change periodically by an air pump device to reproduce the center of mass displacement amplitude recorded in the liver motion database; The constructed three-dimensional simulated human body model is placed on a CT scanning platform, and three-dimensional image data under different respiratory drive states are collected as the real structure image data for virtual training.

5. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: Performing a puncture operation on the simulated human body model and acquiring real-time puncture trajectory data using a puncture needle with an integrated posture sensor includes: A puncture needle with a hollow handle structure is provided, and a nine-axis attitude sensor module is embedded in the hollow handle structure to detect the spatial attitude and position changes of the puncture needle in real time; Inserting the puncture needle into the abdominal wall of the three-dimensional simulated human body model at a specified incident angle to simulate the liver puncture path under clinical CT guidance; During the puncture process, the angle change and displacement information of the puncture needle in the three-dimensional space along the time series is continuously collected and output in the form of attitude quaternion to obtain the real-time puncture trajectory data.

6. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: The real-time puncture trajectory data is registered and fused with the real structural image data, and the fused data is input into a virtual CT image system for playback and analysis to achieve dynamic visualization of the puncture path under different respiratory states, including: Establishing a three-dimensional reconstruction model based on the CT scan image of the three-dimensional simulated human body model, and constructing a spatial reference coordinate system consistent with the posture sensor of the puncture needle; Converting each data point in the real-time puncture trajectory data into the reference coordinate system of the three-dimensional reconstructed model to form a corresponding position vector; A rigid registration algorithm is used to spatially register the puncture path with the liver structure, generating fused three-dimensional image data containing the puncture needle trajectory and liver structure. The three-dimensional image data is input into a virtual CT image system to enable puncture path tracing and analysis under different respiratory states; the virtual CT image system has a built-in dynamic visualization interface that supports multi-phase switching, arbitrary angle rotation and path playback.

7. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: Based on the virtual image system, the spatial error between the real-time puncture trajectory data and the target point is analyzed to construct a large data puncture path library containing multi-user operation records, including: Setting the target area in the virtual image system, wherein the target area is set as a preset spherical or ellipsoidal space voxel according to the voxel coordinates of the lesion center; Calculating a spatial error vector of a single puncture based on a Euclidean distance between an end point of the real-time puncture trajectory data and a center of the target point; Associating and annotating the error vector with the complete trajectory, starting position, puncture angle, and needle insertion depth parameters of the real-time puncture trajectory data to generate a structured puncture record; The puncture records of multiple users in multiple case scenarios are numbered and archived, and the big data puncture path library containing multiple users and multiple scenarios is constructed, and a unified standard format is set.

8. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: The calculation formula of the spatial error vector is: in, is the spatial error vector of a single puncture operation, indicating the spatial deviation between the puncture endpoint and the target point; The coordinates of the end point of the puncture trajectory data in three-dimensional space; is the target center coordinate; θ dev θ is the angle between the puncture needle endpoint direction vector and the target center line, which is used to reflect the directional deviation of the puncture path; ref is a preset reference angle constant, which indicates the reference value of the allowable direction error; ∈ is a minimum positive real number that prevents the denominator from being zero.

9. The method for optimizing CT-guided liver puncture plans based on big data analysis according to claim 1, characterized in that: Statistical analysis and cluster modeling are performed using the big data puncture path library to output the optimal puncture path plan for clinical or training reference, including: Extracting features from the historical trajectory data in the big data puncture path library to obtain feature data; the feature data includes the starting point position, needle insertion angle, path length, end point error vector modulus, and direction deviation angle; Constructing a multidimensional path feature vector set based on the feature data, and clustering similar paths using a density-based clustering algorithm to obtain multiple clusters; Performing mean path calculation on the path trajectory within each cluster to obtain a representative path within the cluster; Based on the real structure image data generated by the three-dimensional simulated human body model corresponding to the object to be trained, the representative path within the cluster adapted to the spatial structure is matched and output as the optimal puncture path solution.

10. A CT-guided liver puncture plan optimization system based on big data analysis, characterized in that: include: a liver motion data construction unit, configured to obtain liver CT image data under multiple respiratory states, and extract the three-dimensional position and morphological parameters of the liver in different respiratory phases from the liver CT image data to construct a liver motion database; a dynamic simulation model generation unit, configured to construct a three-dimensional simulation human body model having a dynamic liver response capability driven by breathing based on the liver motion database, and perform a CT scan on the three-dimensional simulation human body model to obtain real structural image data for virtual training; a puncture trajectory acquisition unit, configured to perform a puncture operation on the simulated human body model and acquire real-time puncture trajectory data using a puncture needle with an integrated posture sensor; an image fusion and playback unit, configured to register and fuse the real-time puncture trajectory data with the real structural image data, and input the fused data into a virtual CT image system for playback and analysis, so as to achieve dynamic visualization of the puncture path under different respiratory states; A spatial error analysis and path recording unit, configured to analyze the spatial error between the real-time puncture trajectory data and the target point based on the virtual image system, so as to construct a large data puncture path library containing multi-user operation records; The path cluster analysis and recommendation unit is used to perform statistical analysis and cluster modeling using the big data puncture path library, and output the optimal puncture path plan for clinical or training reference.

Images