A visualized guided gynecological single-port laparoscopic approach planning system

Abstract

The application discloses a visual guidance gynecological single-hole laparoscopic access planning system and belongs to the technical field of medical assistance. The system comprises the following steps: a three-dimensional mapping module is used to collect multi-modal abdominal images of a patient, and an abdominal dynamic deformation model is constructed with the navel hole as the center and containing the abdominal wall, blood vessels and the position of the uterus. An access optimization module simulates puncture operations in different directions, iteratively calculates the potential energy value of the access resistance in each direction in combination with the collision frequency and penetration depth. A thermal guidance module interpolates the potential energy value, generates a resistance thermal map and superimposes it on the real-time image of the body surface to realize intuitive visual guidance. A trajectory generation module searches the optimal path from the navel hole to the surgical target point in reverse according to the low-resistance access point selected by the operator, relying on the dynamic deformation model through a dynamic programming algorithm, and finally outputs the surgical access trajectory with labeled depth information to assist in completing the laparoscopic access planning.

Description

Technical Field

[0001] This invention relates to the field of medical auxiliary technology, specifically a visually guided single-port laparoscopic approach planning system for gynecology. Background Technology

[0002] Single-port laparoscopic surgery in gynecology involves resection of pelvic lesions through a single incision at the umbilicus. The choice of approach directly impacts surgical safety. The abdominal wall structure surrounding the umbilicus is complex, comprising skin, subcutaneous fat, muscle fascia, peritoneum, and abdominal wall vessels running through these layers. The resistance encountered by instruments varies significantly depending on the puncture angle and depth. Currently, in clinical practice, the puncture direction relies primarily on the surgeon's palpation, anatomical landmarks, and personal experience, lacking objective quantitative assessment methods to address individual differences in abdominal wall structure. Preoperative computed tomography (CT) or magnetic resonance imaging (MRI) can visualize the layered anatomy of the abdominal wall. Some studies utilize 3D reconstruction techniques to create a static abdominal model, simulating a straight puncture path on this model to eliminate puncture directions with significant vascular interference based on the positional relationships of anatomical structures.

[0003] However, static anatomical models fail to reflect the actual morphological changes in the abdominal wall bulge after pneumoperitoneum is established during single-port laparoscopic surgery. Pneumoperitoneum pressure causes significant nonlinear tensile deformation of the abdominal wall, altering the relative positions, thicknesses, and vascular pathways of different tissue layers. Static models reconstructed from supine images cannot accurately represent the tissue stiffness distribution during surgery, leading to significant discrepancies between simulated puncture resistance assessments and real-world conditions. Furthermore, existing methods rely solely on discrete anatomical structures for avoidance, lacking a continuous quantitative description of cumulative tissue resistance along the entire puncture direction. This makes it difficult to quickly identify safe areas with lower overall resistance among numerous potential approach angles. In addition, preoperative planning results are typically presented separately as 3D rendered images or numerical reports, failing to establish an intuitive spatial correspondence with real-time images of the patient's body surface. Surgeons still need to rely on memory and spatial imagination to map virtual planning information to their real-world vision during the procedure, limiting the guiding effect of the planning results on real-time operations.

[0004] To address the aforementioned issues, there is a need for an approach planning scheme that can quantify the puncture resistance in different directions based on the dynamic deformation of the pneumoperitoneum and visually integrate the resistance distribution into the body surface field. This would allow the surgeon to quickly select low-resistance areas under visual guidance and automatically obtain the minimum energy consumption path from the umbilicus to the target point. Summary of the Invention

[0005] This invention provides a visually guided single-port laparoscopic approach planning system for gynecology, aiming to solve the problems of existing static models being unable to express the deformation of pneumoperitoneum, leading to distorted assessment of approach resistance, and the lack of integration between planning information and real-time body surface images, resulting in insufficient guidance effect. By constructing a dynamic deformation model of the patient's abdomen and iteratively calculating the approach resistance potential energy, a resistance potential energy heat map superimposed on the real-time body surface image is generated. Combined with dynamic programming search, a low-resistance approach trajectory output from the umbilicus to the surgical target point is achieved.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a visually guided gynecological single-port laparoscopic approach planning system, which includes a three-dimensional mapping module, an approach optimization module, a thermal guidance module, and a trajectory generation module. By constructing an individualized dynamic deformation model of the patient's abdomen and quantitatively evaluating and visually displaying the puncture approach around the umbilicus, the system assists the operator in quickly selecting a safe and low-resistance approach direction and generating an executable approach trajectory.

[0007] A 3D mapping module is used to acquire preoperative multimodal abdominal imaging data of gynecological patients and map this data onto a 3D voxel mesh centered on the umbilicus, thereby constructing a dynamic deformation model of the patient's abdomen that includes abdominal wall layers, vascular pathways, and the relative position of the uterus. Specifically, voxel labels for skin contours, muscle fascia, peritoneum, and uterine boundaries are extracted from the multimodal abdominal imaging data. A polar coordinate system is established in the 3D voxel mesh with the geometric center of the umbilicus as the origin. Each voxel label is projected onto the polar diameter, polar angle, and depth dimensions to form a layered distance field of the abdominal wall tissue. This layered distance field is then input into a nonlinear finite element deformer. By adjusting the boundary constraints, the abdominal bulge morphology under different pneumoperitoneum pressures is simulated, generating a dynamic deformation model of the patient's abdomen that reflects the tissue deformation characteristics. Preferably, the nonlinear finite element deformer uses a tetrahedral hyperelastic material model to simulate the nonlinear stress-strain relationship of the abdominal wall tissue under pneumoperitoneum pressure, making the model closer to the real pneumoperitoneum state.

[0008] The approach optimization module drives a virtual single-port laparoscope to perform forward and backward simulations along multiple preset puncture directions in a dynamic deformation model of the patient's abdomen. It records the number of collisions and penetration depths between the virtual instrument and abdominal wall tissue voxels in each simulation, and propagates these collision counts and penetration depths back to the model's tissue stiffness field, iteratively updating the approach resistance potential energy value corresponding to each preset puncture direction. In specific implementation, for each preset puncture direction, the virtual single-port laparoscope is controlled to enter the model step-by-step according to a preset step-size sequence. At each step forward, the voxel label at the current position is queried. If it is a blood vessel or fascia, a collision count is accumulated; if it is fat or muscle, the penetration depth at the current step is recorded. The initial resistance value for that direction is obtained by weighted summing of the collision counts and penetration depths corresponding to each step length in the same puncture direction. Then, based on the difference between the initial resistance value and the preset resistance target, the Young's modulus of the tissue voxels corresponding to that direction in the model is adjusted in reverse. The simulation is repeated until the resistance value converges. The converged resistance value is the approach resistance potential energy value. As a preferred implementation, in each advance and retreat simulation, the closest distance between the end effector of the virtual single-port laparoscopy and the uterine boundary is also recorded. When the closest distance is less than the preset safety distance, a penalty potential energy term is added to the approach resistance potential energy value in the puncture direction, and surface interpolation is performed again to update the resistance potential energy value of the corresponding angle region of the resistance potential energy heatmap, thereby avoiding damage to critical structures such as the uterus during the approach.

