Puncture needle path planning method and system based on adaptive local optimization algorithm

Abstract

The present invention relates to the field of medical puncture technology, and specifically to a puncture needle path planning method and system based on an adaptive local optimization algorithm, comprising the following steps: using an optimization algorithm to solve the kinematic model of the puncture needle puncture path reaching each target point in turn, and obtaining the puncture needle puncture global path for puncturing each target point one by one; encapsulating the mapping relationship between the puncture needle puncture global path and the puncture global environmental information through a neural network, and obtaining a global planning model for obtaining the puncture global path according to the puncture global environmental information; performing transfer learning on the global planning model, and obtaining a local planning model for obtaining the puncture local path according to the puncture local environmental information. The present invention realizes real-time local adjustment of preoperative puncture path planning to fully adapt to the real puncture scene and improve the robustness of puncture path planning.

Description

Technical Field

[0001] The present invention relates to the field of medical control technology, and in particular to a puncture needle path planning method and system based on an adaptive local optimization algorithm. Background Art

[0002] With the rapid development of minimally invasive surgery, puncture has been widely used in clinical practice as a minimally invasive surgery, which can complete tasks such as biopsy, brachytherapy, drug delivery and targeted therapy.

[0003] Currently, the most widely used puncture surgeries are basically performed along the puncture path planned before the operation with the assistance of medical imaging equipment. Although the preoperative puncture path planning has completed the global path planning, when the puncture operation is actually performed, it will be difficult to fully adapt the puncture guidance due to the real-time situation of the tissue, resulting in planning failure. If the original path planning is still followed at this time, it will cause puncture failure (the needle tip collides with obstacles, etc.) and cause accidental injuries. If the original path planning is not followed, the operation needs to be suspended. The puncture planning is based on the operator's judgment, and the accuracy is difficult to guarantee, let alone maintaining the consistency of the planning with the original path, which ultimately delays the puncture time.

[0004] Therefore, the existing puncture path planning technology is not robust enough, it is difficult to adapt to real-time complex situations, and the path planning effect is difficult to guarantee. Summary of the invention

[0005] The purpose of the present invention is to provide a puncture needle path planning method and system based on an adaptive local optimization algorithm to solve the technical problems in the prior art that the puncture path planning is insufficiently robust, difficult to adapt to real-time complex situations, and difficult to ensure the path planning effect.

[0006] In order to solve the above technical problems, the present invention specifically provides the following technical solutions:

[0007] The puncture needle path planning method based on the adaptive local optimization algorithm includes the following steps:

[0008] Performing kinematic analysis on the puncture needle, constructing a kinematic model of the puncture needle for puncture path planning of the puncture needle, and constructing a kinematic model of the puncture path of the puncture needle;

[0009] The kinematic model of the puncture needle puncture path reaching each target point is solved in turn by using an optimization algorithm to obtain a global puncture needle puncture path for puncturing each target point one by one;

[0010] The mapping relationship between the puncture needle global path and the puncture global environment information is encapsulated by a neural network to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0011] Transfer learning is performed on the global planning model to obtain a local planning model for obtaining a local puncture path according to the local puncture environment information.

[0012] As a preferred embodiment of the present invention, the method for constructing the kinematic model of the puncture needle includes:

[0013] Set the forward speed of the puncture needle to v The angular velocity of the puncture needle is ω, and the coordinate system of the puncture needle is Ψ p , the puncture needle is in the world coordinate system Ψ g The position and posture of g =[x,y,z,α,β,γ] T , where α, β, and γ are the distances of the puncture needle around the world coordinate system Ψ g The rotation angle of the x, y, and z axes; in the world coordinate system Ψ g In the experiment, the velocity of the puncture needle is determined as V g =x g & ;

[0014] In the needle body coordinate system Ψ p In the experiment, the speed of the puncture needle is determined as V p = v V 1 +W 2 , where V 1 =[001k00] T , V 2 =[000001] T ;

[0015] Based on the world coordinate system Ψ g and the needle body coordinate system Ψ p , the velocity conversion equation of the puncture needle is: V p =QV g , where is the world coordinate system Ψ g and the needle body coordinate system Ψ p The rotation matrix of the puncture needle establishes the world coordinate system Ψ g and the needle body coordinate system Ψ p The speed transformation relationship between Ω is the matrix coefficient of the Euler kinematics equation for rigid body fixed-point motion, which establishes the world coordinate system Ψ g and the needle body coordinate system Ψ p The angular velocity transformation relationship between them;

[0016] According to V g =x g & 、Vp= v V1 +W 2 and V p =QV g , the kinematic model of the puncture needle is determined as: ;

[0017] The time factor in the puncture needle kinematic model is separated to eliminate the puncture speed and rotation speed in the control quantity of the puncture needle motion model, and retain the needle insertion distance s and rotation angle , and the kinematic model of the puncture needle after separation is obtained .

[0018] As a preferred embodiment of the present invention, the method for constructing the kinematic model of the puncture path of the puncture needle includes:

[0019] Kinematic model of puncture needle Apply the stop-and-turn strategy to obtain the control sequence of the puncture needle puncture path , where s 1 、s 2 and n are the lengths of the first, second and nth segments of the puncture path of the puncture needle, , and are the rotation angles at the 1st, 2nd and nth segments of the puncture path of the puncture needle respectively; the control sequence based on the puncture path of the puncture needle , and the needle tip of the puncture needle moves along a fixed radius r to make a standard arc motion, and the puncture needle puncture path is determined to consist of n arc paths Composition, i The first i The length of the segment path, and ,in, For i The corresponding arc, r is the movement radius of the needle tip;

[0020] The control sequence of the puncture path of the puncture needle is composed of the needle insertion distance s and the rotation angle The kinematic model of the puncture path of the puncture needle is obtained by converting the coordinates of the center of the path and the coordinates and length of the starting point of each path segment. , where λ i For the i The starting coordinates of the segment path, λ i c is the i Segment path starting point coordinates λ i The tangent vector at O i For the i The center coordinates of the segment path arc;

[0021] Among them, λ i ,λi c, O i ,s i The relationship between is:

[0022] ;

[0023] ;

[0024] ;

[0025] In the formula, λ i+1 For the i +1 The starting point coordinates of the path, λ i+1 c is the i +1 path starting point coordinates λ i+1 The tangent vector at O i+1 For the i +1 The center coordinates of the arc of the path segment.

