A radial artery puncture robot system

Abstract

The present invention discloses a radial artery puncture robot system, comprising: a puncture robot, an arm fixing platform, a B-ultrasound image acquisition device, a robot controller, and a control platform; the puncture robot comprises a three-axis slide, a puncture robotic arm, a puncture needle, and a puncture pressure sensor; the control platform obtains coordinated control instructions for the puncture robotic arm and the three-axis slide based on continuous radial artery cross-sectional ultrasound images acquired by the B-ultrasound image acquisition device, posture data of the puncture needle during puncture, and motion data of the three-axis slide and the puncture robotic arm fed back by the robot controller, and transmits them to the robot controller; thereby enabling the three-axis slide of the puncture robot to operate in coordination with the puncture robotic arm to complete puncture and withdrawal of the puncture needle, thereby achieving coordinated control of the robotic arm and the three-axis slide.

Description

Technical Field

[0001] The present invention relates to the field of medical devices, and in particular to a radial artery puncture robot system. Background Art

[0002] Radial artery puncture is a common procedure in intensive care units, typically used to collect blood samples for arterial blood gas analysis or for continuous blood pressure monitoring in critically ill patients. Currently, radial artery puncture is performed by medical personnel, who first palpate the pulse to determine the artery's location before proceeding. Preliminary studies have shown a relatively low success rate for radial artery puncture, averaging between 60% and 70%. Without specialized training or expertise, the success rate can drop as low as 53%. Traditional palpation techniques often require repeated punctures, particularly for children or patients with hypotension, edema, and obesity. As puncture failures increase, vasospasm or subcutaneous hematoma can cause the pulse to weaken or even disappear, making the next puncture more difficult. Repeated punctures not only waste puncture supplies, increase treatment costs and the financial burden on patients, but also create additional workload and psychological stress for medical staff. Therefore, automating radial artery puncture and reducing puncture complications are crucial for alleviating both patient injury and medical staff workload.

[0003] With the advancement of artificial intelligence and robotics, researchers have begun to explore the use of robots to perform vascular puncture tasks. Vascular puncture tasks place high demands on the flexibility and accuracy of robot control, mainly involving two key technologies: path planning and collaborative control. Patents with authorization number CN107775632B and application number CN112604094A disclose two types of intravenous puncture robots, focusing on contributions in mechanical devices and structural design. Patent application number CN111968097A relates to a vascular puncture image processing method and a vascular puncture robot, focusing on contributions in vascular segmentation and puncture angle determination based on image recognition. Patent authorization number CN112022294B relates to a method for planning the operation trajectory of an intravenous puncture robot based on ultrasound image guidance, but focuses on contributions in the robot operation process, and does not specifically introduce the robot path planning and collaborative control methods. Therefore, the existing path planning and collaborative control technologies of vascular puncture robots need to be further improved and enhanced.

[0004] Unlike the superficial vein puncture described in the above patents, the radial artery is deeper under the skin, and radial artery puncture is more susceptible to interference from factors such as hypotension, edema, and obesity. Experience has shown that compared with superficial vein puncture, the arterial puncture robot requires higher flexibility and accuracy in path planning and control. The patent with application number CN115708712A provides a radial artery positioning assistance device and a radial artery puncture positioning assistance system. The device does not have the function and potential for autonomous puncture and can only provide assistance for manual puncture by medical staff. Patents with authorization numbers CN109044498B, CN109602497B, and CN116236288B respectively disclose three different puncture robot systems based on robotic arms. However, the robot structure design of these inventions is not compact enough and occupies a large area. In addition, large puncture robotic arms are usually prone to cause strong fear and psychological pressure on patients. Summary of the Invention

[0005] In order to address the deficiencies in the background technology, the present invention provides a radial artery puncture robot system, which can solve the problem of the inability to coordinate control of a robotic arm and a three-axis slide rail.