[0009] The thermal guidance module performs surface interpolation on the access resistance potential energy value according to the circumferential angle of the umbilicus and the puncture depth, generating a resistance potential energy heat map covering all possible access directions around the umbilicus. This heat map is then overlaid on the real-time acquired abdominal surface image of the patient. During surface interpolation, the circumferential angle of the umbilicus is discretized into 360 angle intervals, and the puncture depth is discretized into depth intervals. The combination of each angle interval and depth interval constitutes a candidate access grid point. Using the access resistance potential energy value at each grid point as the height value, radial basis functions are used for smooth interpolation to obtain a continuous resistance potential energy surface. This surface is then projected layer by layer onto a two-dimensional plane according to the depth value, and the resistance potential energy of each layer is encoded with different colors to generate the resistance potential energy heat map. To achieve precise registration with the patient's body surface, the thermal guidance module also acquires the umbilical marker and abdominal contour edge features from the real-time acquired images of the patient's abdomen. It rigidly registers the umbilical center coordinates from the drag potential energy thermal map with the umbilical marker in the body surface image. Then, based on the abdominal contour edge features, it performs elastic deformation correction on the registered thermal map, ensuring that the thermal map boundary coincides with the body surface image boundary, and displays it as a semi-transparent texture overlay. As a further aspect of this invention, the rigid registration is based on the correspondence between the umbilical center and the two bilateral anterior superior iliac spines—three anatomical landmarks—in the two images to calculate the registration matrix, thereby ensuring the accuracy and stability of the overlay.

[0010] The trajectory generation module receives the low-resistance approach seed point selected by the operator on the resistance potential energy heatmap. Using the angle and depth corresponding to the seed point as initial approach parameters, it employs dynamic programming to perform a backward search for the minimum energy consumption path from the umbilicus to the surgical target in the dynamic deformation model of the patient's abdomen, outputting a single-port laparoscopic approach trajectory with depth markings. Specifically, a three-dimensional mesh is established in the model using the angle and depth corresponding to the low-resistance approach seed point as the starting node and the surgical target position as the ending node. The product of the approach resistance potential energy value of each mesh node and the spatial distance between adjacent nodes is used as the state transition cost. Using a dynamic programming algorithm, the module searches backward from the ending node, sequentially recording the previous node that minimizes the cumulative state transition cost, until it backtracks to the starting node. The sequence of backtracked nodes is used as the minimum energy consumption path, and the corresponding puncture depth is marked for each node. To facilitate surgical execution, the trajectory generation module can also obtain the spatial coordinates of each node on the path of minimum energy consumption and the corresponding tissue voxel type, calculate the angle between the spatial vectors of adjacent nodes in the order of nodes as the approach direction deflection angle, arrange the approach direction deflection angles of all nodes in order, and generate an approach trajectory direction sequence with depth marking.

[0011] In addition, the system includes an instrument path preview module, which compares the depth values ​​of each node in the output trajectory with the corresponding abdominal wall thickness in the patient's abdominal dynamic deformation model, marking nodes with depth values ​​greater than the abdominal wall thickness as perforation risk nodes. These perforation risk nodes are highlighted with warning colors on the resistance potential energy heatmap, and a virtual safety boundary box along the approach trajectory is generated in the model, thus visually indicating the potential tissue penetration risk. When the operator inputs a corrected depth for a perforation risk node, this module replaces the original depth value with the corrected depth, uses the replaced node as the new termination node, and re-triggers the trajectory generation module to search for the minimum energy consumption path and output the corrected approach trajectory, achieving interactive optimization of the approach planning results.

[0012] Through the collaborative work of the above modules, this invention integrates multimodal imaging, tissue biomechanical properties, and access resistance assessment, visually presenting the resistance distribution around the umbilicus in the form of a heat map, and automatically planning an access path that balances low resistance and safety. This significantly reduces the reliance on the surgeon's personal experience in single-port laparoscopic surgery access planning, and helps to reduce puncture resistance and improve access accuracy while avoiding blood vessels and important tissues.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: Multimodal abdominal imaging data was mapped onto a 3D voxel mesh centered on the umbilicus. Layered distance fields and nonlinear finite element deformers were used to simulate the abdominal bulge morphology under different pneumoperitoneal pressures, constructing a dynamic deformation model of the patient's abdomen that included abdominal wall layers, vascular orientation, and the relative position of the uterus. This model realistically reproduces the spatial redistribution of tissues under pneumoperitoneal conditions during surgery, allowing subsequent access resistance analysis to be based on an anatomical configuration consistent with the actual surgical environment. In each advance and retreat simulation, the number of collisions and penetration depths between the virtual single-port laparoscope and abdominal wall tissue voxels were recorded, and this collision information was backpropagated to the model's tissue stiffness field, iteratively updating the access resistance potential energy value corresponding to each preset puncture direction. This closed-loop iterative mechanism allows the Young's modulus of the tissue voxels to be gradually adjusted based on simulation feedback. The final converged access resistance potential energy value quantitatively characterizes the overall penetration resistance under dynamic deformation conditions for different puncture directions, avoiding systematic deviations between resistance assessment under static models and actual intraoperative sensations.

[0014] The resistance potential energy values ​​of the access route are interpolated using a surface method based on the circumferential angle of the umbilicus and the puncture depth to generate a resistance potential energy heatmap covering all possible access directions around the umbilicus. By acquiring umbilical markers and abdominal contour edge features from real-time acquired images of the patient's abdominal surface, the resistance potential energy heatmap is rigidly registered and elastically deformed, then overlaid as a semi-transparent texture. This overlay method allows the spatial distribution of resistance potential energy to be directly rendered onto the patient's actual abdominal wall field of view. The surgeon can intuitively see the level of access resistance in various angle regions around the umbilicus without spatial visualization, and quickly identify the low-resistance safety zone. After selecting a low-resistance access seed point on the resistance potential energy heatmap, the operator uses the angle and depth of the seed point as initial parameters and employs a dynamic programming algorithm in the dynamic deformation model of the patient's abdomen to search backward from the surgical target point to the umbilicus for the minimum energy consumption path, outputting a single-port laparoscopic access trajectory with depth markings. During path search, the product of the approach resistance potential energy and the spatial distance is used as the state transition cost, ensuring that the generated path is both the direction with the least anatomical resistance and maintains smooth continuity in depth changes. Each node in the output trajectory carries a depth marker and a sequence of directional deflection angles, providing quantifiable operational references for layer-by-layer puncture during surgery, reducing the probability of damage to high-risk tissues such as blood vessels and fascia. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a schematic diagram of a visually guided single-port laparoscopic approach planning system for gynecology. Figure 2 This is a schematic diagram illustrating the process of constructing a dynamic abdominal deformation model based on multimodal imaging. Figure 3 This is a schematic diagram of the process for iteratively calculating the inlet resistance potential energy value using the inlet optimization module; Figure 4 This is a flowchart of the generation and intraoperative superposition guidance of the resistance potential energy thermogram; Figure 5 It is a curve showing the radial boundary variation of different tissue layers of the abdominal wall at the circumferential angle of the umbilicus; Figure 6 This is a probability density distribution diagram of the initial and updated ingress resistance potential energy values. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] See Figure 1 This invention provides a visualization-guided gynecological single-port laparoscopic approach planning system, including a 3D mapping module, an approach optimization module, a thermal guidance module, and a trajectory generation module. The 3D mapping module acquires multimodal abdominal imaging data of gynecological patients before surgery, maps this data to a 3D voxel mesh centered on the umbilicus, and constructs a dynamic deformation model of the patient's abdomen including abdominal wall layers, vascular orientation, and the relative position of the uterus. The approach optimization module drives a virtual single-port laparoscope to perform advance and retreat simulations along multiple preset puncture directions within the dynamic deformation model of the patient's abdomen. It records the number of collisions and penetration depths between the virtual instrument and abdominal wall tissue voxels in each advance and retreat simulation, and propagates these collision counts and penetration depths back to the tissue stiffness field of the dynamic deformation model of the patient's abdomen, iteratively updating the approach resistance potential energy value corresponding to each preset puncture direction. The thermal guidance module interpolates the approach resistance potential energy value according to the circumferential angle of the umbilicus and the puncture depth, generating a resistance potential energy heat map covering all possible approach directions around the umbilicus, and overlays this heat map onto real-time acquired images of the patient's abdominal surface. The trajectory generation module receives the low-resistance entry seed point selected by the operator on the resistance potential energy heat map. Using the angle and depth corresponding to the low-resistance entry seed point as the initial entry parameters, it uses dynamic programming to search for the minimum energy consumption path from the umbilicus to the surgical target in the dynamic deformation model of the patient's abdomen, and outputs the single-port laparoscopic entry trajectory with depth marking.