[0026] As a preferred solution of the present invention, the method for obtaining the puncture global environment information includes:

[0027] Using medical imaging equipment to collect medical images of the puncture environment;

[0028] The position information of the puncture needle entry point, the target point and obstacles is obtained in the medical image as the puncture global environment information.

[0029] As a preferred solution of the present invention, the method for acquiring the puncture local environment information includes:

[0030] Using the medical image to perform three-dimensional reconstruction to obtain a puncture environment model;

[0031] The SLAM system loaded on the front end of the puncture needle moves along the global puncture path located on the puncture environment model to obtain obstacle information on the global puncture path as the local puncture environment information.

[0032] As a preferred solution of the present invention, the method for constructing the global planning model includes:

[0033] The puncture global environment information is used as an input item of the neural network, and the solution result Q_result of the kinematic model Q of the puncture global path obtained by the optimization algorithm according to the puncture global environment information is used as an output item of the neural network;

[0034] The mapping relationship between the puncture global environment information and the puncture global path is trained by using the neural network between the input items and the output items of the neural network, so as to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0035] The global planning model is:

[0036] Q_resulto=neural_network(G);

[0037] Where Q_resulto is the global puncture path output by the global planning model, G is the global puncture environment information, and neural_network is the neural network.

[0038] As a preferred solution of the present invention, the method for constructing the local planning model includes:

[0039] From the solution result Q_result of the kinematic model of the global puncture path, each path segment having the same position as the local puncture environment information is extracted as the local puncture path Q_m;

[0040] The puncture local environment information is used as a migration input item of the global planning model, and the puncture local path Q_m is used as a migration output item of the global planning model;

[0041] Transfer the mapping relationship between the puncture global environment information and the puncture global path to the mapping relationship between the puncture local environment information and the puncture local path, and obtain a local planning model for obtaining the puncture local path according to the puncture local environment information.

[0042] The local planning model is:

[0043] Q_mo=neural_network(g);

[0044] Where Q_mo is the local puncture path output by the local planning model, g is the local puncture environment information, and neural_network is the neural network;

[0045] The loss function of the global planning model transfer learning is:

[0046] L=||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 +||Q_mo-Q_m || 1 ;

[0047] Where L is the loss function value, Q_mf is the path at the front end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, Q_mb is the path at the back end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, ||(Q_mo+Q_mf+Q_mb)-Q_resulto || 1 and ||Q_mo-Q_m||1 All are L1 norm forms.

[0048] As a preferred solution of the present invention, a method for solving the kinematic model of the puncture path of the puncture needle reaching each target point in sequence by using an optimization algorithm includes:

[0049] Sort each target point from large to small according to the coordinate value in the z-axis direction to obtain the path planning sequence of the target points;

[0050] Using an optimization algorithm based on a multi-objective function to plan the puncture path of the first target point in the path planning sequence of the target points, the puncture path of the puncture needle is obtained;

[0051] After the initial path planning of the puncture needle is completed, the optimization algorithm will be used to perform a new planning operation on the second target point based on the multi-objective function starting from the end point of the initial path of the puncture needle.

[0052] If the puncture path to the second target point cannot be planned at the end point of the initial puncture path of the puncture needle, the puncture needle tip starts to retreat to the position where the needle stopped and turned last on the current path, and the puncture path is planned for the second target point again, and so on, until the path to the second target point is successfully planned;

[0053] If the path planning for the second target point cannot be successfully performed until the puncture needle tip retracts to the needle insertion point, a failure prompt will be given.

[0054] As a preferred solution of the present invention, the multi-objective function is: Y=k 1 Y tip (Q)+k 2 Y s (Q)+Y d (Q);

[0055] Y tip (Q)=|P tip (Q)-P d |;

[0056] Y s (Q)= ;

[0057] Y d (Q)= ;

[0058] ; In the formula, Y is the value of the multi-objective function, k 1 is the weight of the path deviation, k 2 is the weight of the total path length, Y tip (θ) is the error function between the target point and the needle tip of the puncture needle, Ptip (Θ) is the coordinate of the end point of the puncture path determined by Θ, P d is the coordinate of the target point, Y s (Θ) is the path length function, s i It is the first i The length of the segment path, Y d (Θ) is the function to determine whether the needle tip collides with an obstacle, m is the number of obstacles, and y j (d) is the collision function of the jth obstacle, d is the safe distance between the obstacle and the needle tip of the puncture needle, when the needle tip collides with the obstacle, y j (d) The value is +∞, otherwise y j As a preferred embodiment of the present invention, the present invention provides a puncture needle path planning system based on an adaptive local optimization algorithm, which is applied to the puncture needle path planning method based on an adaptive local optimization algorithm, and the system includes:

[0059] A data acquisition unit, used to obtain global puncture environment information and local puncture environment information;

[0060] A data processing unit, used for performing kinematic analysis on the puncture needle, constructing a kinematic model of the puncture needle for planning the puncture path of the puncture needle, and constructing a kinematic model of the puncture path of the puncture needle;

[0061] The kinematic model of the puncture needle puncture path reaching each target point is solved in turn by using an optimization algorithm to obtain a global puncture needle puncture path for puncturing each target point one by one;

[0062] The mapping relationship between the puncture needle global path and the puncture global environment information is encapsulated by a neural network to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0063] Performing transfer learning on the global planning model to obtain a local planning model for obtaining a local puncture path according to local puncture environment information;

[0064] The path planning unit is used to plan the global puncture path of the puncture needle in the global puncture environment based on the global planning model, and to locally adjust the global puncture path of the puncture needle according to the local puncture environment using the local planning model.

[0065] As a preferred solution of the present invention, the method for constructing the global planning model by the data processing unit includes:

[0066] The puncture global environment information is used as an input item of the neural network, and the solution result Q_result of the kinematic model Q of the puncture global path obtained by the optimization algorithm according to the puncture global environment information is used as an output item of the neural network;

[0067] The mapping relationship between the puncture global environment information and the puncture global path is trained by using the neural network between the input items and the output items of the neural network, so as to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0068] The global planning model is:

[0069] Q_resulto=neural_network(G);

[0070] Wherein, Q_resulto is the global puncture path output by the global planning model, G is the global puncture environment information, and neural_network is the neural network. As a preferred solution of the present invention, the method for constructing the local planning model by the data processing unit includes:

[0071] From the solution result Q_result of the kinematic model of the global puncture path, each path segment having the same position as the local puncture environment information is extracted as the local puncture path Q_m;

[0072] The puncture local environment information is used as a migration input item of the global planning model, and the puncture local path Q_m is used as a migration output item of the global planning model;

[0073] Transfer the mapping relationship between the puncture global environment information and the puncture global path to the mapping relationship between the puncture local environment information and the puncture local path, and obtain a local planning model for obtaining the puncture local path according to the puncture local environment information.

