Cardiac intervention surgical robot three-dimensional path planning method and computer program product

Abstract

The invention discloses a three-dimensional path planning method for a cardiac interventional surgical robot and a computer program product, and the method comprises the steps: dynamically adjusting and updating a three-dimensional model in real time according to real-time environment detection and image information in an operation; calculating an optimal three-dimensional path of the surgical robot through an improved genetic algorithm according to the three-dimensional model and the real-time position of the surgical robot; a fitness function F preset by the genetic algorithm considers the path length, the obstacle avoidance capability and the operation safety, and in different stages of the operation, different optimization objectives are emphasized by adjusting values of consideration factor weights. According to the method, the optimal three-dimensional path of the surgical robot is calculated through the genetic algorithm considering the path length, the obstacle avoidance capacity and the surgical safety, intelligent and precise planning of the surgical path is achieved, the method has the intraoperative real-time feedback and dynamic adjustment capacity, and the surgical accuracy and safety are improved.

Description

Technical Field

[0001] The present invention relates to the field of surgical robot-assisted technology, and more specifically, to a three-dimensional path planning method and computer program product for a cardiac interventional surgery robot. Background Art

[0002] As an important branch of modern medical technology, heart valve interventional surgery aims to repair or replace heart valves in a minimally invasive manner, reduce surgical trauma, and improve the quality of life of patients. The heart valve area has a complex anatomical structure, including multiple valves and their surrounding blood vessels, myocardium, pericardium and other tissues. These structures are intertwined to form a highly sophisticated and dynamically changing environment. Therefore, in order to improve surgical accuracy and reduce surgical risks, it is of great significance to find an optimal surgical path in the three-dimensional space of the heart valve for heart valve interventional surgery.

[0003] The path planning of traditional heart valve interventional surgery mainly relies on the doctor's experience and medical imaging data. These methods can guide the selection of surgical paths to a certain extent, but they have obvious limitations: 1. It is highly dependent on the doctor's judgment, has a certain degree of subjectivity and uncertainty, and has high labor costs; 2. Due to the high requirements of cardiac surgery for safety and avoidance of postoperative complications, there is still a lack of appropriate algorithm models to implement more operational path planning; 3. The current path planning algorithm lacks intelligence and adaptability, and cannot be flexibly adjusted and optimized according to the patient's specific situation and surgical needs.

[0004] Therefore, developing a more efficient, safe and adaptable three-dimensional path planning method for cardiac interventional surgery robots is a technical problem that needs to be solved urgently. Summary of the invention

[0005] In response to the above-mentioned defects of the prior art, the present invention provides a three-dimensional path planning method and computer program product for a cardiac interventional surgical robot. The optimal three-dimensional path of the surgical robot is calculated by a genetic algorithm that takes into account the path length, obstacle avoidance capability, and surgical safety, thereby realizing intelligent and precise planning of the surgical path. It also has the ability of real-time feedback and dynamic adjustment during the operation, thereby improving the accuracy and safety of the operation.

[0006] To achieve the above object, on the one hand, the present invention provides a three-dimensional path planning method for a cardiac interventional surgery robot, characterized in that it comprises the following steps:

[0007] Read the three-dimensional structural medical image of the patient's heart and construct a three-dimensional model of the patient's heart based on it; dynamically adjust and update the three-dimensional model in real time according to the real-time environmental detection and image information during the operation;

[0008] According to the three-dimensional model and the real-time position of the surgical robot, the optimal three-dimensional path of the surgical robot is calculated by an improved genetic algorithm; the fitness function F preset by the genetic algorithm takes into account the path length F L , obstacle avoidance capability F o 、Surgery safety S , expressed as:

[0009] F=w L ·F L +w O ·F O +w S ·F S ;

[0010] The weight w L +w O +w S =1,

[0011] Where x i ,y i Indicates the horizontal coordinate, z i is the height coordinate, d i For each point p on the path i The distance to the nearest obstacle, ε is a small positive number; R(P) is the output of the risk assessment model based on machine learning;

[0012] At different stages of the operation, by adjusting w L 、w O and w S to emphasize different optimization objectives.