[0019] Example 1: In specific implementation, please refer to Figure 2 The 3D mapping module acquires multimodal abdominal imaging data of gynecological patients before surgery. This multimodal abdominal imaging data includes computed tomography (CT) images and magnetic resonance imaging (MRI) images. The 3D mapping module performs image registration and fusion of the CT and MRI images to obtain fused data in a unified spatial coordinate system. The 3D mapping module then calls a pre-trained deep learning segmentation network to perform voxel-level semantic segmentation on the fused data, outputting a tissue category label for each voxel. Tissue categories include skin, muscle fascia, fat, peritoneum, blood vessels, uterine border, and bone. The segmentation network adopts a 3D U-Net architecture. The encoder contains four downsampling stages, each consisting of two 3D convolutional layers and a batch normalization layer. The decoder contains four symmetrical upsampling stages, with skip connections passing the encoder feature map to the corresponding decoder layer. The segmentation network is trained on a labeled abdominal image dataset using a weighted sum of the Dice loss function and the cross-entropy loss function. The output is a voxel label map with the same spatial resolution as the fused data, where each voxel label value is an integer code corresponding to the tissue category.

[0020] After obtaining the voxel label map, the 3D mapping module identifies the geometric center of the umbilicus. The geometric center of the umbilicus is determined by the centroid of the skin contour voxel in the concave region of the anterior abdominal wall. In specific implementation, the 3D mapping module extracts the surface mesh composed of skin voxels, calculates the curvature extrema of the surface mesh in the anterior abdominal wall region, locates the center of the umbilicus at the point of maximum negative curvature, and sets the coordinates of the umbilicus center as the origin of the 3D voxel mesh.

[0021] In the 3D voxel mesh, the 3D mapping module establishes a spherical polar coordinate system with the center of the umbilicus as the pole, and extracts a hemispherical region of interest centered at the pole with a radius equal to the preset abdominal radius. All voxel labels within this hemispherical region of interest are transformed from Cartesian coordinates to polar coordinates, with the three dimensions being polar radius, polar angle, and depth. The polar radius is the projected distance from the voxel to the center of the umbilicus on the horizontal plane; the polar angle is the circumferential angle of the voxel relative to the predetermined zero-degree direction in the horizontal plane; and the depth is the vertical distance of the voxel along the direction perpendicular to the tangent plane of the body surface. The 3D mapping module projects the label of each voxel along the three dimensions of polar radius, polar angle, and depth, generating a layered distance field for the abdominal wall tissue. At each combination of polar angle and depth, the layered distance field records the polar radius boundary values ​​encountered along this direction from the center of the umbilicus, representing the changes in voxel labels of various tissues, thus forming an explicit mapping of the tissue layer boundaries.

[0022] After the layered distance field is constructed, the 3D mapping module inputs the layered distance field into the nonlinear finite element deformer. The nonlinear finite element deformer uses a tetrahedral hyperelastic material model to simulate the nonlinear stress-strain relationship of the abdominal wall tissue under pneumoperitoneum pressure. In specific implementation, the nonlinear finite element deformer first generates a 3D tetrahedral mesh covering the abdominal wall region. The mesh nodes are aligned with the sampling points of the layered distance field. The size of the tetrahedral elements is adaptively subdivided according to the estimated distribution of local tissue stiffness, with the mesh size of the skin layer and fascia layer being smaller than that of the fat layer. Each tetrahedral element is assigned hyperelastic material parameters corresponding to the tissue type. The hyperelastic material model selected is the Mooney-Rivlin model, whose strain energy density function expression is: in, This represents the strain energy per unit volume. and These are the first and second principal invariants of the partial variables of the right Cauchy-Green deformed tensor, respectively. and For material constants, Bulk modulus Let be the determinant of the deformed gradient matrix. and The value is set according to the tissue type: skin tissue The value range is from 0.1 MPa to 0.3 MPa. The value ranges from 0.02 MPa to 0.1 MPa; muscle fascia tissue The value range is from 0.3 MPa to 0.8 MPa. The value ranges from 0.05 MPa to 0.2 MPa; adipose tissue The value range is from 0.01 MPa to 0.05 MPa. The value ranges from 0.002 MPa to 0.01 MPa. The value of is determined by the assumption of material incompressibility and is uniformly set to . 1000 times. The displacement field of the nodes was calculated by solving the nodal displacement field using the finite element method.

[0023] The boundary constraints of the nonlinear finite element deformer are set as follows: fixed displacement constraints are applied to the nodes on the patient's back surface, and a uniformly distributed pressure load is applied to the inner surface of the abdominal cavity to simulate pneumoperitoneum pressure. The pneumoperitoneum pressure value is set according to the pressure required for surgery, with a typical value of 12 mmHg to 15 mmHg. The nonlinear finite element deformer uses the Newton-Raphson iterative method to solve the nonlinear equilibrium equations. Iteration stops when the residual force norm of the nodes is less than a preset convergence threshold, obtaining the node displacement field after the abdominal bulge deformation. The deformed node displacement field is superimposed with the original tetrahedral mesh, and the spatial position of the tissue voxel labels is updated to form a dynamic deformation model of the patient's abdomen. This dynamic deformation model of the patient's abdomen can express the hierarchical deformation of abdominal wall tissues, the stretching and displacement of blood vessel orientation, and the movement of the relative position of the uterus under different pneumoperitoneum pressures.