[0006] A radial artery puncture robot system includes: a puncture robot, an arm fixing platform, a B-ultrasound image acquisition device, a robot controller and a control platform; the puncture robot includes a three-axis slide rail, a puncture robot arm, a puncture needle, and a puncture pressure sensor;

[0007] The three-axis slide is used to coordinate the movement of the puncture robot arm;

[0008] The base of the puncture robot arm is fixed to one end of the Z-axis of the three-axis slide rail and is used to control the puncture needle to complete the puncture task;

[0009] The puncture pressure sensor is used to monitor the force applied to the puncture needle and send the collected data to the control platform;

[0010] The puncture needle serves as the end effector of the puncture robotic arm and is connected to the other end of the puncture robotic arm through a puncture pressure sensor;

[0011] The arm fixation table is used to fix the patient's forearm in a preset position;

[0012] The B-ultrasound image acquisition device is used to continuously acquire a cross-sectional ultrasound image of the radial artery of the patient's arm fixed on the arm fixing table, and an image of the position of the puncture needle in the radial artery during the puncture action, and transmit the images to the control platform;

[0013] The control platform is used to process continuous radial artery cross-sectional ultrasound images, puncture needle posture data, and motion data of the three-axis slide and puncture manipulator fed back by the robot controller, and obtain coordinated control instructions for the puncture manipulator and the three-axis slide and send them to the robot controller;

[0014] The robot controller is used to run the control program of the puncture pressure sensor driver, puncture robot arm and three-axis slide.

[0015] Furthermore, the control platform includes:

[0016] Image processing module: used to segment the radial artery in continuous cross-sectional ultrasound images;

[0017] 3D vascular reconstruction module: used to reconstruct the 3D structure of the radial artery wall based on the segmented radial artery image and determine the optimal puncture posture data, where the puncture posture data includes the puncture position and puncture angle;

[0018] Robot collaborative control module: used to plan and obtain the path of the robot arm based on the optimal puncture posture data; used to control the movement of the three-axis slide to cooperate with the puncture robot arm through the three-axis slide control strategy, so that the movement of the three-axis slide actively cooperates with the motion control of the robot arm;

[0019] Puncture force analysis module: used to monitor and record the force data of the needle tip passing through the skin, subcutaneous tissue and blood vessel wall during puncture and withdrawal based on the data collected by the puncture pressure sensor, assisting in detecting whether the needle tip has penetrated the arterial wall and tracking the force applied to the needle tip;

[0020] Radial artery puncture module: It is used to perform puncture based on the determined optimal puncture position and puncture angle. It combines the real-time acquired B-ultrasound images and data collected by the puncture pressure sensor to control the puncture robot arm to accurately achieve blood vessel puncture and withdrawal.

[0021] Emergency brake module: used to urgently stop the robot's movement and puncture operation.

[0022] Furthermore, the radial artery puncture module includes a high-precision puncture control submodule, a puncture point positioning submodule, and a puncture detection and needle tip tracking submodule;

[0023] High-precision puncture control submodule: This module provides submillimeter motion control for the robotic arm to manipulate the puncture needle during vascular puncture. Specifically, it provides high-precision control of the robotic arm joints during puncture and withdrawal, thereby achieving subtle needle tip displacement and rotation.

[0024] Puncture point positioning submodule: used to determine the puncture position and angle based on 3D radial artery reconstruction, so that the puncture position is located on the central axis of the radial artery and the puncture action is performed at a preset tilt angle;

[0025] Puncture detection and needle tip tracking submodule: used to monitor the position of the puncture needle tip based on the puncture force monitoring data and B-ultrasound images, and whether the puncture needle tip has broken through the blood vessel wall.

[0026] Furthermore, the specific steps for reconstructing the 3D structure of the radial artery wall are as follows:

[0027] S01: The 3D vascular reconstruction module converts the segmented 2D radial artery wall pixel positions into 3D point cloud coordinates in the 3D Cartesian coordinate system of the robot system based on the radial artery subcutaneous depth distance provided by the segmented ultrasound image for each frame of the B-ultrasound image processed by the image processing module;

[0028] S02: Based on the continuous posture data of the ultrasound probe during the B-ultrasound image acquisition process, including the position and angle of the ultrasound probe, the generated continuous multi-frame 3D point clouds are stitched into a tubular 3D radial artery wall.

[0029] Furthermore, the process of determining the optimal puncture posture data is as follows:

[0030] S031: calculating the centroid pixel of the blood vessel for the segmented blood vessel wall in each frame of the ultrasound image, and determining the 3D coordinates of the centroid using a 3D Cartesian coordinate system based on the subcutaneous depth distance of the centroid pixel provided by the segmented ultrasound image;

[0031] S032: connecting the 3D coordinates of the centroid of the blood vessels in the continuous multiple frames of images into the central axis of the blood vessels according to the continuous posture of the ultrasound probe during the B-ultrasound image acquisition process;

[0032] S033: Projecting the central axis of the blood vessel upward onto the blood vessel wall to determine candidate puncture locations on the blood vessel wall, and randomly selecting a candidate puncture location on the end of the blood vessel wall near the palm as the optimal puncture location; specifically, randomly selecting refers to sampling on the projection line of the central axis onto the blood vessel wall;

[0033] S034: Determine the optimal puncture angle along the direction of the central axis of the blood vessel, and at the same time, tilt the puncture needle downward at a preset angle relative to the horizontal plane where the central axis is located, wherein the preset range of the tilt angle is 15° to 30°.

