An ultrasound-guided intelligent indwelling needle puncture robot and a system thereof

Abstract

The application discloses an intelligent indwelling needle puncture robot under ultrasonic guidance and a system thereof, and belongs to the field of medical instruments, and comprises a multi-degree-of-freedom mechanical arm system, a high-definition ultrasonic imaging module, a central processing and control unit, a man-machine interaction interface and a safety monitoring and emergency unit; the multi-degree-of-freedom mechanical arm system comprises a probe control mechanical arm and a puncture execution mechanical arm, and the probe control mechanical arm is used for clamping and adjusting the position and posture of an ultrasonic probe; the application is characterized in that the high-definition ultrasonic imaging module is used for optimizing the image, an AI blood vessel recognition and segmentation module is combined to accurately recognize the blood vessel and obtain relevant parameters, an intelligent path planning module is used for planning an optimal path for avoiding dangerous areas, and the multi-degree-of-freedom mechanical arm system is used for fine operation, so that the problem of insufficient accuracy caused by poor blood vessel conditions in manual puncture is effectively solved, the probability of repeated puncture is reduced, and complications such as blood vessel injury and hematoma are reduced.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to an ultrasound-guided intelligent indwelling needle puncture robot and its system. Background Technology

[0002] Intravenous catheterization is one of the most basic and frequently performed invasive procedures in clinical medicine, serving as a crucial step in establishing intravenous access for drug infusion, blood collection, and nutritional support. Although this procedure seems routine, its success rate and safety are constrained by various factors in actual clinical practice, especially when dealing with specific challenging patient groups, where existing techniques and manual operation methods have significant limitations.

[0003] First, the success rate of puncture is limited by the patient's vascular condition and the operator's experience. In pediatric, oncology, and emergency departments, medical staff often encounter situations where blood vessels are small (e.g., the diameter of blood vessels in infants is only 2-3 mm), located deep within the vessel, or have collapsed or poorly filled vessels due to obesity, edema, or shock. Statistical data shows that for these patients with difficult vascular access (DIVA), the first-time success rate of manual blind puncture is often less than 60%. Repeated trial punctures not only cause patients significant physical pain and psychological fear, but also easily lead to complications such as endothelial damage, puncture site hematoma, phlebitis, and even nerve damage. In severe cases, the inability to establish intravenous access in time may delay rescue efforts.

[0004] Secondly, existing ultrasound-guided techniques have high operational barriers and insufficient stability. To improve the success rate of punctures in difficult vessels, ultrasound-guided intravenous puncture (USGIV) has been gradually introduced into clinical practice. However, traditional USGIV procedures require extremely high skill levels from medical personnel, demanding a high degree of hand-eye-brain coordination: one hand holds the ultrasound probe to find and maintain the optimal cutting plane, while the other hand holds the puncture needle for blind operation, and both eyes must keep a close eye on the two-dimensional image on the screen, constructing a three-dimensional spatial relationship in the brain. This non-direct-vision operation mode not only has a long training period (usually 6-12 months), but is also highly susceptible to human factors. For example, the average physiological tremor amplitude of the human hand is about 0.5-1 mm. During prolonged operation, fatigue, or mental stress, this tremor can cause probe slippage or needle tip deviation from the intended path, resulting in puncture failure.

[0005] Furthermore, there is a lack of standardized operating procedures and closed-loop control throughout the entire process. Different operators exhibit significant differences in needle insertion angle, speed, and depth control, making it difficult to establish unified clinical standards. Most existing auxiliary puncture devices are single-function devices, providing only simple mechanical fixation or image enhancement, and cannot achieve closed-loop automated control from "ultrasound image perception" to "intelligent decision-making" and then to "precise mechanical execution." In addition, some puncture robots have complex structures and lack the ability to track subtle patient movements in real time. If the patient shifts during the puncture process (such as reflexive twitching caused by breathing or pain), the robot cannot respond promptly, posing a significant safety hazard.

[0006] To address the aforementioned issues, the development of an intelligent puncture robot system that integrates high-definition ultrasound imaging, AI-powered intelligent recognition, path planning, and high-precision mechanical control is particularly urgent. This system needs to possess intelligent capabilities across the entire process of perception, decision-making, and execution to overcome the limitations of manual operation and improve puncture accuracy and efficiency. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent air conditioning, cooling, heating and temperature control system for tram traction converter boxes, in order to solve the problems pointed out in the background art.

[0008] This invention provides an ultrasound-guided intelligent indwelling needle puncture robot, comprising a multi-degree-of-freedom robotic arm system, a high-definition ultrasound imaging module, a central processing and control unit, a human-machine interface, and a safety monitoring and emergency unit.