[0024] See Figure 5 In the figure, the horizontal axis represents the circumferential angle of the umbilicus, ranging from 0° to 360°, and the vertical axis represents the extreme diameter boundary value of the tissue layer, in millimeters, indicating the thickness or boundary distance of each tissue layer at the corresponding circumferential angle. The curves correspond to the changing trends of the extreme diameter boundary values ​​of five tissue layers: skin, muscle fascia, blood vessels, fat, and peritoneum.

[0025] The black solid line represents the skin layer. The overall polar diameter boundary value is between 14mm and 22mm, showing a periodic fluctuation. It reaches a maximum value of about 22mm at about 230°. There is a clear trough in the polar diameter boundary value in the 90° to 150° region, which is about 15mm. This reflects the variation of skin layer thickness at different circumferential angles.

[0026] The black dashed line represents the fat layer, whose polar diameter boundary value fluctuates between 21 mm and 35 mm, showing a relatively obvious bimodal characteristic. The peak values ​​are distributed at circumferential angles close to 0° and 270°, respectively. The highest peak is about 35 mm, and the lowest trough appears at about 180°, about 22 mm, indicating that the thickness of the fat layer has a significant non-uniform distribution along the circumferential angle of the umbilicus.

[0027] The black dotted lines represent the muscle fascia layer. The extreme diameter boundary values ​​are between 28mm and 43mm. The curve shows a large fluctuation range, reaching a maximum peak of about 43mm between 150° and 210°, indicating that the fascia layer is thicker in this region. At other angles, it gradually decreases to about 28mm, showing the obvious variability in the thickness of the muscle fascia layer.

[0028] The black dotted lines represent the peritoneum. The extreme diameter boundary values ​​vary from 30 mm to 42 mm. The curve shows a fluctuating trend, with the peak value mainly located around 350°, approximately 42 mm, and the lowest trough appearing in the range of 150° to 210°, approximately 30 mm. This reflects that the morphology of the peritoneum has a certain regular distribution around the umbilicus.

[0029] The black dots and lines represent the vascular layer. The extreme diameter boundary value fluctuates between 30 mm and 44 mm, with multiple local peaks and valleys. In particular, there is a significant peak value of 44 mm near 150° to 210°. At other angles, the diameter shows a decreasing trend, with the lowest point being about 28 mm. This indicates that there is a relatively thick distribution of vascular tissue in the vascular layer in this angle region.

[0030] Overall, the five curves exhibit periodic variations across the circumferential angle from 0° to 360°, and the extreme diameter boundary values ​​of different tissue layers show a staggered distribution. The skin layer has the smallest thickness and the smallest variation, while the fat layer and muscle fascia layer have larger thicknesses and more significant fluctuations. The extreme diameter boundary values ​​of the vascular layer and peritoneum layer are relatively high, revealing a complex tissue hierarchy. This figure accurately reflects the thickness distribution of the abdominal wall tissue layers around the umbilicus obtained by the three-dimensional mapping module in Example 1 through multimodal image data fusion and voxel-level semantic segmentation. It is used to construct the tissue layer distance field in the dynamic deformation model of the patient's abdomen, demonstrating the spatial positioning and thickness differences of different tissue layers around the umbilicus, providing a precise anatomical information basis for subsequent laparoscopic approach planning.

[0031] Example 2: In specific implementation, please refer to Figure 3 The approach optimization module is activated after the dynamic deformation model of the patient's abdomen is constructed. The module first generates multiple preset puncture directions. These preset puncture directions are uniformly sampled within the hemispherical space of a 3D voxel mesh, with the umbilicus center as the vertex. Sampling employs a spiral scanning method based on the Fibonacci spherical distribution, generating N equally spaced angle intervals in the polar angle dimension and M elevation angle layers in the depth elevation angle dimension. N is set to 360°, and M to 180°, resulting in 64,800 preset puncture direction vectors. Each preset puncture direction vector originates from the umbilicus center and extends along a unit direction to a preset maximum puncture depth. The maximum puncture depth is automatically set to 1.5 times the maximum abdominal wall thickness within the model, based on the statistical values ​​of the abdominal wall thickness distribution in the dynamic deformation model of the patient's abdomen.

[0032] The approach optimization module controls the virtual single-port laparoscope to perform forward and backward simulations along each preset puncture direction. The virtual single-port laparoscope is represented by a cylindrical rigid body model with a diameter of 5 mm or 10 mm. A simplified geometric model of the instrument's end effector is attached to the front end of the cylindrical rigid body model. The forward and backward simulations are executed according to a preset step sequence, which consists of a forward entry step sequence and a backward step sequence. The forward entry step sequence starts from zero depth and increases the step value incrementally, with each step value set to 0.5 times the instrument diameter. After each step forward, the approach optimization module extracts the voxel tag corresponding to the current position of the virtual single-port laparoscope from the dynamic deformation model of the patient's abdomen. The voxel tag values ​​include skin voxels, fat voxels, muscle voxels, fascia voxels, blood vessel voxels, peritoneum voxels, and uterine boundary voxels.

[0033] When the extracted voxel label is a vascular voxel or a fascia voxel, the approach optimization module accumulates one collision count. The collision weight for vascular voxels is set to 1.5, and the collision weight for fascia voxels is set to 1.2. When the extracted voxel label is a fat voxel or a muscle voxel, the approach optimization module does not accumulate collision counts but records the penetration depth value corresponding to the current step length. The penetration depth value is defined as the Euclidean distance from the skin voxel surface to the current position along the preset puncture direction at the tip of the virtual single-port laparoscope. The approach optimization module continues to advance in the same preset puncture direction until it touches the uterine boundary voxel, the penetration depth exceeds the preset maximum puncture depth, or the number of collisions reaches the preset upper limit. At this point, it stops advancing and records all step length data in that preset puncture direction.

[0034] After all step lengths are completed, the approach optimization module performs a weighted summation of the collision count and penetration depth for all step lengths in the same preset puncture direction. The weighted summation formula is as follows: in, This indicates the initial resistance value in the preset puncture direction. This represents the total number of steps in the simulated forward and backward movement along the preset puncture direction. The step size index is used for indexing. For the first The cumulative number of collisions over a step length. For the first The penetration depth value recorded for each step size. To preset the maximum puncture depth, The weighting factor for the number of collisions. This is the penetration depth weighting coefficient. The value was set at 0.75, which was determined based on the regression analysis results of the correlation between the number of collisions and mechanical resistance through an isolated porcine abdominal wall puncture experiment. The value is 0.25, which is determined based on the biomechanical calibration results of the contribution of puncture depth to frictional resistance. This is a constant automatically set by the approach optimization module based on the statistical values ​​of the abdominal wall thickness distribution from the patient's abdominal dynamic deformation model.