[0034] Furthermore, based on the optimal posture determined by the 3D vascular reconstruction module, the specific process of the robot collaborative control module obtaining the movement path of the end of the robotic arm is as follows:

[0035] A graph search algorithm or a sampling-based path planning algorithm is used for path planning to obtain the optimal path that the end effector of the robotic arm takes to move from the starting position to the target position, where the optimal path consists of a series of ordered three-dimensional coordinate points in a Cartesian coordinate system; a series of intermediate sub-targets are obtained by sampling the coordinate points in the optimal path at equal intervals at a preset interval distance L; the yaw angle, pitch angle, and roll angle of the series of intermediate sub-targets in the Euler coordinate system are calculated, which is the orientation of the end effector of the robotic arm when it passes through each intermediate sub-target.

[0036] Furthermore, in the Euler coordinate system, the yaw angle, pitch angle, or roll angle of the intermediate sub-target is calculated as follows:

[0037] Y i+1 =Y i +ΔY, i=0,1,2,3,...,n-1,

[0038]

[0039] Where Y0 represents the yaw angle, pitch angle, or roll angle of the starting posture g0; Y n represents the puncture target posture g n The yaw angle, pitch angle, or roll angle; Y i+1 Indicates the yaw angle, pitch angle, or roll angle of the i+1th sub-target; Y i represents the yaw angle, pitch angle, or roll angle of the i-th sub-target; ΔY represents the change in the yaw angle, pitch angle, or roll angle between two adjacent sub-targets; π represents the pi constant.

[0040] Furthermore, the neural network reinforcement learning training process of the three-axis slide control strategy is as follows:

[0041] S001: Obtain observation data of the puncture robot at the current sub-target gn; the observation data includes the states of the joints of the puncture robot arm, the position and speed of the puncture needle, the sub-target position and posture, the position and speed of the three-axis slide, and the movement of the three-axis slide at the last moment;

[0042] S002: According to the current state of each joint of the puncture robot arm and the sub-target posture, the inverse kinematics solver is used to solve the motion speed v of each joint of the robot arm joint , IK_Fail, and judge whether the solution is successful: if the solution is successful, the flag bit IK_Fail is set to 0; if the solution fails, the flag bit IK_Fail is set to 1;

[0043] S003: Take the movement speed of each joint of the puncture robot arm, the state of each joint of the puncture robot arm, the position of the puncture needle, the speed of the puncture needle, the position of the sub-target, the position of the three-axis slide, the speed of the three-axis slide, and the action of the three-axis slide at the last moment as the input of the three-axis slide control neural network, and output the movement action of the three-axis slide

[0044] S004: When the flag IK_Fail is set to 0, the movement of each joint of the robot arm and the movement of the three-axis slide are executed simultaneously, and the reward after the three-axis slide executes the movement is calculated;

[0045] S005: Based on the calculated rewards, the neural network model of the three-axis slide control strategy is optimized using the reinforcement learning algorithm until the expected total future rewards is maximized, and the sub-goal is updated to gn +1 , return to S001 until the sub-goal is updated to the final puncture posture g N .

[0046] Furthermore, the reward calculation formula in S004 is:

[0047] r(s t , a t )=r IK +αr acc ,

[0048]

[0049]

[0050] r acc =||a t -a t-1 || 2

[0051] Among them, s t is the state space of the robot; a t is the motion of the three-axis slide at time t; r IK is the error between the desired position of the puncture needle and the actual position; r acc is the difference in acceleration between two adjacent moments of the triaxial slide; α and c rot is the weight coefficient; is the three-dimensional coordinate position that the puncture needle is expected to reach; is the three-dimensional coordinate position actually reached by the puncture needle; d rot The rotation distance between the actual position of the puncture needle and the expected position; The angle of the desired position of the puncture needle; is the angle of the actual position reached by the puncture needle; a t and at-1 are the motion accelerations of the three-axis slide at time t and t-1 respectively.