[0074] The local planning model is:

[0075] Q_mo=neural_network(g);

[0076] Where Q_mo is the local puncture path output by the local planning model, g is the local puncture environment information, and neural_network is the neural network;

[0077] The loss function of the global planning model transfer learning is:

[0078] L=||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 +||Q_mo-Q_m || 1;

[0079] Where L is the loss function value, Q_mf is the path at the front end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, Q_mb is the path at the back end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, ||(Q_mo+Q_mf+Q_mb)-Q_resulto || 1 and ||Q_mo-Q_m|| 1 All are L1 norm forms.

[0080] Compared with the prior art, the present invention has the following beneficial effects:

[0081] The present invention uses a neural network to encapsulate the mapping relationship between the global puncture path of the puncture needle obtained by the optimization algorithm and the global puncture environment information, and obtains a global planning model for obtaining the global puncture path according to the global puncture environment information, and transfers the learning to a local planning model for obtaining the local puncture path according to the local puncture environment information, thereby realizing real-time local adjustment of the preoperative puncture path planning to fully adapt to the actual puncture scene and improve the robustness of the puncture path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0083] Figure 1 A flow chart of a puncture needle path planning method based on an adaptive local optimization algorithm provided in an embodiment of the present invention;

[0084] Figure 2 A block diagram of a puncture needle path planning system based on an adaptive local optimization algorithm provided in an embodiment of the present invention;

[0085] Figure 3 A schematic diagram of a kinematic model of a puncture needle provided in an embodiment of the present invention;

[0086] Figure 4 A schematic diagram of a kinematic model of a puncture needle puncture path provided in an embodiment of the present invention;

[0087] Figure 5 A schematic diagram of needle tip retraction provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] like Figure 1 As shown, the present invention provides a puncture needle path planning method based on an adaptive local optimization algorithm, comprising the following steps:

[0090] Performing kinematic analysis on the puncture needle, constructing a kinematic model of the puncture needle for puncture path planning of the puncture needle, and constructing a kinematic model of the puncture path of the puncture needle;

[0091] The kinematic model of the puncture needle puncture path reaching each target point is solved in turn by using an optimization algorithm to obtain a global puncture needle puncture path for puncturing each target point one by one;

[0092] The mapping relationship between the global puncture path of the puncture needle and the global puncture environment information is encapsulated by a neural network to obtain a global planning model for obtaining the global puncture path according to the global puncture environment information;

[0093] Transfer learning is performed on the global planning model to obtain a local planning model for obtaining a local puncture path according to the local puncture environment information.

[0094] In order to improve the robustness of puncture path planning and adapt to the real-time complex environment of the puncture scene, the present invention first performs global planning to obtain the global puncture path, and then performs local planning according to the actual situation during the puncture process to achieve real-time adjustment of the global path, so that the planned puncture path can adapt to the changes in the complex puncture environment.

[0095] Specifically, the present invention uses a swarm intelligence algorithm or an optimization algorithm (particle swarm algorithm, genetic algorithm, simulated annealing algorithm, etc.) to solve the optimal global puncture path according to the global environmental information of the puncture scene, which can be used for global puncture guidance.

[0096] However, global planning is based on the environmental information fixed before the operation. It can be seen that the global puncture path is a historical plan, which is difficult to fully adapt to the current puncture scene. Therefore, the present invention adds local planning on the basis of global guidance to provide real-time local guidance based on the actual situation of the puncture scene. For example, if an obstacle appears on the global path during the puncture process, local planning is used to achieve local adjustment of the global path.

[0097] The present invention performs local planning through a local planning model, and the local planning model is learned by transfer learning from the global planning model. Since the training samples of the global planning model come from the search and solution of a swarm intelligence algorithm or an optimization algorithm, the output result of the global planning model is consistent with the solution result of the kinematic model Q of the global puncture path. Therefore, the local planning model will inherit the planning style of the global planning model, or in other words, the local path planned by the local planning model is consistent with the global path planned by the global planning model. Furthermore, the local path planned by the local planning model is consistent with the solution result of the kinematic model Q of the global puncture path, and can be spliced ​​to the original global path to maintain path continuity, thereby ensuring the continuity of the puncture operation and improving the puncture effect.

[0098] In order to plan the path of the puncture needle, the kinematics of the puncture needle must be analyzed first. First, assume that the soft tissue being punctured is uniform and the tissue does not deform during the puncture process; when the needle tail is rotated, the bevel direction of the needle tip follows the rotation of the needle tail; the needle body completely follows the movement of the needle tip. After entering the soft tissue, the needle body is fixed by the tissue and does not change position in the radial direction. Based on the above assumptions, when puncturing in the soft tissue, the needle tip moves along an arc with a curvature of k. The arc is determined by the needle tip posture and the direction of the needle tip bevel. The kinematic model of the puncture needle is established as follows:

[0099] like Figure 3 As shown, the method for constructing the kinematic model of the puncture needle includes:

[0100] Set the forward speed of the puncture needle to v The angular velocity of the puncture needle is ω, and the coordinate system of the puncture needle is Ψ p , the puncture needle is in the world coordinate system Ψ g The position and posture of g =[x,y,z,α,β,γ] T , where α, β, and γ are the distances of the puncture needle around the world coordinate system Ψ g The rotation angle of the x, y, and z axes; in the world coordinate system Ψ g In the experiment, the speed of the puncture needle is determined as V g =x g & ; In the needle body coordinate system Ψ p In the experiment, the speed of the puncture needle is determined as V p = v V 1 +W 2 , where V 1 =[001k00] T , V 2 =[000001] T ; Based on the world coordinate system Ψ g and the needle body coordinate system Ψp , the velocity conversion equation of the puncture needle is: V p =QV g , where is the world coordinate system Ψ g and the needle body coordinate system Ψ p The rotation matrix of the puncture needle establishes the world coordinate system Ψ g and the needle body coordinate system Ψ p The speed transformation relationship between Ω is the matrix coefficient of the Euler kinematics equation for rigid body fixed-point motion, which establishes the world coordinate system Ψ g and the needle body coordinate system Ψ p According to the angular velocity transformation relationship between V g =x g & 、Vp= v V 1 +W 2 and V p =QV g , the kinematic model of the puncture needle is determined as: ;

[0101] The time factor in the puncture needle kinematic model is separated to eliminate the puncture speed and rotation speed in the control quantity of the puncture needle motion model, and retain the needle insertion distance s and rotation angle ,get .