[0035] After calculating the initial resistance value, the approach optimization module compares it with a preset resistance target. The preset resistance target is set to a zero-resistance state, i.e., the target value is set to 0. The approach optimization module calculates the difference between the initial resistance value and the preset resistance target. This difference is used as an error signal and backpropagated to the Young's modulus of the corresponding tissue voxel in the dynamic deformation model of the patient's abdomen. Backpropagation uses gradient descent, locating the corresponding voxel at each step along the preset puncture direction trajectory, calculating the partial derivative of the error signal with respect to the Young's modulus of the tissue voxel, and updating the Young's modulus value according to the learning rate parameter set to 0.001. After the update, the approach optimization module re-executes the approach-reverse simulation, recalculating the collision response and tissue deformation using the updated Young's modulus parameter, recalculating the initial resistance value, and entering an iterative loop. The iterative loop terminates when the relative change in the initial resistance value between two consecutive iterations is less than a convergence threshold set to 0.01. The approach optimization module stores the initial resistance value at convergence as the approach resistance potential energy value in the corresponding record for that preset puncture direction. The approach optimization module performs the above iterative advance and retreat simulation sequentially for all preset puncture directions to complete the calculation of the approach resistance potential energy value for all preset puncture directions.

[0036] During each simulated advance and retreat, the approach optimization module simultaneously records the closest distance between the end effector of the virtual single-port laparoscopy and the uterine boundary. The end effector position is represented by the center point of the ball head at the front end of the cylindrical rigid body model of the virtual single-port laparoscopy. The uterine boundary is represented by an isosurface composed of voxels of the uterine boundary in the dynamic deformation model of the patient's abdomen. The closest distance is obtained by calculating the shortest Euclidean distance from the center point of the ball head of the end effector to all triangular facets on the isosurface of the uterine boundary. After calculating the closest distance at each step position, the approach optimization module compares the closest distance with a preset safety distance. The preset safety distance is set to 10 mm, based on the safe operating distance specifications between instruments and the uterine serosal layer during single-port laparoscopic surgery.

[0037] When the closest distance is less than 10 mm of the preset safety distance, the approach optimization module adds a penalty potential energy term to the approach resistance potential energy value in that preset puncture direction. The penalty potential energy term is calculated as a potential energy increment function, which is exponential in form, with the natural constant e as the base and the exponent being the negative ratio of the closest distance to the preset safety distance multiplied by a penalty coefficient. The penalty coefficient is set to 5.0, based on calibration obtained through physician operation preferences on the surgical simulator. The approach optimization module marks the approach resistance potential energy value after adding the penalty potential energy term as the updated approach resistance potential energy value. The approach optimization module transmits the updated approach resistance potential energy value to the thermal guidance module, triggering the thermal guidance module to re-execute the surface interpolation process, incorporating the approach resistance potential energy value after adding the penalty potential energy term into the radial basis function interpolation calculation, thereby updating the resistance potential energy value of the corresponding angle region in the resistance potential energy heatmap. The updated resistance potential energy heatmap shows a higher potential energy region in the direction near the uterine boundary, guiding the operator to avoid risky approach directions where the instrument is too close to the uterus.

[0038] See Figure 6 In the figure, the horizontal axis represents the inbound resistance potential energy value, ranging from 0 to 25, and the vertical axis represents the probability density of the corresponding inbound resistance potential energy value. The solid curve in the figure is the probability density distribution of the initial resistance value before the addition of the penalty potential energy term by the inbound optimization module, and the dashed curve is the probability density distribution of the updated inbound resistance potential energy value after the addition of the penalty potential energy term.

[0039] The solid curve shows that the initial resistance values ​​are mainly concentrated in the low resistance range of 0 to 10, with the probability density peaking at around 3.5. This indicates that the initial resistance is relatively low in most of the preset puncture directions, making them suitable as potential entry paths. As the resistance potential energy value increases above 10, the probability density decreases rapidly, indicating that there are fewer directions with high resistance potential energy.

[0040] The dashed curve shows that the updated resistance potential energy distribution shifts significantly to the right in the high resistance range, and the peak probability density increases from about 3.5 to about 9 to 10. The probability density remains at a high level in the 10 to 20 range, indicating that after iterative optimization and the addition of the uterine boundary safety distance penalty, the resistance potential energy in some of the original low resistance directions has increased significantly.

[0041] The two curves show significant differences in the 0 to 5 range. The initial resistance value has a high probability density in this range, while the resistance density decreases significantly after the update. This reflects that the approach optimization module effectively improves the resistance potential energy of the puncture direction that is easy to approach the uterine boundary by iteratively updating the tissue stiffness parameters, thus playing a role in eliminating high-risk approach directions.

[0042] Overall, the trend of probability density distribution shown in the figure reflects that in Example 2, the approach optimization module uses the number of collisions and penetration depth to calculate the initial resistance value, and combines it with the safety distance penalty potential energy term to adjust tissue stiffness through iterative backpropagation. This achieves the optimization and updating of the resistance potential energy of all preset puncture directions, enhances the ability to avoid high-resistance areas of vascular fascia and high-risk paths close to the uterine boundary, and thus provides more accurate resistance potential energy data for the subsequent thermal guidance module.

[0043] Example 3: In specific implementation, please refer to Figure 4 The thermal guidance module receives the potential energy values ​​of the access resistance corresponding to all preset puncture directions from the access optimization module. The thermal guidance module discretizes the circumferential angle of the umbilicus into 360 angle intervals, each spanning one degree. The zero-degree direction of an angle interval is defined as the direction from the center of the umbilicus to the right anterior superior iliac spine within the transverse section of the patient's abdomen. The puncture depth is discretized into depth intervals, starting from zero depth at the skin surface to the preset maximum puncture depth, divided with a fixed step size of 0.5 times the diameter of the virtual single-port laparoscope. The combination of each angle interval and each depth interval defines a candidate access grid point in three-dimensional space; the total number of candidate access grid points is equal to the product of the total number of angle intervals and the total number of depth intervals.

[0044] The thermal guidance module uses the entry drag potential energy value at each candidate entry grid point as the height value of that grid point. For candidate entry grid points not directly hit by the preset puncture direction, the thermal guidance module performs an initial assignment by querying the entry drag potential energy value in the nearest preset puncture direction and applying an inverse distance-weighted average. The thermal guidance module then uses radial basis functions to smoothly interpolate the height values ​​of all candidate entry grid points, generating a continuous drag potential energy surface. The expression for radial basis function interpolation is: In the above formula, Represents the angular coordinates of the drag potential energy surface in the circumferential direction of the umbilicus. and puncture depth coordinates The interpolated drag potential energy value at the location; This represents the total number of candidate ingress grid points; The index numbers of the candidate ingress grid points are incremented from 1 to 1. ; For the first The radial basis function weights corresponding to each candidate ingress grid point The coefficient matrix of the linear equation system is obtained by constructing and solving a system of linear equations constrained by the known height values ​​of all candidate ingress grid points. The coefficient matrix of the linear equation system is composed of the values ​​of the radial basis functions between any two candidate ingress grid points. For radial basis functions, multiple quadratic functions are chosen, with the form: , The distance variable is the input to the radial basis function. For smoothing parameters, The value is 0.1, which is a constant set to ensure that the interpolation surface remains second-order continuous and differentiable when passing through candidate ingress grid points without generating excessive fluctuations. For the first The circumferential angular coordinates of the umbilicus of each candidate ingress grid point; For the first The puncture depth coordinates of the candidate approach grid points. The right-hand side of the above linear equation system is the th... The known ingress resistance potential energy values ​​at each candidate ingress grid point are used in the solution process, which employs the LU decomposition method to output all weights. Substituting the radial basis function interpolation expression, a continuous drag potential energy surface is constructed.