[0052] Beneficial effects

[0053] The present invention proposes a radial artery puncture robot system with the following advantages: adopting a priority control concept, taking the control of the robotic arm as a high priority and the three-axis slide as a low priority, emphasizing that the three-axis slide will perform corresponding coordinated movements only after knowing the movement intention of the robotic arm, ensuring that its own movement will not violate the movement intention of the robotic arm, and can improve the success rate of the radial artery puncture robot in performing tasks; adopting deep reinforcement learning technology to learn a three-axis slide control strategy with a certain degree of foresight, ensuring that the puncture movement of the three-axis slide and the robotic arm is carried out simultaneously, and the movement of the three-axis slide does not need to wait for the completion of the execution of the robotic arm components; it can replace nursing staff to perform radial artery puncture, improve the puncture success rate, reduce nursing workload, and save medical resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 This is a structural block diagram of a radial artery puncture robot system provided by an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of the structure of the puncture robot arm, puncture needle, and puncture pressure sensor provided by an embodiment of the present invention;

[0057] Figure 3 This is a schematic block diagram of a control platform provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] To make the objectives, technical solutions, and advantages of the present invention more apparent, the technical solutions of the present invention will be described in detail below. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other implementations obtained by those of ordinary skill in the art without inventive effort are within the scope of protection of the present invention.

[0059] Example 1

[0060] like Figure 1 As shown, a radial artery puncture robot system includes: a puncture robot, an arm fixing platform, a B-ultrasound image acquisition device, a robot controller and a control platform;

[0061] like Figure 2 As shown, the puncture robot includes a three-axis slide, a puncture robotic arm (the puncture robotic arm has multiple joints 1), a puncture needle 3, and a puncture pressure sensor 2; the three-axis slide is used to coordinate the movement of the puncture robotic arm; the base of the puncture robotic arm is fixed at one end of the Z-axis of the three-axis slide, and is used to control the puncture needle 3 to complete the puncture task; the puncture pressure sensor 2 is used to monitor the force on the puncture needle 3 and send the collected data to the control platform; the puncture needle 3 serves as the end effector of the puncture robotic arm and is connected to the other end of the puncture robotic arm through the puncture pressure sensor 2.

[0062] Preferably, the base of the puncture robot arm is fixed to the lower end of the Z axis of the three-axis slide rail.

[0063] Preferably, the puncture pressure sensor 2 is connected to the end of the robotic arm via a micro slide rail, which is used to telescopically control the position of the puncture needle 3 and the puncture pressure sensor 2 and can push the puncture needle 3 outward.

[0064] Preferably, the four pillars of the three-axis slide rail are heightened to a certain extent, and the degree of freedom of the robotic arm is greater than or equal to 4.

[0065] The arm fixation platform is used to fix the patient's forearm in a preset position; preferably, the arm fixation platform is provided with a first strap and a second strap, the first strap is used to fix the palm, and the second strap is used to fix the arm.

[0066] The B-ultrasound image acquisition device is used to continuously acquire a cross-sectional ultrasound image of the radial artery of the patient's arm fixed on the arm fixing table, and an image of the position of the puncture needle 3 in the radial artery during the puncture action, and send the image to the control platform;

[0067] The control platform is used to process the continuous radial artery cross-sectional ultrasound images, the puncture posture data of the puncture needle 3, and the motion data of the three-axis slide and the puncture manipulator fed back by the robot controller, and obtain the coordinated control instructions of the puncture manipulator and the three-axis slide and send them to the robot controller;

[0068] The robot controller is used to run the control program of the puncture pressure sensor driver, puncture robot arm and three-axis slide.

[0069] like Figure 3 As shown, the control platform includes:

[0070] Image processing module: used to segment the radial artery in continuous cross-sectional ultrasound images;

[0071] 3D vascular reconstruction module: used to reconstruct the 3D structure of the radial artery wall based on the segmented radial artery image and determine the optimal puncture posture data, where the puncture posture data includes the puncture position and puncture angle.

[0072] Robot collaborative control module: used to plan and obtain the path of the robotic arm based on the optimal puncture posture data; used to control the movement of the three-axis slide in conjunction with the puncture robotic arm through the three-axis slide control strategy, so that the movement of the three-axis slide actively coordinates with the motion control of the robotic arm;

[0073] Puncture force analysis module: used to monitor and record the force data of the puncture needle 3 when it passes through the skin, subcutaneous tissue and blood vessel wall during puncture and withdrawal based on the data collected by the puncture pressure sensor 2, to assist in detecting whether the needle tip has broken through the arterial wall and tracking the force applied to the needle tip;

[0074] Radial artery puncture module: used to perform puncture based on the determined optimal puncture position and puncture angle, and combined with the real-time acquired B-ultrasound images and data collected by the puncture pressure sensor 2, the puncture robot arm controls the puncture needle 3 to accurately achieve blood vessel puncture and withdrawal;

[0075] Emergency brake module: used to urgently stop the robot's movement and puncture operation.