[0009] The multi-degree-of-freedom robotic arm system serves as the execution entity, comprising a probe-controlling robotic arm and a puncture-execution robotic arm. The probe-controlling robotic arm is configured to hold and adjust the position and orientation of the ultrasound probe to ensure that the ultrasound imaging plane can completely cover and accurately locate the target blood vessel area. The puncture-execution robotic arm is configured to hold and drive the indwelling needle puncturist to perform precise puncture actions, including needle insertion, needle withdrawal, and multi-angle adjustment. The probe-controlling robotic arm and the puncture-execution robotic arm can move independently or be coordinated and controlled in the same coordinate system.

[0010] The high-definition ultrasound imaging module is used to acquire cross-sectional or longitudinal ultrasound video streams of the target vascular region in real time, providing deep imaging information including blood vessels, nerves and surrounding tissues.

[0011] The central processing and control unit includes an AI vessel recognition and segmentation module, an intelligent path planning module, and a motion control module. The AI ​​vessel recognition and segmentation module, based on a pre-trained deep learning model, identifies target vessels in the ultrasound video stream in real time, segments the vessel contour, and calculates the vessel depth, diameter, and centerline coordinates. The intelligent path planning module, based on the three-dimensional spatial information of the vessel, plans the optimal puncture path that avoids dangerous areas such as nerves and tendons and conforms to clinical puncture standards. The motion control module converts the puncture path into robotic arm motion commands, controlling the puncture needle to perform the needle insertion operation at a preset angle and speed.

[0012] The human-computer interaction interface includes a touch screen and operating components. The touch screen is used to display real-time ultrasound images, blood vessel contours, puncture paths and system status. The operating components include physical buttons and a foot switch, and have one-click blood vessel recognition, path confirmation, puncture start and emergency stop functions.

[0013] The safety monitoring and emergency unit includes a force sensing feedback module, an optical / electromagnetic positioning module, and a dual confirmation module; the force sensing feedback module is used to monitor the needle insertion resistance, the optical / electromagnetic positioning module is used to track the displacement of the patient's puncture site, and the dual confirmation module is used to reconfirm the blood vessel position before the puncture operation is performed.

[0014] Optionally, the high-definition ultrasound imaging module includes an ultrasound host and an image acquisition card. The image acquisition card communicates with the central processing and control unit via a PCIe interface. The central processing and control unit converts the ultrasound video stream acquired by the image acquisition card at a frame rate of not less than 25fps into digital image frames, and sequentially performs inter-frame denoising algorithms such as Gaussian filtering or median filtering, and image enhancement algorithms such as histogram equalization or edge enhancement on the digital image frames to improve the contrast between blood vessels and surrounding tissues.

[0015] Optionally, the AI ​​vessel recognition and segmentation module employs a pre-trained U-Net++ deep learning model. This model has been trained on no fewer than 10,000 clinical ultrasound vessel images, achieving a vessel recognition accuracy of no less than 95%, and can accurately identify target vessels including the basilic vein and cephalic vein. The loss function of the U-Net++ deep learning model is a weighted sum of Dice loss and binary cross-entropy loss, with the Dice loss weight coefficient ranging from 0.6 to 0.8. After training, the model achieves a Dice coefficient of no less than 0.95 for vessel segmentation and a vessel edge recognition error of ≤0.1mm.

[0016] Optionally, the calculation process of the intelligent path planning module includes:

[0017] Based on the three-dimensional coordinates of blood vessels output by the AI ​​blood vessel recognition and segmentation module, including the planar coordinates and depth values ​​of the blood vessel centerline, and taking the point on the blood vessel centerline closest to the skin as the target puncture point, the coordinates of the needle insertion point are calculated: the horizontal distance between the needle insertion point and the target puncture point is 1.5-2 times the blood vessel depth, and it is located on the extended line of the long axis of the blood vessel.

[0018] The AI ​​vessel recognition and segmentation module's calculation process for vessel depth, diameter, and centerline coordinates includes:

[0019] Blood vessel depth: Based on the depth calibration parameters of ultrasound images, the actual depth value corresponding to each pixel is preset, the pixel coordinates of the upper boundary of blood vessels in the segmentation mask are extracted and converted into the actual depth value based on the skin surface;

[0020] Blood vessel diameter: In the segmentation mask, samples are taken along the direction perpendicular to the direction of the blood vessel, the pixel distance between the two sides of the blood vessel at each sampling point is calculated, and the average value is taken as the blood vessel diameter after being converted into the actual size.