[0013] When designing the fitness function of the genetic algorithm in the method, multiple key factors of the surgical path are fully considered, such as path length, obstacle avoidance ability, and surgical safety. The obstacle avoidance ability is evaluated by calculating the distance from each point on the path to the nearest obstacle; the surgical safety is quantified by combining the machine learning risk assessment model R(P). The weights of each factor are dynamically adjusted according to the surgical stage to ensure that the planned path is both physically feasible and surgically safe. In this way, path planning not only considers physical feasibility, but also takes into account the safety and success rate of the operation.

[0014] Furthermore, due to F L 、F O 、F S The consideration criteria of each function are very different. In order to make each function have the same sensitivity to the overall fitness function, the path length F L , obstacle avoidance capability F o 、Surgery safety SStandardization is performed, and the standardization formula is:

[0015]

[0016] Where P i Indicates path; F k (k=L,O,S) is standardized to The fitness function is:

[0017]

[0018] Furthermore, constructing a three-dimensional model of the patient's heart includes the following sub-steps:

[0019] Step S1.1, using median filtering, bilateral filtering and Gaussian filtering to process the image in order to effectively remove noise in the three-dimensional structure medical image of the patient's heart and improve the image quality;

[0020] Step S1.2, using a deep learning model to segment the image processed in step S1.1, identifying and marking heart valves, blood vessels and tissue structures;

[0021] Step S1.3, reconstruct the structures including heart valves and blood vessels in the image in three dimensions, and use the registration technology to fuse the heart valve segmentation module, blood vessel feature module and robot segmentation module to obtain a three-dimensional model that integrates multiple information.

[0022] Furthermore, the deep model of step S1.2 is a deep model based on U-Net, which has an encoder-decoder structure, can fuse bottom-level and high-level features, is suitable for medical image segmentation, and improves segmentation accuracy. On the basis of U-Net, other CNN (Convolutional Neural Network) structures (such as VGG (Visual Geometry Group), ResNet (Residual Network), etc.) can also be combined to enhance feature extraction capabilities.

[0023] Furthermore, when performing three-dimensional reconstruction in step S1.3, the three-dimensional model is finely processed using Taubin smoothing and detail enhancement.

[0024] Furthermore, calculating the optimal three-dimensional path of the surgical robot by genetic algorithm includes the following sub-steps:

[0025] Step S2.1, initialize the population, randomly generate a certain number of individuals (i.e. potential paths) as the initial population; determine the solution space, each individual consists of a series of genes (path point coordinates), representing a path from the starting position to the target;

[0026] Step S2.2, calculating the fitness of each individual according to the preset fitness function;

[0027] Step S2.3, select individuals with high fitness as parents;

[0028] Step S2.4: Randomly select two parent individuals from the selected individuals, perform a crossover operation, generate new offspring individuals, and form a new population;

[0029] Step S2.5, determine whether the new population reaches the termination condition (such as reaching a preset number of iterations or finding a solution that meets the characteristic conditions); if so, end the algorithm and output the optimal three-dimensional path; if not, repeat steps 2.2 to 2.4.

[0030] Furthermore, in step S2.3, the roulette method is used to select the parent: first, the probability Pi of each individual being inherited to the next generation population is calculated; then the cumulative probability qi of each individual is calculated, which represents the sum of the selection probabilities of the previous i individuals; then a uniformly distributed pseudo-random number r is generated in the interval [0,1], and this random number is used to simulate the pointer position in the roulette; the cumulative probability array q is traversed, and the first k value that satisfies q[k-1]<r≤q[k] is found, and the kth individual is selected as the parent; in order to avoid the loss of excellent individuals, an elite retention strategy is introduced in the selection process to retain individuals with high fitness directly to the next generation.

[0031] Furthermore, when selecting the parent generation for crossover in step S2.4, a multi-point crossover method is used. This method can more flexibly combine the path features of different parent generation individuals.