[0076] More specifically, the radial artery puncture module includes a high-precision puncture control submodule, a puncture point positioning submodule, and a puncture detection and needle tip tracking submodule;

[0077] High-precision puncture control submodule: This module is used to provide submillimeter motion control for the robotic arm to manipulate the puncture needle 3 to perform vascular puncture. Specifically, it provides high-precision control of the robotic arm joints during the puncture and withdrawal process, thereby achieving subtle needle tip displacement and rotation.

[0078] Puncture point positioning submodule: used to determine the puncture position and angle based on 3D radial artery reconstruction, so that the puncture position is located on the central axis of the radial artery and the puncture action is performed at a preset tilt angle;

[0079] Puncture detection and needle tip tracking submodule: used to monitor the position of the needle tip of the puncture needle 3 and whether the needle tip of the puncture needle 3 breaks through the blood vessel wall based on the puncture force monitoring data and B-ultrasound images.

[0080] Furthermore, the specific steps for reconstructing the 3D structure of the radial artery wall are as follows:

[0081] S01: The 3D vascular reconstruction module converts the segmented 2D radial artery wall pixel positions into 3D point cloud coordinates in the 3D Cartesian coordinate system of the robot system based on the radial artery subcutaneous depth distance provided by the segmented ultrasound image for each frame of the B-ultrasound image processed by the image processing module;

[0082] S02: Based on the continuous posture data of the ultrasound probe during the B-ultrasound image acquisition process, including the position and angle of the ultrasound probe, the generated continuous multi-frame 3D point clouds are stitched into a tubular 3D radial artery wall.

[0083] The specific process of determining the optimal puncture posture data is as follows:

[0084] S031: calculating the centroid pixel of the blood vessel for the segmented blood vessel wall in each frame of the ultrasound image, and determining the 3D coordinates of the centroid using a 3D Cartesian coordinate system based on the subcutaneous depth distance of the centroid pixel provided by the segmented ultrasound image;

[0085] S032: connecting the 3D coordinates of the centroid of the blood vessels in the continuous multiple frames of images into the central axis of the blood vessels according to the continuous posture of the ultrasound probe during the B-ultrasound image acquisition process;

[0086] S033: Projecting the central axis of the blood vessel upward onto the blood vessel wall to determine candidate puncture locations on the blood vessel wall, and randomly selecting a candidate puncture location on the end of the blood vessel wall near the palm as the optimal puncture location; in specific implementation, the random selection is not a random selection, but a sampling is performed on the projection line of the central axis onto the blood vessel wall;

[0087] S034: Determine the optimal puncture angle along the direction of the central axis of the blood vessel, and tilt the puncture needle 3 downward relative to the horizontal plane of the central axis by a preset angle. In this embodiment, the preset range of the tilt angle is 15° to 30°.

[0088] More specifically, based on the optimal posture determined by the 3D vascular reconstruction module, the specific process of the robot collaborative control module obtaining the movement path of the end of the robotic arm is as follows:

[0089] A graph search algorithm or a sampling-based path planning algorithm is used for path planning to obtain the optimal path for the robot end effector to move from the starting position to the target position. The optimal path consists of a series of ordered three-dimensional coordinate points in a Cartesian coordinate system. By sampling the coordinate points in the optimal path at equal intervals at a preset interval distance L, a series of intermediate sub-targets are obtained. To determine the orientation of the robot end effector as it passes through each sub-target, the yaw angle, pitch angle, and roll angle describing its orientation are sequentially assigned to the series of sub-targets in the Euler coordinate system. The specific details are as follows:

[0090] In order to reduce the pressure of solving the inverse kinematics of the robot arm and improve the success rate of motion planning. Let the end effector of the robot arm (puncture needle tip) pass through n-1 intermediate sub-target postures in sequence and finally reach the specified puncture posture. Use a graph search algorithm (such as A*) or a sampling-based path planning algorithm (such as RRT*) to determine the position of an intermediate sub-target at each interval distance L. The setting of distance L should ensure the reachability of the end effector of the robot arm between two adjacent sub-targets. In the Euler coordinate system, in the graph search algorithm and the sampling-based path planning algorithm, the yaw angle distribution of the intermediate sub-targets is announced as follows:

[0091] Y i+1 =Y i +ΔY, i=0,1,2,3,...,n-1,

[0092]

[0093] Among them, Y0 represents the yaw angle of the starting posture g0; Y n represents the puncture target posture g n The yaw angle; Y i+1 Indicates the yaw angle of the i+1th sub-target; Y i represents the yaw angle of the i-th sub-target; ΔY represents the change in yaw angle between two adjacent sub-targets; and π represents the circumference constant. The pitch and roll angles of intermediate sub-targets are assigned in the same manner as the yaw angle, so this will not be described in detail.