[0045] The thermal guidance module projects the drag potential energy surface layer by layer according to depth values. At each discrete depth layer, the drag potential energy surface corresponds to a drag potential energy value curve that varies angularly along the circumferential angle of the umbilicus. The thermal guidance module projects the drag potential energy value curves of all depth layers onto a two-dimensional plane and uses a color mapping table to encode different drag potential energy values ​​with different colors. The color mapping table uses a color gradient from blue to red, with blue corresponding to low drag potential energy values ​​and red corresponding to high drag potential energy values. The thermal guidance module stores and outputs the two-dimensional color distribution image as a drag potential energy heatmap.

[0046] The thermal imaging module acquires real-time images of the patient's abdominal surface using a visible light camera positioned in the operating room, directly facing the patient's disinfected and prepared abdominal area. The module then uses a pre-trained umbilical detection network to process the images, outputting the two-dimensional coordinates of the umbilical markers in the image's pixel coordinate system. The umbilical detection network employs a convolutional neural network structure with a MobileNetV2 backbone and a feature pyramid network for multi-scale feature fusion, outputting a heatmap of key umbilical points. The coordinates of the umbilical markers are obtained by taking the local maxima of the output heatmap. Simultaneously, the module uses a contour segmentation network to acquire abdominal contour edge features. This network uses a U-Net architecture and outputs a binary mask of the abdominal region. Canny edges are extracted from this mask to obtain a set of abdominal contour edge feature points.

[0047] The thermal guidance module rigidly registers the coordinates of the umbilicus center in the drag potential energy heatmap with the umbilicus marker in the patient's abdominal surface image. Rigid registration calculates the registration matrix based on the correspondence between the three anatomical landmarks—the umbilicus center, the right anterior superior iliac spine, and the left anterior superior iliac spine—in the two images. In the drag potential energy heatmap, the positions of the right and left anterior superior iliac spines are obtained by mapping the bony markers of the anterior superior iliac spines, calibrated in the preoperative multimodal abdominal imaging data, to the heatmap coordinate system. In the patient's abdominal surface image, the positions of the right and left anterior superior iliac spines are confirmed by manual selection by the operator on the image or automatically identified by a pre-trained bony landmark detection network. The registration matrix for rigid registration is a two-dimensional similarity transformation matrix, containing rotation angles, translation vectors, and a uniform scaling factor. The thermal guidance module solves the following least-squares problem to obtain the registration matrix: minimizing the sum of the squared Euclidean distances between the three anatomical landmarks in the heatmap coordinate system and their corresponding anatomical landmarks in the patient's abdominal surface image. After the transformation, the thermal guidance module aligns the coordinate system of the drag potential energy thermal map to the coordinate system of the patient's abdominal surface image.

[0048] After rigid registration, the thermal guidance module performs elastic deformation correction on the registered drag potential energy heatmap based on the abdominal contour edge features. Elastic deformation correction employs a moving least squares deformation method, using abdominal contour edge feature points as control points to deform the boundary of the drag potential energy heatmap until it coincides with the abdominal contour boundary of the patient's abdominal surface image. The correspondence between control points is established by using a non-rigid iterative nearest-point algorithm between the abdominal contour edges in the drag potential energy heatmap and the abdominal contour edges in the patient's abdominal surface image. The thermal guidance module calculates the displacement vector for each control point pair, uses the moving least squares method to interpolate across the entire image plane to generate a dense deformation field, and applies this dense deformation field to the registered drag potential energy heatmap to complete the elastic deformation correction.

[0049] The thermal guidance module converts the corrected drag potential energy heatmap into a semi-transparent texture, with the alpha channel value of the semi-transparent texture uniformly set to 0.6. The thermal guidance module displays the semi-transparent texture on the real-time acquired abdominal surface image of the patient in a pseudo-color overlay, allowing the operator to simultaneously observe the patient's actual abdominal wall anatomical landmarks and the access drag potential energy distribution displayed through color coding.

[0050] Example 4: In practice, the trajectory generation module receives a low-resistance approach seed point selected by the operator on the resistance potential energy heatmap. This seed point is selected by the operator via a human-computer interaction device by clicking on the overlaid image of the patient's abdomen. The click location is mapped to the corresponding pixel coordinates on the resistance potential energy heatmap. The module then uses the color coding of the heatmap to look up the corresponding circumferential angle value of the umbilicus and the puncture depth value. The trajectory generation module extracts the circumferential angle value of the umbilicus corresponding to the low-resistance approach seed point as the starting angle parameter and the puncture depth value as the starting depth parameter.

[0051] The trajectory generation module uses the angle and depth corresponding to the low-resistivity approach seed point as the starting node and the surgical target location as the ending node. The surgical target location is determined by the three-dimensional coordinates of the centroid of the uterine lesion or the surgical operation target point marked in the preoperative multimodal abdominal imaging data. These three-dimensional coordinates have been mapped to the voxel coordinate system of the patient's abdominal dynamic deformation model through coordinate transformation. The trajectory generation module establishes a three-dimensional mesh graph from the starting node to the ending node in the patient's abdominal dynamic deformation model. The node set of the three-dimensional mesh graph consists of all voxel center points located within the hemispherical region of interest in the patient's abdominal dynamic deformation model. The starting node is located on the voxel center coordinates with the umbilicus center as the center, the starting depth parameter as the radius, and the starting angle parameter as the direction. The ending node is located on the voxel whose voxel center coordinates are closest to the surgical target location. The edge set of the three-dimensional mesh graph connects each node to its twenty-six adjacent voxel nodes in three-dimensional space. The adjacency relationship is based on the three-dimensional six-connectivity plus eighteen diagonal connectivity definition of the voxel mesh.

[0052] In the 3D mesh graph, the trajectory generation module calculates the inbound drag potential energy value at each mesh node. For mesh nodes already covered by candidate inbound mesh points, the inbound drag potential energy value is directly sampled from the drag potential energy surface. For mesh nodes located in the gaps between candidate inbound mesh points, the inbound drag potential energy value is obtained by trilinear interpolation of the drag potential energy surface. The trajectory generation module uses the product of the inbound drag potential energy value at each mesh node and the spatial distance to adjacent nodes as the state transition cost. The formula for calculating the state transition cost is: In the above formula, Indicates from grid node Transfer to adjacent grid nodes The cost of state transition; The current grid node; For the target's adjacent grid nodes; Indicates the target's adjacent grid nodes The ingress drag potential energy value is obtained by drag potential energy surface sampling or trilinear interpolation. Represents grid nodes Three-dimensional spatial coordinate vector in the voxel coordinate system of the dynamic deformation model of the patient's abdomen; Represents grid nodes Three-dimensional spatial coordinate vector in the voxel coordinate system of the dynamic deformation model of the patient's abdomen; This represents the Euclidean distance between two grid nodes, i.e., the spatial distance. The range of values ​​for is real numbers greater than zero, when The value is higher when the location is in an area with a high concentration of vascular or fascial voxels. The value is lower when the location is in a region where fat or muscle cells are evenly distributed.