[0102] The puncture needle model is a nonlinear system and a non-holonomic constraint system, so it is difficult to plan the path directly based on this model. The puncture needle trajectory is independent of the puncture speed and rotation speed, so the time factor can be separated from the control quantity, and only the needle insertion distance s and the rotation angle θ are used as the control quantity.

[0103] The method for constructing the kinematic model of the puncture needle puncture path includes:

[0104] Kinematic model of puncture needle Apply the stop-and-turn strategy to obtain the control sequence of the puncture needle puncture path , where s 1 、s 2 and n are the lengths of the first, second and nth segments of the puncture path of the puncture needle, , and are the rotation angles at the 1st, 2nd and nth segments of the puncture path of the puncture needle respectively; the control sequence based on the puncture path of the puncture needle , and the needle tip of the puncture needle moves along a fixed radius r to make a standard arc motion, and the puncture needle puncture path is determined to consist of n arc paths Composition, i The first i The length of the segment path, and ,in, for The corresponding arc, r is the movement radius of the needle tip.

[0105] like Figure 4 As shown, the control sequence of the puncture path of the puncture needle is composed of the needle insertion distance s and the rotation angle The kinematic model of the puncture path of the puncture needle is obtained by converting the coordinates of the center of the path and the coordinates and length of the starting point of each path segment. , where λ i For the i The starting coordinates of the segment path, λ i c is the i Segment path starting point coordinates λ i The tangent vector at O i For the i The center coordinates of the segment path arc;

[0106] Among them, λ i ,λ i c, O i ,s i The relationship between is:

[0107] ;

[0108] ;

[0109] ;

[0110] In the formula, λ i+1 For the i +1 The starting point coordinates of the path, λ i+1 c is the i +1 path starting point coordinates λ i+1 The tangent vector at O i+1 For the i +1 The center coordinates of the arc of the path segment.

[0111] Using swarm intelligence algorithms / optimization algorithms, such as particle swarm optimization, genetic algorithm, simulated annealing algorithm, etc., based on multi-objective functions, search and solve to obtain the solution result Q_result of the kinematic model of the global puncture path. The method for obtaining the global puncture environment information includes:

[0112] Using medical imaging equipment to collect medical images of the puncture environment;

[0113] The location information of the puncture needle point, target point and obstacles is obtained in the medical image as the global environment information of the puncture.

[0114] The method for obtaining the local environment information of the puncture includes:

[0115] The puncture environment model is obtained by three-dimensional reconstruction using medical images;

[0116] The SLAM system loaded at the front end of the puncture needle moves along the global puncture path located on the puncture environment model to obtain obstacle information on the global puncture path as the local puncture environment information.

[0117] The SLAM (Simultaneous Localization and Mapping) system can sense the pressure of the blood vessel wall, the blood flow speed and resistance in the blood vessel after puncturing into the blood vessel model through visual sensors and pressure sensors or laser sensors that perceive the blood vessel wall. This is equivalent to sensing obstacles. When the puncture needle senses an obstacle while moving along the global puncture path, it collects its information and transmits it to the local planning model to plan the local path at the current position, so as to make local adjustments to the global path to adapt to the current puncture scene.

[0118] The present invention performs local planning through a local planning model, and the local planning model is learned by transfer learning from the global planning model. Since the training samples of the global planning model come from the search and solution of a swarm intelligence algorithm or an optimization algorithm, the output result of the global planning model is consistent with the solution result of the kinematic model Q of the global puncture path. Therefore, the local planning model inherits the planning style of the global planning model, or the local path planned by the local planning model is consistent with the global path planned by the global planning model. Then, the local path planned by the local planning model is consistent with the solution result of the kinematic model Q of the global puncture path, and can be spliced ​​to the original global path to maintain the path coherence, thereby ensuring the coherence of the puncture operation and improving the puncture effect, as follows:

[0119] The construction method of the global planning model includes:

[0120] The global environment information of the puncture is used as the input item of the neural network, and the solution result Q_result of the kinematic model Q of the global puncture path obtained by the optimization algorithm according to the global environment information of the puncture is used as the output item of the neural network;

[0121] The mapping relationship between the puncture global environment information and the puncture global path is trained by using the neural network between the input items and the output items of the neural network, so as to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0122] The global planning model is:

[0123] Q_resulto=neural_network(G);

[0124] Where Q_resulto is the global puncture path output by the global planning model, G is the global puncture environment information, and neural_network is the neural network.

[0125] The construction method of the local planning model includes:

[0126] From the solution result Q_result of the kinematic model of the global puncture path, each path segment having the same position as the local puncture environment information is extracted as the local puncture path Q_m;

[0127] The local environment information of the puncture is used as the migration input item of the global planning model, and the local path Q_m of the puncture is used as the migration output item of the global planning model;

[0128] Transfer the mapping relationship between the puncture global environment information and the puncture global path to the mapping relationship between the puncture local environment information and the puncture local path, and obtain a local planning model for obtaining the puncture local path according to the puncture local environment information.

[0129] The local planning model is:

[0130] Q_mo=neural_network(g);

[0131] Where Q_mo is the local puncture path output by the local planning model, g is the local puncture environment information, and neural_network is the neural network;

[0132] The loss function of global planning model transfer learning is:

[0133] L=||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 +||Q_mo-Q_m || 1 ;

[0134] Where L is the loss function value, Q_mf is the path at the front end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, Q_mb is the path at the back end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, ||(Q_mo+Q_mf+Q_mb)-Q_resulto || 1 and ||Q_mo-Q_m|| 1 All are L1 norm forms.