[0094] The coordinated control of the puncture robot's puncture arm and the three-axis slide can be defined as a goal-conditioned reinforcement learning problem. The control of the three-axis slide can be expressed as a partially observed Markov decision process under the deep reinforcement learning framework, consisting of a tuple (S, A, O, T (s t+1 |s t , a t ), P(o t |s t ), r(s t , a t ), γ) are defined, where S is the state space of the puncture robot; A is the action space of the three-axis slide; O is the observation space; T is the state transfer function; P is the observation probability; r is the reward; γ is the discount factor. Under the setting of target-conditioned reinforcement learning, the three-axis slide control strategy π(a t |o t , g) The goal is to be in the target distribution Next, find a sequence of actions for a three-axis slide that makes the expected sum of future rewards Maximize, where g is the sub-target point; observation space o t ∈O is composed of the joint states s of the robot arm joint, the pose p of the end effector ee , the speed of the end effector v ee , sub-target pose g, pose data of the three-axis slide (including position and angle) p s ={p x , p y , p z}、Speed ​​v of the three-axis slide s And the action a of the three-axis slide at time t-1 t-1 Composition. Action space a t ∈A is composed of the motion speeds of the three axes of X, Y, and Z of the three-axis slide, that is, s t ∈S is the state space of the robot, P(o t |s t ) is the observation probability, T(s t+1 |s t , a t ) is the state transition function.

[0095] The three-axis slide control strategy requires N rounds of reinforcement learning training until the rewards obtained converge. The specific neural network reinforcement learning training process of the three-axis slide control strategy is as follows:

[0096] S001: Get the puncture robot's current sub-goal g n Observation data; wherein the observation data includes the joint states s of the puncture robot arm joint , the position of the puncture needle p ee , the speed of the puncture needle v ee , sub-target pose , the position p of the three-axis slide s ={p x , p v , p z}、Speed ​​v of the three-axis slide s 、The action of the three-axis slide at the last moment a t-1 ;

[0097] S002: According to the current state of each joint of the puncture robot arm joint And the sub-target posture g1, the inverse kinematics solver is used to solve the motion velocity v of each joint of the manipulator joint , IK_Fail, and judge whether the solution is successful: if the solution is successful, the flag bit IK_Fail is set to 0; if the solution fails, the flag bit IK_Fail is set to 1;

[0098] S003: Set the movement speed v of each joint of the puncture robot arm joint , IK_Fail, each joint state of the puncture robot arm s joint , the position of the puncture needle pee , the speed of the puncture needle v ee , sub-target pose The position p of the three-axis slide s ={p x , p y , p z}、Speed ​​v of the three-axis slide s 、The action of the three-axis slide at the last moment a t-1 As the input of the three-axis slide control neural network, it outputs the motion of the three-axis slide

[0099] S004: When the flag IK_Fail is set to 0, the movement of each joint of the robot arm and the movement of the three-axis slide are executed simultaneously, and the reward after the three-axis slide executes the movement is calculated;

[0100] S005: Based on the calculated rewards, the reinforcement learning algorithm is used to optimize the neural network model of the three-axis slide control strategy until the expected total future rewards are maximized, and the sub-goal is updated to g n+1 , return to S001 until the sub-goal is updated to the final puncture posture g N .

[0101] Furthermore, the reward calculation formula in S004 is:

[0102] r(s t , a t )=r IK +αr acc ,

[0103]

[0104]

[0105] r acc =||a t -a t-1 || 2

[0106] Among them, s t is the state space of the robot; a t is the motion of the three-axis slide at time t; r IK is the error between the desired position of the puncture needle and the actual position; r acc is the difference in acceleration between two adjacent moments of the triaxial slide; α and C rot is the weight coefficient; is the three-dimensional coordinate position that the puncture needle is expected to reach; is the three-dimensional coordinate position actually reached by the puncture needle 3; d rotis the rotation distance between the actual position reached by the puncture needle 3 and the expected position; The angle of the desired position of the puncture needle; is the angle of the posture actually reached by the puncture needle 3; a t and a t-1 are the motion accelerations of the three-axis slide at time t and t-1 respectively.