[0032] Furthermore, the steps for implementing the risk assessment model based on machine learning to predict possible complications or risks under different pathways are as follows:

[0033] Extracting features related to the safety of the surgical path from the three-dimensional model; the features include the distance from each point on the path to key anatomical structures (such as blood vessel walls, heart valves, etc.), the curvature of the path, whether the path passes through high-risk areas (such as valve areas with severe calcification), and the relative position of the path to areas where known complications occur;

[0034] Collect historical surgical data, including successful and failed surgical cases, and the safety assessment results of the surgical path in each case as label data;

[0035] The extracted features and label data are used to train any supervised learning model including logistic regression, random forest, gradient boosted tree (GBDT), and neural network. After optimizing the parameters of the supervised learning model, a risk assessment model based on machine learning is obtained, and its output is R(P).

[0036] On the other hand, the present invention provides a computer program product, characterized in that when the computer program product is run on a computer or device, the computer or device executes the three-dimensional path planning method for the cardiac interventional surgery robot as described above.

[0037] Compared with the prior art, the present invention has the following technical effects:

[0038] (1) When designing the fitness function, the present invention fully considers multiple key factors of the surgical path, such as path length, obstacle avoidance ability, and surgical safety. By dynamically adjusting the weights of each factor according to the specific situation, the present invention can flexibly respond to the needs of different surgical scenarios. In addition, the introduction of a risk assessment model based on machine learning further improves the safety and accuracy of path planning, ensuring that the surgical path is both efficient and safe.

[0039] (2) The present invention adopts a genetic algorithm to perform path planning for a cardiac interventional surgery robot. This method simulates the natural evolution process, automatically searches for and optimizes the path, and makes path selection more objective and accurate. It significantly improves the calculation speed and efficiency of path search, effectively reduces human errors, and greatly saves labor costs and time.

[0040] (3) When constructing a three-dimensional heart model, the present invention integrates information such as the heart valve segmentation module and the blood vessel feature module. This comprehensive information fusion strategy makes the constructed three-dimensional model more precise and accurate, and can more realistically reflect the actual structure of the patient's heart, providing a solid foundation for subsequent path planning.

[0041] (4) The present invention closely combines path planning with real-time environmental detection and image information, achieving real-time feedback and dynamic adjustment during surgery. During surgery, the system can dynamically update the three-dimensional model based on real-time detected environmental changes (such as vascular deformation, instrument deviation, etc.), and optimize the surgical path accordingly. This real-time feedback and dynamic adjustment mechanism ensures the accuracy and safety of the surgical process and reduces surgical risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a path planning method for a cardiac interventional surgery robot according to an embodiment of the present invention;

[0043] Figure 2 A flowchart of a genetic algorithm in one embodiment of the present invention;

[0044] Figure 3 Schematic diagram of the solution space of the genetic algorithm in one embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0046] In the following detailed description, many specific details are set forth to provide a more thorough understanding of the present invention. However, it is apparent to those skilled in the art that well-known algorithms and models (such as U-Net deep model, VGG model, ResNet model, Taubin smoothing algorithm, supervised learning model) are not shown in detail to avoid obscuring the subject matter of the present invention.

[0047] In addition, the order of execution of actions, steps, etc. in the devices and methods shown in the claims, specifications and drawings can be implemented in any order as long as there is no special explicit limitation on the order and the output of the previous processing is not used in the subsequent processing.

[0048] Example

[0049] This embodiment first provides a three-dimensional path planning method for a cardiac interventional surgery robot. Figure 1 , including the following steps:

[0050] Step S1: read a three-dimensional medical image of a patient's heart, and process and construct a three-dimensional model of the patient's heart.

[0051] More specifically, step S1 includes the following sub-steps:

[0052] Step S1.1, use filtering technology to remove noise in the image: first perform median filtering, for each pixel in the image, take a neighborhood (such as 3x3, 5x5, etc.) with it as the center, sort the pixel values ​​in the neighborhood and take the median as the new value of the pixel, the purpose is to remove salt and pepper noise in the image.