[0021] Centerline coordinates: The skeleton of the segmentation mask is extracted and a morphological thinning algorithm is used to obtain the pixel-level centerline. Combined with the planar coordinate calibration parameters of the ultrasound image, the actual planar distance corresponding to each pixel is preset and converted into actual three-dimensional coordinates. The actual three-dimensional coordinates are composed of planar coordinates and depth values.

[0022] Optionally, the needle insertion angle can be determined based on the depth and diameter of the blood vessel: when the blood vessel depth is ≤5mm, the needle insertion angle is 25°-30°, and when the depth is >5mm, the angle is 15°-20°, ensuring that the angle between the puncture needle path and the long axis of the blood vessel is ≤10°.

[0023] Based on a pre-set anatomical structure database (including the coordinate range of dangerous areas such as nerves and bones), the system checks whether the path avoids dangerous areas. If there is overlap, it automatically adjusts the horizontal distance or angle of the needle insertion point until the path is safe.

[0024] The three-dimensional coordinate parameters of the final path are output, including the coordinates of the needle insertion point, the coordinates of the target puncture point, and the needle insertion angle, and are converted into motion commands that can be executed by the robotic arm.

[0025] Optionally, the motion control module is based on forward and inverse kinematics algorithms and combined with a photoelectric encoder with a resolution of not less than 1000 lines / revolution to achieve closed-loop position control.

[0026] The servo motor is driven by a pulse width modulation signal to control the insertion angle of the puncture needle to be 15°-30° and the insertion speed to be 1-3mm / s. The speed can be adaptively adjusted according to the depth of the blood vessel.

[0027] Optionally, the force sensing feedback module integrates a miniature force sensor at the tip of the puncture needle and the joint of the robotic arm, with a preset normal resistance range of 0.5-2N; when the resistance exceeds the range or changes abruptly, the needle insertion is immediately stopped and an audible and visual alarm is triggered; the displacement monitoring threshold of the optical / electromagnetic positioning module is 1mm, and when the patient's puncture site displacement exceeds the threshold, the system pauses and requests recalibration.

[0028] Optionally, the central processing and control unit may further include a biomechanical compensation module;

[0029] The end effector of the probe-controlled robotic arm integrates a high-sensitivity triaxial force sensor for real-time monitoring of the normal force applied to the skin surface by the ultrasound probe. ;

[0030] The biomechanical compensation module is configured to perform the following steps:

[0031] Before the puncture procedure begins, an in-situ tissue elasticity calibration procedure is performed: by controlling the probe to manipulate the robotic arm to apply a series of small, known preset force increments ΔF, and based on the corresponding tissue displacement ΔZ captured by the high-definition ultrasound imaging module, the effective Young's modulus of the soft tissue in the target vascular region is calculated using the least squares method. ;

[0032] During puncture guidance, the predicted depth offset of the target vessel centerline caused by probe pressure is calculated in real time. The predicted depth offset is dynamically corrected based on this offset, adjusting the target puncture point depth coordinates output by the intelligent path planning module. The calculation formula is:

[0033]

[0034] in, The predicted depth offset (m) of the vessel centerline; The probe normal force (N) is monitored in real time. The Poisson's ratio for soft tissues is 0.45-0.49 (dimensionless), with a preset empirical value for mixed subcutaneous fat and muscle tissue. The effective Young's modulus (Pa) calculated by the in-situ calibration procedure; Width of the contact surface of the ultrasonic probe (m); Based on the initial depth H and initial diameter of the blood vessel The calculated geometric influence factor (dimensionless) is used to characterize the non-uniformity of stress distribution.

[0035] Optionally, the central processing and control unit further includes an acoustic coupling adaptive optimization module; the acoustic coupling adaptive optimization module is configured as follows:

[0036] The image frames of the ultrasound video stream are analyzed in real time to calculate the acoustic coupling quality index (ACQ). The ACQ index is comprehensively evaluated by analyzing the reverberation artifact intensity in the near-field region (NFR) of the image and the spatial frequency energy distribution in the target vascular region (ROI).

[0037] Set the ACQ optimization threshold T_opt and tolerance threshold T_min;

[0038] When ACQ is below T_opt, the module activates the probe attitude fine-tuning protocol: controlling the probe manipulator (11) to perform a sequence of micro-movements, including micro-angle deflections (≤0.5°) along the long and short axes of the probe and periodic fine-tuning of the normal pressure (amplitude ≤0.1N, frequency 1-2Hz); this micro-movement is designed to redistribute the ultrasonic gel between the probe and the skin and to expel microbubbles from the contact surface;

[0039] If ACQ remains below T_min after executing the probe posture fine-tuning protocol, the system will pause the puncture procedure and prompt the operator to intervene manually through the human-machine interface.