[0053] The trajectory generation module employs a dynamic programming algorithm to search backwards from the termination node. The dynamic programming algorithm maintains a cumulative state transition cost table, which records the minimum cumulative cost for each grid node in the 3D mesh graph to reach the termination node. During algorithm initialization, the cumulative cost of the termination node is set to its own ingress resistance potential energy value, and the cumulative costs of all other grid nodes are set to positive infinity. The trajectory generation module uses a priority queue for breadth-first cost propagation. Each time, the grid node with the minimum cumulative cost is taken from the priority queue and expanded. For all its adjacent grid nodes, candidate cumulative costs are calculated for reaching the termination node from the adjacent grid nodes through the current grid node. These candidate cumulative costs consist of the ingress resistance potential energy value of the adjacent grid nodes, the cumulative cost of the current grid node, and the state transition cost. When a candidate cumulative cost is less than the existing cumulative cost of an adjacent grid node, the cumulative cost of that adjacent grid node is updated, and its predecessor node is recorded as the current grid node. Simultaneously, the adjacent grid node is added to the priority queue. This cost propagation process continues until the cumulative cost of the starting node is determined or the priority queue is empty.

[0054] The trajectory generation module starts from the starting node and backtracks sequentially along the recorded predecessor node relationships towards the ending node. During the backtracking process, the traversed grid nodes are arranged into a node sequence. If the predecessor node record of the starting node is empty, the trajectory generation module selects the neighboring grid node with the minimum cumulative cost around the starting node as the first backtracking node. The node sequence obtained through backtracking is the minimum energy consumption path. The trajectory generation module labels the corresponding puncture depth for each grid node on the minimum energy consumption path. The puncture depth value is defined as the shortest distance from the grid node along the radial direction from the center of the umbilicus to the voxel on the skin surface.

[0055] The trajectory generation module obtains the spatial coordinates of each grid node on the minimum energy consumption path and the corresponding tissue voxel type. The tissue voxel type is directly retrieved from the voxel label map of the patient's abdominal dynamic deformation model. Each grid node records tissue voxel types including fat voxels, muscle voxels, fascia voxels, vascular voxels, peritoneal voxels, and uterine boundary voxels. The trajectory generation module calculates the spatial vectors of adjacent grid nodes according to their order on the minimum energy consumption path. For the i-th... The spatial vector of the nth grid node is defined as starting from the nth grid node. The spatial coordinates of the nth grid node point to the first The spatial coordinates of the nth grid node are represented by a three-dimensional direction vector. The trajectory generation module will then... The spatial vector of the nth grid node and the nth The angle between the spatial vectors of the nth grid node is used as the angle between the spatial vectors of the nth grid node. The approach direction deflection angle of each grid node is calculated by dividing the dot product of two 3D direction vectors by the product of their magnitudes and then taking the inverse cosine. The value of the approach direction deflection angle ranges from 0 to 180 degrees. The trajectory generation module arranges the approach direction deflection angles of all grid nodes in order of their position on the path of minimum energy consumption, generating a direction sequence of single-port laparoscopic approach trajectories with depth markings. The approach direction deflection angles of the first and last grid nodes in the direction sequence are marked as zero.

[0056] Example 5: In practice, the instrument path preview module obtains the single-port laparoscopic approach trajectory with depth markers from the trajectory generation module. The single-port laparoscopic approach trajectory consists of sequentially arranged grid nodes, each carrying three-dimensional spatial coordinates, puncture depth value, and tissue voxel type.

[0057] The instrument path preview module accesses the dynamic deformation model of the patient's abdomen to obtain the abdominal wall thickness corresponding to each voxel position in the model. Abdominal wall thickness is defined as the maximum distance from the skin surface voxel along a direction perpendicular to the body surface to the peritoneal voxel. The abdominal wall thickness field is generated during the construction of the dynamic deformation model of the patient's abdomen, and the instrument path preview module directly reads the value of this abdominal wall thickness field at each grid node position on the single-port laparoscopic approach trajectory.

[0058] The instrument path preview module compares the puncture depth value of each grid node on the single-port laparoscopic approach trajectory with the corresponding abdominal wall thickness. When the puncture depth value of a grid node is greater than the abdominal wall thickness at the corresponding location, the instrument path preview module marks that grid node as a perforation risk node. A perforation risk node indicates that the puncture depth of the virtual single-port laparoscopy at this location exceeds the safe thickness of the abdominal wall tissue, posing a risk of peritoneal penetration or damage to intra-abdominal organs.

[0059] The device path preview module highlights the perforation risk nodes on the drag potential energy heatmap. The highlighting uses a warning color, with a high-brightness yellow color value of RGB(255,255,0). Based on the color mapping table currently overlaid on the drag potential energy heatmap, the module overlays the warning color onto the corresponding heatmap pixels of the perforation risk nodes using alpha blending. The alpha value of the warning color layer is set to 0.9, resulting in a prominent yellow mark on the heatmap for the perforation risk node area.

[0060] In the dynamic deformation model of the patient's abdomen, the instrument path preview module generates a virtual safety bounding box along the single-port laparoscopic approach trajectory. The virtual safety bounding box is generated as follows: for each mesh node on the single-port laparoscopic approach trajectory, a cuboid bounding box is constructed centered on the mesh node. The major axis of the cuboid bounding box is aligned with the tangent direction of the approach trajectory at that mesh node. The side length of the cuboid bounding box is set to twice the diameter of the virtual single-port laparoscopy, and the length of the cuboid is set to half the distance between adjacent mesh nodes. The instrument path preview module merges the cuboid bounding boxes of all mesh nodes into a connected volume rendering model, rendered using a semi-transparent red material with an alpha value set to 0.3, forming a virtual safety bounding box along the approach trajectory, used to indicate the safe range of the instrument path in the model's 3D view.

[0061] The instrument path preview module receives the operator's corrected depth input for perforation risk nodes. This corrected depth input is achieved through a human-computer interface. The operator clicks on the selected perforation risk node on the resistance potential energy heat map or the 3D model view, and a depth value input box pops up, allowing the operator to manually enter the corrected depth value. The allowed range for the corrected depth value is a real number between zero skin surface depth and the corresponding abdominal wall thickness. The instrument path preview module replaces the original depth value of the selected perforation risk node with the corrected depth input. Only the depth value of the perforation risk node modified by the operator is replaced; the depth values ​​of other grid nodes on the single-port laparoscopic approach trajectory remain unchanged.

[0062] The instrument path preview module sets the mesh node with the replaced depth value as the new termination node. If the operator modifies multiple perforation risk nodes, the instrument path preview module selects the modified node that is furthest from the starting node on the approach trajectory as the new termination node. The instrument path preview module passes the spatial coordinates of the new termination node to the trajectory generation module, triggering the trajectory generation module to re-execute the dynamic programming search process in the patient's abdominal dynamic deformation model.