[0053] Then perform bilateral filtering on the image after median filtering, taking into account the spatial proximity of pixels and the similarity of pixel values, and taking a weighted average of the two. The formula is:

[0054]

[0055] Among them, g(x,y) is the output pixel value, f(k,l) is the pixel value in the neighborhood S, and w(i,j,k,l) ​​is the weight, which depends on the spatial proximity d(i,j,k,l) ​​and the pixel value similarity r(i,j,k,l):

[0056]

[0057] Finally, Gaussian filtering is performed to perform convolution on the image, and the convolution kernel is a Gaussian function. The calculation formula is as follows:

[0058]

[0059] Among them, G(x,y) is the Gaussian kernel and σ is the standard deviation, which determines the smoothness of the filter.

[0060] Step S1.2, using a deep learning model to segment the image processed in step S1.1, identify and mark the heart valve, blood vessels and tissue structures. The model used is a U-Net model, which excels in terms of small data volume requirements and fast training speed. In addition, this embodiment introduces convolution blocks of VGG or ResNet to enhance feature extraction capabilities by embedding these structural blocks in the encoder of U-Net.

[0061] Step S1.3, the heart valve, blood vessel and other structures of the image are reconstructed in three dimensions using volume rendering (Marching Cubes algorithm), and the images of different modalities (such as heart valve segmentation module, blood vessel feature module and robot segmentation module) are aligned to the same coordinate system using image registration technology, and then the segmentation results of each modality are fused. The three-dimensional model is smoothed by applying the Taubin smoothing algorithm, and the visual effect and accuracy of the three-dimensional model are improved by using edge sharpening, texture enhancement and other technologies.

[0062] Step S2: Calculate the optimal three-dimensional path of the surgical robot using an improved genetic algorithm based on the three-dimensional model and the real-time position of the surgical robot. Figure 2 As shown, including:

[0063] Step S2.1: Initialize the population and randomly generate a certain number of individuals (i.e., potential paths) as the initial population. For example, the population size is Q and the crossover probability is P. c Before solving the problem, it is important to encode the solution space and determine the genetic encoding method, so as to establish a one-to-one mapping relationship between the problem and the genetic algorithm. The actual encoding method is usually represented by real numbers or floating-point numbers within a certain range, which can directly decode the solution space, reduce the amount of calculation, and improve the search efficiency of the algorithm. Therefore, this embodiment adopts the actual encoding method as the genetic encoding. The genetic encoding method is as follows: Figure 3 As shown, where X n , Y n Indicates the horizontal coordinate, Z n is the height. Such a set of genes represents a path from the starting position to the target. Determine the solution space, each individual consists of a series of genes (path point coordinates), representing a path from the starting position to the target.

[0064] Step S2.2, calculate the fitness of each individual according to the preset fitness function. The fitness function is designed using the penalty function. The core of three-dimensional path planning is to design a path that meets the constraints and is the shortest and safest. We consider path length, obstacle avoidance ability, and surgical safety as factors for fitness function evaluation.

[0065] The path length is:

[0066]

[0067] Obstacle avoidance capabilities:

[0068]

[0069] Let each point p on the path i The distance to the nearest obstacle is d i ,ε is a small positive number used to avoid division by zero. The reciprocal of the distance is used here because the larger the distance (i.e., the farther away from the obstacle), the smaller the obstacle avoidance ability evaluation value, which is opposite to the fitness maximization direction usually pursued in the genetic algorithm, so the reciprocal is needed.

[0070] The formula for calculating surgical safety is:

[0071]

[0072] Suppose the output of the risk assessment model based on machine learning is R(P), where P represents the path. Similarly, the inverse form is used here to ensure the consistency of the fitness function.

[0073] The machine learning-based risk assessment model is implemented through the following steps:

[0074] Historical surgical data were collected, including successful and failed surgical cases, as well as safety assessment results of the surgical path in each case, which could be scored by experienced physicians or determined based on postoperative complications.