[0135] In order to ensure that the local path planned by the local planning model is consistent with the solution result of the kinematic model Q of the global puncture path, and can be spliced ​​to the original global path to maintain the path coherence, thereby ensuring the coherence of the puncture operation and improving the puncture effect, the loss function is set to L=||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 +||Q_mo-Q_m || 1 , the coherence of splicing the local path planned by the local planning model to the original global path, that is, ||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 , and the obstacle avoidance deviation of the local path planned by the local planning model, i.e., ||Q_mo-Q_m|| 1 , construct a loss function to ensure that the local path obtained by the local planning model can be coherently spliced ​​to the original global path with the minimum obstacle avoidance deviation, and ensure that the local adjustment can maintain the coherence of the puncture path.

[0136] In the present invention, the mapping relationship between the puncture path and the environmental information obtained in the preoperative CT / MRI image is transferred to the mapping relationship between the puncture path and the real-scene image during the puncture process through transfer learning, so that there is no need to continue DSA image acquisition during the puncture process, avoiding the operator's continuous exposure to radiation and improving safety.

[0137] Therefore, the local planning model in the present invention can directly obtain the same path planning performance as the environmental information obtained by the optimization algorithm through CT / MRI images through the real-scene images during the puncture process, without being exposed to radiation.

[0138] When dealing with the problem of multiple target points, the general oblique needle tip puncture needle path planning method needs to re-plan the path from the needle entry point to the target target for each target point. The overall distance of the puncture path planned by this needle insertion strategy is relatively long, which will undoubtedly increase the time of the operation, cause more unnecessary damage to healthy tissues, and also increase the pain of the patient. Therefore, the present invention uses the needle tip insertion and retraction strategy to plan multiple target points, effectively shorten the operation time, and reduce the damage and pain caused to the patient. The needle tip retraction strategy is combined with the optimization algorithm to solve the kinematic model of the puncture needle puncture path containing multiple target points, that is, the planning method of the optimized puncture needle puncture path for puncturing each target target point one by one is as follows:

[0139] Sort each target point from large to small according to the coordinate value in the z-axis direction to obtain the path planning sequence of the target points;

[0140] Using an optimization algorithm based on a multi-objective function to plan the puncture path of the first target point in the path planning sequence of the target points, the puncture path of the puncture needle is obtained;

[0141] After the initial path planning of the puncture needle is completed, the optimization algorithm will be used to perform a new planning operation on the second target point based on the multi-objective function starting from the end point of the initial path of the puncture needle.

[0142] If the puncture path to the second target point cannot be planned at the end point of the initial puncture path of the puncture needle, the puncture needle tip starts to retreat to the position where the needle stopped and turned last on the current path, and the puncture path is planned for the second target point again, and so on, until the path to the second target point is successfully planned;

[0143] If the path planning for the second target point cannot be successfully performed until the puncture needle tip retracts to the needle insertion point, a failure prompt will be given.

[0144] like Figure 5 As shown, in process ①, the needle tip retreats along the original trajectory; in process ②, the needle tip rotates a certain angle; in process ③, the needle tip moves forward in other directions.

[0145] The constraints for optimizing the kinematic model of the puncture path of the puncture needle by the optimization algorithm include:

[0146] The total length of the optimal puncture path is less than or equal to the total length of the puncture needle;

[0147] The arc curvature of each segment of the optimal puncture path is less than or equal to the maximum bending curvature of the puncture needle stiffness.

[0148] Construct a multi-objective function for use by the optimization algorithm as follows:

[0149] Determining the total path length, path deviation, and whether the needle tip encounters an obstacle as multiple optimization objectives of the puncture needle puncture path;

[0150] According to the kinematic model of the puncture needle puncture path and multiple optimization objectives, the multi-objective function of the puncture needle puncture path planning is determined; the multi-objective function is:

[0151] Y=k 1 Y tip (Q)+k 2 Y s (Q)+Y d (Q);

[0152] Y tip (Q)=|P tip (Q)-P d |;

[0153] Y s(Q)= ;

[0154] Y d (Q)= ;

[0155] ;

[0156] In the formula, Y is the value of the multi-objective function, k 1 is the weight of the path deviation, k 2 is the weight of the total path length, k 1 and k 2 The values ​​are all [0,1] and can be set according to the actual situation during use. tip (θ) is the error function between the target point and the needle tip of the puncture needle, P tip (Θ) is the coordinate of the end point of the puncture path determined by Θ, P d is the coordinate of the target point, Y s (Θ) is the path length function, s i It is the first i The length of the segment path, Y d (Θ) is the function to determine whether the needle tip collides with an obstacle, m is the number of obstacles, and y j (d) is the collision function of the jth obstacle, d is the safe distance between the obstacle and the needle tip of the puncture needle. When the needle tip collides with the obstacle, y j (d) The value is +∞, otherwise y j (d) The value is zero.

[0157] The optimization algorithm in the present invention can adopt the currently commonly used swarm intelligence search algorithm and particle swarm algorithm for path planning.

[0158] Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence. The algorithm simulates the foraging behavior of bird flocks and finds the optimal solution to the problem through information sharing and collaboration between particles. The path planning process includes:

[0159] Step 1: Initialization: Randomly generate a group of particles, each with a random position and speed, with each segment of the puncture path (λ i ,λ i c, O i ,s i ), i ∈[1,n] construct particle position X h (t)=[λ h ,λ h c, O h ,s h ]T , and set the particle velocity to V h (t)=[ v 1h , v 2h , v 3h , v 4h ] T , where X h (t) is the position of the hth particle after the tth iteration, V h (t) is the velocity of the hth particle after the tth iteration, λ h ,λ h c, O h ,s h are the starting point coordinate vectors of the n-segment path of the h-th particle, λ h c is the coordinate vector λ of the starting point of the n-segment path h The tangent vector at O h is the coordinate vector of the center of the n-segment path arc, s n is the length vector of the n-segment path, v 1h , v 2h , v 3h , v 4h They are h ,λ h c, O h ,s h The second step: evaluate the fitness: construct the fitness function based on the objective function and calculate the fitness value of each particle.

[0160] Step 3: Individual best and global best update: For each particle, compare its current fitness value with the historical best position (P besth )’s fitness value, if the current value is better, update P besth Find the particle with the best fitness value among all particles and set its position as the global best (G best ).