[0107] Preferably, the three-axis slide control strategy is trained using other reinforcement learning algorithms, and the three-axis slide control strategy uses other path planning algorithms to determine the sub-target pose.

Claims

1. A radial artery puncture robot system, characterized in that: include: Puncture robot, arm fixing platform, B-ultrasound image acquisition device, robot controller and control platform; the puncture robot includes a three-axis slide rail, a puncture robotic arm, a puncture needle, and a puncture pressure sensor; The three-axis slide is used to coordinate the movement of the puncture robot arm; The base of the puncture robot arm is fixed to one end of the Z-axis of the three-axis slide rail and is used to control the puncture needle to complete the puncture task; The puncture pressure sensor is used to monitor the force applied to the puncture needle and send the collected data to the control platform; The puncture needle serves as the end effector of the puncture robotic arm and is connected to the other end of the puncture robotic arm through a puncture pressure sensor; The arm fixation table is used to fix the patient's forearm in a preset position; The B-ultrasound image acquisition device is used to continuously acquire a cross-sectional ultrasound image of the radial artery of the patient's arm fixed on the arm fixing table, and an image of the position of the puncture needle in the radial artery during the puncture action, and transmit the images to the control platform; The control platform is used to process continuous radial artery cross-sectional ultrasound images, puncture needle posture data, and motion data of the three-axis slide and puncture manipulator fed back by the robot controller, and obtain coordinated control instructions for the puncture manipulator and the three-axis slide and send them to the robot controller; The robot controller is used to run the control program of the puncture pressure sensor driver, the puncture robot arm and the three-axis slide; The control platform includes a robot collaborative control module, which is used to control the movement of the three-axis slide in conjunction with the puncture robot arm through the three-axis slide control strategy, so that the movement of the three-axis slide is actively coordinated with the movement control of the robot arm; The neural network reinforcement learning training process of the three-axis slide control strategy is as follows: S001: Get the puncture robot's current sub-goal Observation data; wherein the observation data includes the state of each joint of the puncture robot arm, the position and posture of the puncture needle, the speed of the puncture needle, the position and posture of the sub-target, the position and posture of the three-axis slide, the speed of the three-axis slide, and the action of the three-axis slide at the last moment; S002: According to the current state of each joint of the puncture robot arm and the sub-target posture, the inverse kinematics solver is used to solve the motion speed of each joint of the robot arm , and judge whether the solution is successful: If the solution is successful, the flag is Set to 0; if the solution fails, the flag Set to 1; S003: Take the movement speed of each joint of the puncture robot arm, the state of each joint of the puncture robot arm, the position of the puncture needle, the speed of the puncture needle, the position of the sub-target, the position of the three-axis slide, the speed of the three-axis slide, and the action of the three-axis slide at the last moment as the input of the three-axis slide control neural network, and output the movement action of the three-axis slide ; S004: In the flag position When set to 0, the robot arm joint motions and the three-axis slide motions are executed simultaneously, and the rewards after the three-axis slide motions are calculated; S005: Based on the calculated rewards, the reinforcement learning algorithm is used to optimize the neural network model of the three-axis slide control strategy until the expected total future rewards are maximized, and the sub-goal is updated to , return to S001 until the sub-goal is updated to the final puncture posture .

2. The radial artery puncture robot system according to claim 1, characterized in that: The control platform includes: Image processing module: used to segment the radial artery in continuous cross-sectional ultrasound images; 3D vascular reconstruction module: used to reconstruct the 3D structure of the radial artery wall based on the segmented radial artery image and determine the optimal puncture posture data, where the puncture posture data includes the puncture position and puncture angle; Robot collaborative control module: also used to plan and obtain the path of the robotic arm based on the optimal puncture posture data; Puncture force analysis module: used to monitor and record the force data of the needle tip passing through the skin, subcutaneous tissue and blood vessel wall during puncture and withdrawal based on the data collected by the puncture pressure sensor, assisting in detecting whether the needle tip has penetrated the arterial wall and tracking the force applied to the needle tip; Radial artery puncture module: It is used to perform puncture based on the determined optimal puncture position and puncture angle. It combines the real-time acquired B-ultrasound images and data collected by the puncture pressure sensor to control the puncture robot arm to accurately achieve blood vessel puncture and withdrawal. Emergency brake module: used to urgently stop the robot's movement and puncture operation.