[0075] Secondly, a supervised learning model is selected. Since it involves predicting unknown data based on known data, it is suitable to use a supervised learning model, such as logistic regression, random forest, gradient boosting tree (GBDT), neural network, etc. This embodiment uses a neural network model.

[0076] Finally, use the data to train the machine learning model, adjust the model parameters, and optimize the model performance.

[0077] For a new surgical path P, its relevant features are extracted, including the distance from each point on the path to the heart valve, the curvature of the path, whether the path passes through a high-risk area (such as a valve area with severe calcification), and the relative position of the path and the area where known complications occur. These features are input into the trained machine learning model to obtain the risk assessment value R(P) of path P.

[0078] Assume that the weights of path length, obstacle avoidance ability and surgical safety are w L 、w O and w S , and these weights can be dynamically adjusted according to the surgical stage, then the comprehensive fitness function F can be expressed as:

[0079] F=w L ·F L +w O ·F O +w S ·F S

[0080] in

[0081] w L +w O +w S =1

[0082] At different stages of the operation, the L 、w O and w S In the early stage of surgery, when the robot needs to reach the target area quickly, the optimization of path length is particularly important. At this time, w should be increased. L However, when approaching the surgical target, if the path length conflicts with other factors (obstacle avoidance and safety), w needs to be appropriately reduced. L The weight of w should be increased to ensure the safety of the path and the ability to avoid obstacles. During surgery, if the robot encounters complex or unknown anatomical structures, w should be increased. O The weight of w should be increased to ensure that the robot path planning can bypass all potential obstacles and avoid unnecessary damage to the patient. S weights to ensure that surgical instruments can safely pass through complex areas.

[0083] Because F L 、F O 、F S The consideration criteria of each function are very different. In order to make each function have the same sensitivity to the total fitness function, we standardize the above three functions. The standardization formula is as follows:

[0084]

[0085] Where P i Indicates a path.

[0086] After normalization, F k (k=L,O,S) becomes So the fitness function becomes:

[0087]

[0088] Step S2.3, select individuals with high fitness as parents. For example, this embodiment uses the roulette method as the selection strategy. The roulette selection method has a certain randomness, which helps the algorithm to jump out of the local optimal solution. The basic idea is that the probability of each individual being selected is proportional to its fitness. The implementation process can be summarized into the following steps:

[0089] First, calculate the probability P of each individual being inherited to the next generation population. i , this probability can be directly obtained by dividing the individual fitness by the sum of the group fitness:

[0090]

[0091] Then calculate the cumulative probability q for each individual i , which represents the sum of the selection probabilities of the first i individuals:

[0092]

[0093] Then generate a uniformly distributed pseudo-random number r in the interval [0,1], which is used to simulate the pointer position in roulette. Traverse the cumulative probability array q, find the first k value that satisfies q[k-1]<r≤q[k], and select the kth individual as the parent. This process is repeated until a sufficient number of parent individuals are selected.

[0094] In addition, in this process, in order to avoid the loss of excellent individuals, an elite retention strategy can be introduced in the selection process, that is, the individuals with the highest fitness are directly retained to the next generation.

[0095] Step S2.4: Randomly select two parent individuals from the selected individuals, perform a crossover operation, generate new offspring individuals, and form a new population. For example, multi-point crossover is used to cross the parents. This method increases the diversity of the population and helps search a wider solution space by selecting multiple intersection points on the three-dimensional path and exchanging parts of the path between these points. The following is a detailed description of this process:

[0096] First, two parent individuals are selected from the current population according to the fitness function.

[0097] Then, multiple intersections are randomly selected on the three-dimensional path. The number of intersections can be determined based on the complexity of the path and the specific requirements of the problem. In heart valve intervention surgery, 2 to 3 intersections can be selected appropriately. Ensure that the intersections on the two parent paths are arranged in the same order to facilitate subsequent path exchange.

[0098] Next, the parent path is split into multiple sub-segments at each intersection. For each parent, the sub-segments are alternately selected for exchange from the first intersection to the last intersection. For example, the first sub-segment of parent A is exchanged with the first sub-segment of parent B, the second sub-segment of parent A is exchanged with the second sub-segment of parent B, and so on. Finally, the exchanged sub-segments are reconnected to form two new child paths.