[0161] Step 4: Speed ​​and position update: Update the speed and position of each particle according to the particle's current speed, individual best position, and global best position. ;X h (t+1)=X h (t)+V h (t+1); where V h (t+1) is the velocity of the hth particle after the t+1th iteration, is the inertia weight, c1 is the individual learning factor of each particle, c 2 is the social learning factor of the particle group, r 1 and r 2 is a random number that follows a uniform distribution in the interval [0,1], P Besth is the historical optimal position of the hth particle, G Best It is the best historical position of all particles. Iteration: Repeat steps 2 to 4 until the termination condition is met, such as reaching the maximum number of iterations T to obtain the optimal solution of the puncture path of the puncture needle.

[0162] In actual use, swarm intelligence algorithms such as genetic algorithms and simulated annealing algorithms can also be used as optimization algorithms, which will not be described in detail here.

[0163] In order to further improve the puncture path planning effect of the optimization algorithm, the present invention solves the defects of the simple particle swarm algorithm in puncture path planning (for example, most of them do not have a strategy to enhance the search for particles that perform well in the search process; secondly, the fitness values ​​of some particles no longer continue to improve after a certain number of iterations, and these particles may stagnate and there is no strategy to deal with this problem, resulting in poor optimization results).

[0164] The present invention also provides another embodiment, a particle swarm algorithm of a swarm strategy obtained by combining a swarm strategy with a particle swarm algorithm, that is, a method for planning a puncture needle puncture path among multiple target points by using an improved particle swarm algorithm, which is specifically as follows:

[0165] Sort each target point from large to small according to the coordinate value in the z-axis direction to obtain the path planning sequence of the target points;

[0166] The bee colony strategy particle swarm algorithm is used to plan the puncture path of the first target point in the path planning sequence of the target points, and the initial puncture path of the puncture needle is obtained;

[0167] After the initial path planning of the puncture needle is completed, the bee colony strategy particle swarm algorithm will be used to perform a new planning operation on the second target point starting from the end point of the initial path of the puncture needle.

[0168] If the puncture path to the second target point cannot be planned at the end point of the initial puncture path of the puncture needle, the puncture needle tip starts to retreat to the position where the needle stopped and turned last on the current path, and the puncture path is planned for the second target point again, and so on, until the path to the second target point is successfully planned;

[0169] If the path planning for the second target point cannot be successfully performed until the puncture needle tip retracts to the needle insertion point, a failure prompt will be given.

[0170] The starting point of the puncture needle puncture path planning is the needle entry point, wherein the needle entry point, target point and obstacle position information are obtained by obtaining the puncture environment information mark through medical imaging equipment.

[0171] The path of each segment of the puncture needle puncture path (λ i ,λ i c, O i ,s i ), i ∈[1,n] construct particle position X h (t)=[λ h ,λ h c, O h ,s h ] T , and set the particle velocity to V h (t)=[ v 1h , v 2h , v 3h , v 4h ] T , where X h (t) is the position of the hth particle after the tth iteration, V h (t) is the velocity of the hth particle after the tth iteration, λ h ,λ h c, O h ,s h are the starting point coordinate vectors of the n-segment path of the h-th particle, λ h c is the coordinate vector λ of the starting point of the n-segment path h The tangent vector at O h is the coordinate vector of the center of the n-segment path arc, s n is the length vector of the n-segment path, v 1h , v 2h , v 3h , v 4h They are h ,λ h c, O h ,s h The movement speed is calculated; the employed learning stage, the spectator learning stage and the reconnaissance learning stage of the bee colony particle swarm algorithm are used to perform particle optimization respectively, and an enhanced search is performed around particles with better fitness values, and particles whose fitness values ​​have not improved after a preset threshold number of iterations are initialized, and the puncture needle puncture path is obtained after the total number of iterations is completed.

[0172] Like other optimization algorithms, artificial bee colony is a swarm intelligence algorithm strategy inspired by the foraging behavior of bee colonies, but the bee colony strategy has the characteristics of division of labor, which divides the bee colony into three groups, namely employed bee colonies, spectator bee colonies and scout bee colonies. The employed bee colonies are responsible for searching near a food source to find a better food source, the spectator bee colonies perform enhanced search around the food source with better fitness value, and the scout bee colonies perform random search when a food source is exhausted. By combining the bee foraging search mechanism with the particle swarm algorithm, a new bee foraging learning particle swarm algorithm is proposed. BFL divides the original particle swarm optimization algorithm into three stages, namely the employed learning stage, the spectator learning stage and the scout learning stage. The employed learning stage is equivalent to the optimization stage of the traditional particle swarm optimization algorithm. The spectator learning stage further performs enhanced search around the particles with better fitness value, and the scout learning initializes the particles whose fitness value has not improved after a certain number of iterations.

[0173] During the employment learning phase, the update status of each particle will be recorded, among which,

[0174] If the fitness value of the hth particle is not updated after the iteration, its counter count(h) is incremented by one;

[0175] If the fitness value of the hth particle is updated after the iteration, its counter count(h) is set to zero;

[0176] The particle position and velocity update formula is: ;

[0177] X h (t+1)=X h (t)+V h (t+1);

[0178] ; Where V h (t+1) is the velocity of the hth particle after the t+1th iteration, is the inertia weight, c 1 is the individual learning factor of each particle, c 2 is the social learning factor of the particle group, r 1 and r 2 is a random number that follows a uniform distribution in the interval [0,1], P Besth is the historical optimal position of the hth particle, G Best is the best historical position of all particles, count(h) and x h-new are the counter value of the hth particle and the new particle value, respectively, f(x h-new ) is the h The fitness of the new particle value of the particle.

[0179] In the spectator learning phase, the fitness value is calculated based on the optimal position of each particle, and the probability of each particle being selected is calculated;

[0180] The fitness value calculation formula is: ; In the formula, F(x h ) is the fitness value of the hth particle;

[0181] The probability formula for a particle to be selected is: ;

[0182] In the formula, p h is the probability of the hth particle being selected, and N is the total number of particles. According to the probability of the particle being selected, the roulette strategy is used to select the particle, where

[0183] When the hth particle is selected, the hiring learning phase is performed on the particle alone, and the particle's position and speed update formula in the hiring learning phase is used to update the particle's speed and position. If the updated position is better than the original one, the particle's counter count(h) is reset to zero, otherwise it is incremented by 1. In the reconnaissance learning phase, if the hth particle's fitness value does not improve after the preset number of iterations, the particle's counter reaches the upper limit of the count, and the particle's position X is reset to zero. h and speed V h Perform initialization processing.