[0063] After the trajectory generation module is triggered, it keeps the starting node unchanged and recalculates the minimum energy consumption path from the starting node to the new ending node in the 3D mesh of the dynamic deformation model of the patient's abdomen, using the new ending node as the endpoint. The state transition cost definition used in the dynamic programming search is the same as in the original search, still using the product of the approach resistance potential energy value at each mesh node and the spatial distance between adjacent nodes. The trajectory generation module outputs the newly calculated minimum energy consumption path and labels the depth values ​​of all mesh nodes on the path, forming the corrected single-port laparoscopic approach trajectory. In the corrected single-port laparoscopic approach trajectory, the locations of perforation risk nodes are bypassed or replanned as paths with a depth not exceeding the abdominal wall thickness, thus meeting the safety operation requirements.

[0064] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A visually guided single-port laparoscopic approach planning system for gynecology, characterized in that, include: The three-dimensional mapping module acquires multimodal abdominal image data of gynecological patients before surgery, maps the multimodal abdominal image data to a three-dimensional voxel mesh centered on the umbilicus, and constructs a dynamic deformation model of the patient's abdomen that includes abdominal wall layers, blood vessel orientation, and the relative position of the uterus. The approach optimization module drives the virtual single-port laparoscope to perform advance and retreat simulations along multiple preset puncture directions in the dynamic deformation model of the patient's abdomen. It records the number of collisions and the penetration depth between the virtual instrument and the abdominal wall tissue voxel in each advance and retreat simulation. The number of collisions and the penetration depth are then propagated back to the tissue stiffness field of the dynamic deformation model of the patient's abdomen, and the approach resistance potential energy value corresponding to each preset puncture direction is iteratively updated. The thermal guidance module performs surface interpolation on the entry resistance potential energy value according to the circumferential angle of the umbilicus and the puncture depth to generate a resistance potential energy thermal map covering all possible entry directions around the umbilicus, and superimposes the resistance potential energy thermal map on the real-time acquired abdominal surface image of the patient. The trajectory generation module receives the low-resistance entry seed point selected by the operator on the resistance potential energy heat map, uses the angle and depth corresponding to the low-resistance entry seed point as the initial entry parameters, and uses dynamic programming to search for the minimum energy consumption path from the umbilicus to the surgical target in the dynamic deformation model of the patient's abdomen, and outputs a single-port laparoscopic entry trajectory with depth marking.

2. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The three-dimensional mapping module is specifically used for: Obtain voxel labels for skin contours, muscle fascia, peritoneum, and uterine boundaries contained in the multimodal abdominal image data; A polar coordinate system is established in the three-dimensional voxel grid according to the geometric center of the umbilicus. The voxel labels are projected onto the three dimensions of polar diameter, polar angle and depth to form a layered distance field of abdominal wall tissue. The layered distance field is input into a nonlinear finite element deformer. By adjusting the boundary constraints of the nonlinear finite element deformer, the abdominal bulging morphology under different pneumoperitoneum pressures is simulated to obtain the dynamic deformation model of the patient's abdomen.

3. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 2, characterized in that, The nonlinear finite element deformer uses a tetrahedral hyperelastic material model to simulate the nonlinear stress-strain relationship of abdominal wall tissue under pneumoperitoneum pressure.

4. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The ingress optimization module is specifically used for: In each of the preset puncture directions, the virtual single-port laparoscope is controlled to enter the dynamic deformation model of the patient's abdomen step by step according to the preset step length sequence. Each time it advances by one step, the voxel label at the current position is queried. If the voxel label is a blood vessel or fascia tissue, the number of collisions is accumulated once. If the voxel label is fat or muscle tissue, the penetration depth of the current step length is recorded. The initial resistance value for the puncture direction is obtained by weighting and summing the number of collisions and the penetration depth for all step lengths in the same puncture direction. Based on the difference between the initial resistance value and the preset resistance target, the Young's modulus of the corresponding tissue voxel in the dynamic deformation model of the patient's abdomen is adjusted in reverse, and the forward and backward simulation is repeated until the initial resistance value converges. The converged initial resistance value is then used as the approach resistance potential energy value.

5. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The thermal guiding module is specifically used for: The circumferential angle of the umbilicus is discretized into 360 angle intervals, and the puncture depth is discretized into depth intervals. Each combination of angle interval and depth interval corresponds to a candidate access grid point. The ingress drag potential energy value at each candidate ingress grid point is used as the height value of that grid point. Radial basis functions are used to smoothly interpolate the height values ​​of all grid points to generate a continuous drag potential energy surface. The drag potential energy surface is projected onto a two-dimensional plane in layers according to depth values, and the drag potential energy of each layer is encoded with different colors to form the drag potential energy heat map.

6. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The thermal guiding module is also used for: Acquire umbilical markers and abdominal contour features from real-time acquired images of the patient's abdominal surface; The coordinates of the umbilicus center in the resistance potential energy heat map are rigidly registered with the umbilicus marker in the patient's abdominal surface image. Based on the abdominal contour edge features, the registered resistance potential energy heat map is elastically deformed to make the boundary of the resistance potential energy heat map coincide with the boundary of the patient's abdominal surface image, and the corrected resistance potential energy heat map is superimposed and displayed in the form of a semi-transparent texture.

7. A visually guided single-port laparoscopic approach planning system for gynecology according to claim 6, characterized in that, The rigid registration is based on the correspondence between the three anatomical landmarks—the center of the umbilicus and the bilateral anterior superior iliac spines—in the two images, which is used to calculate the registration matrix.

8. The visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The trajectory generation module is specifically used for: Using the angle and depth corresponding to the low-resistance approach seed point as the starting node and the surgical target point position as the ending node, a three-dimensional mesh diagram from the starting node to the ending node is established in the dynamic deformation model of the patient's abdomen. In the three-dimensional mesh diagram, the product of the ingress resistance potential energy value at each mesh node and the spatial distance between adjacent nodes is used as the state transition cost; A dynamic programming algorithm is used to search backwards from the termination node, recording the previous node that minimizes the cumulative state transition cost, until the starting node is reached. The sequence of nodes obtained by backtracking is taken as the minimum energy consumption path, and the corresponding puncture depth is marked for each node.

9. A visually guided single-port laparoscopic approach planning system for gynecology according to claim 8, characterized in that, The trajectory generation module is also used for: Obtain the spatial coordinates of each node on the path of minimum energy consumption and the tissue voxel type corresponding to the node; Calculate the spatial vectors of adjacent nodes in the order of nodes, and take the angle between the spatial vector of each node and the spatial vector of the previous node as the inbound direction deflection angle of that node. Arrange the approach direction deflection angles of all nodes in the node order to generate the direction sequence of the single-port laparoscopic approach trajectory with depth marking.

10. A visually guided single-port laparoscopic approach planning system for gynecology according to claim 1, characterized in that, The ingress optimization module is also used for: In each simulation of advance and retreat, the closest distance between the end effector of the virtual single-port laparoscope and the boundary of the uterus is recorded; When the nearest distance is less than the preset safety distance, a penalty potential energy term is added to the approach resistance potential energy value in the puncture direction; The ingress resistance potential energy value after adding the penalty potential energy term is re-interpolated using surface interpolation to update the resistance potential energy value of the corresponding angular region in the resistance potential energy heatmap.

Images