[0099] Step S2.5, determine whether the new population reaches the termination condition, which determines when the algorithm stops iterating and outputs the optimal solution. If it is satisfied, the algorithm ends and outputs the optimal three-dimensional path. If it is not satisfied, repeat the steps 2.2 to 2.4. More specifically, the termination condition is set first. One is to reach the preset number of iterations. Before the algorithm starts, a maximum number of iterations (T) is set. When the number of algorithm iterations reaches this value, the algorithm stops;

[0100] The second is to determine whether a solution space satisfies certain conditions (such as the shortest path and the safest). If so, the solution is considered good enough and the algorithm can be terminated early.

[0101] For the first termination condition, an iteration counter (t) can be maintained in the algorithm, and the value of this counter is updated at the end of each iteration (t=t+1).

[0102] For the second termination condition, the present invention chooses to use the path as the judgment condition, sets a suitable threshold L0, and compares the path length L of each solution with L0 after each iteration. If L≤L0, the termination condition is met.

[0103] Step S3: According to the actual situation during the operation (such as blood vessel deformation, robot deviation, etc.), the three-dimensional model is dynamically updated through real-time environmental detection and image information, and step S2 is repeated.

[0104] If the above-mentioned three-dimensional path planning method of the cardiac interventional surgery robot is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Therefore, the essence of this technical solution or the part that contributes to the prior art or the part of this technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0105] In summary, the present invention provides a three-dimensional path planning method and computer program product for a cardiac interventional surgery robot, which dynamically adjusts and updates the three-dimensional model in real time according to real-time environmental detection and image information during surgery; calculates the optimal three-dimensional path of the surgical robot through an improved genetic algorithm according to the three-dimensional model and the real-time position of the surgical robot; the fitness function F preset by the genetic algorithm takes into account the path length F L , obstacle avoidance capability F o 、Surgery safety S , at different stages of the operation, different optimization goals are emphasized by adjusting the weights of the factors considered. The present invention calculates the optimal three-dimensional path of the surgical robot through a genetic algorithm that considers path length, obstacle avoidance, and surgical safety, realizes intelligent and precise planning of the surgical path, and has real-time feedback and dynamic adjustment capabilities during the operation, thereby improving the accuracy and safety of the operation.

[0106] Those skilled in the art should understand that those skilled in the art can implement variations by combining the prior art and the above embodiments, which will not be described in detail here. Such variations do not affect the essential content of the present invention, and will not be described in detail here.

[0107] The above describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and the devices and structures that are not described in detail should be understood to be implemented in a common manner in the art; any technician familiar with the art can use the above-disclosed methods and technical contents to make many possible changes and modifications to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, or modify them into equivalent embodiments of equivalent changes, which does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention are still within the scope of protection of the technical solutions of the present invention.

[0108] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Claims

1. A three-dimensional path planning method for a cardiac interventional surgery robot, characterized in that: The following steps are involved: Read the three-dimensional structural medical image of the patient's heart and construct a three-dimensional model of the patient's heart based on it; dynamically adjust and update the three-dimensional model in real time according to the real-time environmental detection and image information during the operation; According to the three-dimensional model and the real-time position of the surgical robot, the optimal three-dimensional path of the surgical robot is calculated by an improved genetic algorithm; the fitness function F preset by the genetic algorithm takes into account the path length F L , obstacle avoidance capability F o 、Surgery safety S , expressed as: F=w L ·F L +w O ·F O +w S ·F S ; The weight w L +w O +w S =1, Where x i ,y i Indicates the horizontal coordinate, z i is the height coordinate, d i For each point p on the path i The distance to the nearest obstacle, ε is a small positive number; R(P) is the output of the risk assessment model based on machine learning; At different stages of the operation, by adjusting w L 、w O and w S to emphasize different optimization objectives.