[0184] like Figure 2 As shown, the present invention provides a puncture needle path planning system based on an adaptive local optimization algorithm, which is applied to a puncture needle path planning method based on an adaptive local optimization algorithm. The system includes:

[0185] A data acquisition unit, used to obtain global puncture environment information and local puncture environment information;

[0186] A data processing unit, used for performing kinematic analysis on the puncture needle, constructing a kinematic model of the puncture needle for planning the puncture path of the puncture needle, and constructing a kinematic model of the puncture path of the puncture needle;

[0187] The kinematic model of the puncture needle puncture path reaching each target point is solved in turn by using an optimization algorithm to obtain a global puncture needle puncture path for puncturing each target point one by one;

[0188] The mapping relationship between the global puncture path of the puncture needle and the global puncture environment information is encapsulated by a neural network to obtain a global planning model for obtaining the global puncture path according to the global puncture environment information;

[0189] Performing transfer learning on the global planning model to obtain a local planning model for obtaining a local puncture path according to local puncture environment information;

[0190] The path planning unit is used to plan the global puncture path of the puncture needle in the global puncture environment based on the global planning model, and to locally adjust the global puncture path of the puncture needle according to the local puncture environment using the local planning model.

[0191] The method for constructing the global planning model by the data processing unit includes:

[0192] The global environment information of the puncture is used as the input item of the neural network, and the solution result Q_result of the kinematic model Q of the global puncture path obtained by the optimization algorithm according to the global environment information of the puncture is used as the output item of the neural network;

[0193] The mapping relationship between the puncture global environment information and the puncture global path is trained by using the neural network between the input items and the output items of the neural network, so as to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information;

[0194] The global planning model is:

[0195] Q_resulto=neural_network(G);

[0196] Where Q_resulto is the global puncture path output by the global planning model, G is the global puncture environment information, and neural_network is the neural network.

[0197] The method for constructing the local planning model by the data processing unit includes:

[0198] Extract the path segments with the same position as the puncture local environment information from the solution result Q_result of the kinematic model of the puncture global path as the puncture local path Q_m; use the puncture local environment information as the migration input item of the global planning model, and use the puncture local path Q_m as the migration output item of the global planning model;

[0199] Transfer the mapping relationship between the puncture global environment information and the puncture global path to the mapping relationship between the puncture local environment information and the puncture local path, and obtain a local planning model for obtaining the puncture local path according to the puncture local environment information.

[0200] The local planning model is: Q_mo=neural_network(g);

[0201] Where Q_mo is the local puncture path output by the local planning model, g is the local puncture environment information, and neural_network is the neural network;

[0202] The loss function of global planning model transfer learning is:

[0203] L=||(Q_mo+Q_mf+Q_mb)-Q_resulto|| 1 +||Q_mo-Q_m || 1 ; Where L is the loss function value, Q_mf is the path at the front end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, Q_mb is the path at the back end of the puncture local path Q_m in the puncture global path Q_resulto output by the global planning model, ||(Q_mo+Q_mf+Q_mb)-Q_resulto || 1 and ||Q_mo-Q_m|| 1 The neural network in the present invention can use BP neural network, which is a concept proposed by scientists led by Rumelhart and McClelland. It is a multi-layer feedforward neural network trained according to the error back propagation algorithm and is one of the most widely used neural network models.

[0204] BP neural network has the ability to classify patterns of any complexity and excellent multi-dimensional function mapping ability. It learns certain rules through its own training and obtains the result closest to the expected output value when the input value is given, that is, it determines the mapping relationship between input and output. The BP algorithm includes two processes: forward propagation of signals and reverse propagation of errors. That is, when calculating the error output, it is done from input to output, while adjusting the weights and thresholds from output to input. During forward propagation, the input signal acts on the output node through the hidden layer, and after nonlinear transformation, the output signal is generated. If the actual output does not match the expected output, the error reverse propagation process is entered. Error back propagation is to propagate the output error through the hidden layer to the input layer layer by layer, and distribute the error to all units in each layer, and use the error signal obtained from each layer as the basis for adjusting the weights of each unit. By adjusting the connection strength between the input node and the hidden layer node, the connection strength between the hidden layer node and the output node, and the threshold, the error is reduced along the gradient direction. After repeated learning and training, the network parameters (weights and thresholds) corresponding to the minimum error are determined, and the training stops. At this point, the trained neural network can process the input information of similar samples and obtain the output information by itself after nonlinear transformation with the smallest output error.

[0205] By utilizing the function mapping property of the BP neural network, the present invention applies the BP neural network to path planning, trains the BP neural network to learn the mapping relationship between environmental information and the puncture path, and then puts it into use during the puncture process.

[0206] The present invention uses a neural network to encapsulate the mapping relationship between the global puncture path of the puncture needle obtained by the optimization algorithm and the global puncture environment information, and obtains a global planning model for obtaining the global puncture path according to the global puncture environment information, and transfers the learning to a local planning model for obtaining the local puncture path according to the local puncture environment information, thereby realizing real-time local adjustment of the preoperative puncture path planning to fully adapt to the actual puncture scene and improve the robustness of the puncture path planning.

[0207] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present application. The protection scope of the present application is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present application within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present application.

Claims

1. A puncture needle path planning system based on an adaptive local optimization algorithm, characterized in that: The method for puncture needle path planning based on the adaptive local optimization algorithm comprises the following steps: Performing kinematic analysis on the puncture needle, constructing a kinematic model of the puncture needle for puncture path planning of the puncture needle, and constructing a kinematic model of the puncture path of the puncture needle; The kinematic model of the puncture needle puncture path reaching each target point is solved in turn by using an optimization algorithm to obtain a global puncture needle puncture path for puncturing each target point one by one; The mapping relationship between the puncture needle global path and the puncture global environment information is encapsulated by a neural network to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information; Performing transfer learning on the global planning model to obtain a local planning model for obtaining a local puncture path according to local puncture environment information; The construction method of the local planning model includes: Solution results of the kinematic model of the puncture needle puncture path Extract the paths with the same position as the local environment information of the puncture as the local puncture path ; The local environment information of the puncture is used as the migration input of the global planning model, and the local path of the puncture is As the migration output of the global planning model; The global planning model is used to transfer the mapping relationship between the global puncture environment information and the global puncture path to the mapping relationship between the local puncture environment information and the local puncture path based on the setting of the transfer input item and the transfer output item, so as to obtain a local planning model for obtaining the local puncture path according to the local puncture environment information; The local planning model is: ; In the formula, is the local puncture path output by the local planning model. To puncture local environmental information, for neural networks; The loss function of global planning model transfer learning is: ; In the formula, is the loss function value, The puncture global path output by the global planning model The path at the front end of the local puncture path, The puncture global path output by the global planning model Located in the local puncture path The path to the backend, and All are L1 norm forms.

2. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 1, characterized in that: The method for constructing the kinematic model of the puncture needle comprises: Set the forward speed of the puncture needle to The angular velocity of the axial rotation of the puncture needle is , the needle body coordinate system of the puncture needle is , the puncture needle is in the world coordinate system The position and posture of , where The puncture needle is in the world coordinate system Identifiers of the position and attitude matrices in , The puncture needle revolves around the world coordinate system of The rotation angle of the axis, The puncture needle revolves around the world coordinate system of The position coordinates of the axis, is the transpose operator; in the world coordinate system In the experiment, the speed of the puncture needle is determined as , where The puncture needle is in the world coordinate system The speed of In the needle coordinate system In the experiment, the speed of the puncture needle is determined as ,in , , where The puncture needle is in the needle body coordinate system The speed of Based on the world coordinate system and needle body coordinate system , the velocity conversion equation of the puncture needle is obtained as: ,in, , , In the formula, The world coordinate system and needle body coordinate system The rotation matrix that establishes the world coordinate system of the puncture needle and needle body coordinate system The speed transformation relationship between is the matrix coefficient of the Euler kinematics equation for rigid body fixed-point motion, which establishes the world coordinate system and needle body coordinate system The angular velocity transformation relationship between them; according to , as well as , the kinematic model of the puncture needle is determined as: ; Separate the time factor in the puncture needle kinematic model to eliminate the puncture speed and rotation speed in the control quantity of the puncture needle motion model and retain the needle insertion distance and rotation angle , and the kinematic model of the puncture needle after separation is obtained .

3. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 2, characterized in that: The method for constructing the kinematic model of the puncture needle puncture path includes: Kinematic model of puncture needle Apply the stop-and-turn strategy to obtain the control sequence of the puncture needle puncture path ,in, is the identifier of the control sequence, , and are the lengths of the first, second and nth segments of the puncture path of the puncture needle, , and The first, second and The rotation angle at the segment path; Control sequence based on puncture needle puncture path , and the tip of the puncture needle moves along a fixed radius Make standard circular motion to determine the puncture path of the puncture needle. Segment arc path composition, The first The length of the segment path, and ,in, for The corresponding arc, is the moving radius of the needle tip; The control sequence of the puncture needle puncture path is changed from the needle insertion distance and rotation angle The kinematic model of the puncture path of the puncture needle is obtained by converting the coordinates of the center of the path and the coordinates and length of the starting point of each path segment. , where is the kinematic model identifier of the puncture needle puncture path, For the The starting coordinates of the segment path, For the Segment path starting point coordinates The tangent vector at For the The coordinates of the center of the arc of the segment path; where: The relationship between is: ; ; ; In the formula, For the The starting coordinates of the segment path, For the Segment path starting point coordinates The tangent vector at For the The coordinates of the center point of the segment path arc.

4. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 3, characterized in that: The method for obtaining the puncture global environment information comprises: Using medical imaging equipment to collect medical images of the puncture environment; The position information of the puncture needle entry point, the target point and obstacles is obtained in the medical image as the puncture global environment information.

5. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 4, characterized in that: The method for obtaining the puncture local environment information comprises: Using the medical image to perform three-dimensional reconstruction to obtain a puncture environment model; The SLAM system loaded on the front end of the puncture needle moves along the global puncture path located on the puncture environment model to obtain obstacle information on the global puncture path as the local puncture environment information.

6. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 5, characterized in that: The method for constructing the global planning model includes: The global puncture environment information is used as the input item of the neural network, and the kinematic model of the puncture path of the puncture needle obtained by the optimization algorithm is The solution result of As the output of the neural network; The mapping relationship between the puncture global environment information and the puncture global path is trained by using the neural network based on the setting of the neural network input items and the neural network output items, so as to obtain a global planning model for obtaining the puncture global path according to the puncture global environment information; The global planning model is: ; In the formula, is the puncture global path output by the global planning model, To puncture the global environment information, For neural network.

7. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 1, characterized in that: The method for solving the kinematic model of the puncture path of the puncture needle reaching each target point in turn by using an optimization algorithm includes: Sort each target point from large to small according to the coordinate value in the z-axis direction to obtain the path planning sequence of the target points; Using an optimization algorithm based on a multi-objective function to plan the puncture path of the first target point in the path planning sequence of the target points, the puncture path of the puncture needle is obtained; After the initial path planning of the puncture needle is completed, the optimization algorithm will be used to perform a new planning operation on the second target point based on the multi-objective function starting from the end point of the initial path of the puncture needle. If the puncture path to the second target point cannot be planned at the end point of the initial puncture path of the puncture needle, the puncture needle tip starts to retreat to the position where the needle stopped and turned last on the current path, and the puncture path is planned for the second target point again, and so on, until the path to the second target point is successfully planned; If the path planning for the second target point cannot be successfully performed until the puncture needle tip retracts to the needle insertion point, a failure prompt will be given.

8. The puncture needle path planning system based on the adaptive local optimization algorithm according to claim 7, characterized in that: The multi-objective function is: ; ; ; ; ; In the formula, is the value of the multi-objective function, is the weight of the path deviation, is the weight of the total path length, is the error function between the target point and the needle tip of the puncture needle, for The determined endpoint coordinates of the puncture needle puncture path, are the coordinates of the target point, is the path length function, It is the first The length of the segment path, is a function to determine whether the needle tip of the puncture needle collides with an obstacle. is the number of obstacles, For the The collision function of obstacles, is the safe distance between the obstacle and the needle tip of the puncture needle. When the needle tip collides with the obstacle, The value is +∞, otherwise The value is zero.

Images