3. The radial artery puncture robot system according to claim 2, characterized in that: The radial artery puncture module includes a high-precision puncture control submodule, a puncture point positioning submodule, and a puncture detection and needle tip tracking submodule; High-precision puncture control submodule: used to provide submillimeter motion control for the robotic arm to manipulate the puncture needle to perform vascular puncture; Puncture point positioning submodule: used to determine the puncture position and puncture angle based on 3D radial artery reconstruction, so that the puncture position is located on the central axis of the radial artery and the puncture action is performed at a preset tilt angle; Puncture detection and needle tip tracking submodule: used to monitor the position of the puncture needle tip based on the puncture force monitoring data and B-ultrasound images, and whether the puncture needle tip has broken through the blood vessel wall.

4. The radial artery puncture robot system according to claim 2, characterized in that: The specific steps for reconstructing the 3D structure of the radial artery wall are: S01: The 3D vascular reconstruction module converts the segmented 2D radial artery wall pixel positions into 3D point cloud coordinates in the 3D Cartesian coordinate system of the robot system based on the radial artery subcutaneous depth distance provided by the segmented ultrasound image for each frame of the B-ultrasound image processed by the image processing module; S02: Based on the continuous posture data of the ultrasound probe during the B-ultrasound image acquisition process, including the position and angle of the ultrasound probe, the generated continuous multi-frame 3D point clouds are stitched into a tubular 3D radial artery wall.

5. The radial artery puncture robot system according to claim 2, characterized in that: The specific process of determining the optimal puncture posture data is as follows: S0031: Calculate the centroid pixel of the blood vessel wall segmented in each frame of the ultrasound image, and determine the 3D coordinates of the centroid using a 3D Cartesian coordinate system based on the subcutaneous depth distance of the centroid pixel provided by the segmented ultrasound image; S0032: Connecting the 3D coordinates of the centroid of the blood vessels in the consecutive multiple frames of images into the central axis of the blood vessels according to the continuous posture of the ultrasound probe during the B-ultrasound image acquisition process; S0033: Projecting the central axis of the blood vessel upward onto the blood vessel wall to determine candidate puncture positions on the blood vessel wall, and randomly selecting a candidate puncture position on the end of the blood vessel wall close to the palm as the optimal puncture position; S0034: Determine the optimal puncture angle along the direction of the central axis of the blood vessel, and at the same time, tilt the puncture needle downward at a preset angle relative to the horizontal plane where the central axis is located.

6. The radial artery puncture robot system according to claim 2, characterized in that: Based on the optimal posture determined by the 3D vascular reconstruction module, the robot collaborative control module plans the movement path of the robotic arm end as follows: A graph search algorithm or a sampling-based path planning algorithm is used for path planning to obtain the optimal path that the end effector of the robotic arm takes to move from the starting position to the target position, where the optimal path consists of a series of ordered three-dimensional coordinate points in a Cartesian coordinate system; a series of intermediate sub-targets are obtained by sampling the coordinate points in the optimal path at equal intervals at a preset interval distance L; the yaw angle, pitch angle, and roll angle of the series of intermediate sub-targets in the Euler coordinate system are calculated, which is the orientation of the end effector of the robotic arm when it passes through each intermediate sub-target.

7. The radial artery puncture robot system according to claim 6, characterized in that: In the Euler coordinate system, the yaw angle, pitch angle, or roll angle of the intermediate sub-target is calculated as follows: ； ； in, Indicates the starting position yaw angle, pitch angle, or roll angle; Indicates the yaw angle, pitch angle, or roll angle of the puncture target posture; Indicates the yaw angle, pitch angle, or roll angle of the i+1th sub-target; represents the yaw angle, pitch angle, or roll angle of the i-th sub-target; Indicates the change in the yaw angle, pitch angle, or roll angle of two adjacent sub-targets.

8. The radial artery puncture robot system according to claim 1, characterized in that: The reward calculation formula in S004 is: , , ， ； in, is the state space of the robot; is the motion of the three-axis slide at time t; The error between the desired position of the puncture needle and the actual position it reaches; is the difference in acceleration between two adjacent moments of the triaxial slide; and is the weight coefficient; is the three-dimensional coordinate position that the puncture needle is expected to reach; is the three-dimensional coordinate position actually reached by the puncture needle; The rotation distance between the actual position of the puncture needle and the expected position; The angle of the desired position of the puncture needle; is the angle of the position actually reached by the puncture needle; are the motion accelerations of the three-axis slide at time t and t-1 respectively.

Images