2. The three-dimensional path planning method of a cardiac interventional surgery robot according to claim 1, characterized in that: For path length F L , obstacle avoidance capability F o 、Surgery safety S Standardization is performed, and the standardization formula is: Where P i Indicates path; F k (k=L,O,S) is standardized to The fitness function is:

3. The three-dimensional path planning method of a cardiac interventional surgery robot according to claim 1 or 2, characterized in that: Building a 3D model of the patient's heart includes the following sub-steps: Step S1.1, using median filtering, bilateral filtering and Gaussian filtering to process the image in sequence; Step S1.2, using a deep learning model to segment the image processed in step S1.1, identifying and marking heart valves, blood vessels and tissue structures; Step S1.3, reconstruct the structures including heart valves and blood vessels in the image in three dimensions, and use the registration technology to fuse the heart valve segmentation module, blood vessel feature module and robot segmentation module to obtain a three-dimensional model that integrates multiple information.

4. The three-dimensional path planning method of a cardiac interventional surgery robot according to claim 3, characterized in that: The deep model of step S1.2 is a U-Net-based deep model, which has an encoder-decoder structure and can fuse low-level and high-level features.

5. The three-dimensional path planning method for a cardiac interventional surgery robot according to claim 3, characterized in that: When performing three-dimensional reconstruction in step S1.3, the three-dimensional model is finely processed using Taubin smoothing and detail enhancement.

6. The three-dimensional path planning method of a cardiac interventional surgery robot according to claim 1 or 2, characterized in that: Calculating the optimal three-dimensional path of the surgical robot by genetic algorithm includes the following sub-steps: Step S2.1, initialize the population, randomly generate a certain number of individuals (i.e., potential paths) as the initial population; Determine the solution space, where each individual consists of a series of genes (path point coordinates), representing a path from the starting position to the target; Step S2.2, calculating the fitness of each individual according to the preset fitness function; Step S2.3, select individuals with high fitness as parents; Step S2.4: Randomly select two parent individuals from the selected individuals, perform a crossover operation, generate new offspring individuals, and form a new population; Step S2.5, determine whether the new population reaches the termination condition; if it does, end the algorithm and output the optimal three-dimensional path; if it does not, repeat steps 2.2 to 2.

4.

7. The three-dimensional path planning method for a cardiac interventional surgery robot according to claim 6, characterized in that: In step S2.3, the roulette method is used to select the parent: first, the probability Pi of each individual being inherited to the next generation population is calculated; then the cumulative probability qi of each individual is calculated, which represents the sum of the selection probabilities of the previous i individuals; then a uniformly distributed pseudo-random number r is generated in the interval [0,1], and this random number is used to simulate the pointer position in the roulette; the cumulative probability array q is traversed to find the first k value that satisfies q[k-1]<r≤q[k], and the kth individual is selected as the parent; an elite retention strategy is introduced in the selection process to directly retain individuals with high fitness to the next generation.

8. The three-dimensional path planning method for a cardiac interventional surgery robot according to claim 6, characterized in that: When selecting the parent generation for crossover in step S2.4, a multi-point crossover method is used.

9. The three-dimensional path planning method of a cardiac interventional surgery robot according to claim 1 or 2, characterized in that: The steps for implementing the machine learning-based risk assessment model to predict possible complications or risks under different pathways are as follows: Extracting features related to the safety of the surgical path from the three-dimensional model; the features include the distance from each point on the path to the key anatomical structure, the curvature of the path, whether the path passes through a high-risk area, and the relative position of the path to a known complication-prone area; Collect historical surgical data, including successful and failed surgical cases, and the safety assessment results of the surgical path in each case as label data; The extracted features and label data are used to train any supervised learning model including logistic regression, random forest, gradient boosting tree, and neural network. After optimizing the parameters of the supervised learning model, a risk assessment model based on machine learning is obtained, and its output is R(P).

10. A computer program product, characterized in that When the computer program product runs on a computer or device, the computer or device executes the three-dimensional path planning method for a cardiac interventional surgery robot as described in any one of claims 1 to 9.

Images