[0040] An ultrasound-guided intelligent indwelling needle puncture system provided by this invention includes the ultrasound-guided intelligent indwelling needle puncture robot described in any one of the first aspects, and further includes a body positioning fixation component and a storage module;

[0041] The body positioning fixation component is used to fix the patient's puncture site and prevent limb movement during the puncture process;

[0042] The storage module is used to record all data during the puncture process in real time, supporting data traceability and postoperative analysis.

[0043] The body positioning fixation component includes a flexible strap and a body positioning support pad; the flexible strap has a Velcro adjustment structure to adapt to different body types of adults and children, such as the forearm and back of the hand, for puncture sites; the body positioning support pad has a built-in pressure sensor that sends a displacement warning signal to the central processing and control unit when the pressure change caused by the patient's limb movement exceeds 5N.

[0044] The storage module uses a solid-state drive with a capacity of no less than 512GB and a data storage time of no less than 30 days; it supports data export via USB 3.0 interface, and the exported data includes structured data such as ultrasound image frames, puncture path coordinates, robotic arm motion parameters, and force sensing curves.

[0045] The present invention has achieved the following beneficial effects:

[0046] This invention, through the combination of AI image segmentation and sub-millimeter-level robotic arm control, effectively overcomes deviations caused by hand tremors and visual errors in manual operation, significantly improving the success rate of punctures in small and deep blood vessels and reducing the pain caused by repeated punctures. It achieves an automated closed loop from image recognition to puncture execution, eliminating reliance on the operator's superior hand-eye coordination, greatly shortening the training cycle for medical personnel, and alleviating the pressure of a shortage of skilled professionals.

[0047] This invention employs a triple-redundant safety mechanism involving force, vision, and displacement monitoring, coupled with a dual-confirmation process, to minimize the risk of medical accidents. By planning the path according to a unified algorithmic logic, human error is eliminated, establishing a standardized puncture procedure (SOP). The automatic recording of all data provides valuable data assets for clinical research and quality control.

[0048] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0049] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0050] Figure 1 This is a schematic diagram of the puncture robot in this invention;

[0051] Figure 2 This is a rear-view structural diagram of the puncture robot in this invention;

[0052] Figure 3 This is a schematic diagram of the puncture system in this invention.

[0053] In the figure: 1. Multi-degree-of-freedom robotic arm system; 11. Probe-controlled robotic arm; 12. Puncture execution robotic arm; 2. Human-machine interface. Detailed Implementation

[0054] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0055] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Example 1:

[0057] Reference Figure 1 , Figure 2 and Figure 3The intelligent puncture robot described in this embodiment adopts a split or integrated trolley design. It includes a multi-degree-of-freedom robotic arm system 1, a high-definition ultrasonic imaging module, a central processing and control unit, a human-machine interface 2, and a safety monitoring and emergency unit; the multi-degree-of-freedom robotic arm system 1 is mounted on a stable base.

[0058] The multi-degree-of-freedom robotic arm system 1 serves as the execution entity, comprising a probe-controlling robotic arm 11 and a puncture-execution robotic arm 12. The probe-controlling robotic arm 11 is configured to clamp and adjust the position and orientation of the ultrasound probe to ensure that the ultrasound imaging plane can completely cover and accurately locate the target blood vessel area. The puncture-execution robotic arm 12 is configured to clamp and drive the indwelling needle puncturist to perform precise puncture actions, including needle insertion, needle withdrawal, and multi-angle adjustment. The probe-controlling robotic arm 11 and the puncture-execution robotic arm 12 can move independently or be coordinated and controlled in the same coordinate system.

[0059] The high-definition ultrasound imaging module is used to acquire cross-sectional or longitudinal ultrasound video streams of the target vascular region in real time, providing deep imaging information including blood vessels, nerves and surrounding tissues.

[0060] The central processing and control unit, acting as the control center, can connect with the multi-degree-of-freedom robotic arm system 1 and the high-definition ultrasound imaging module to achieve control. The central processing and control unit includes an AI blood vessel recognition and segmentation module, an intelligent path planning module, and a motion control module. The AI ​​blood vessel recognition and segmentation module, based on a pre-trained deep learning model, identifies target blood vessels in the ultrasound video stream in real time, segments the blood vessel contour, and calculates the blood vessel depth, diameter, and centerline coordinates. The intelligent path planning module, based on the three-dimensional spatial information of the blood vessel, plans the optimal puncture path that avoids dangerous areas such as nerves and tendons and conforms to clinical puncture standards. The motion control module converts the puncture path into robotic arm motion commands, controlling the puncture needle to perform the needle insertion operation at a preset angle and speed.

[0061] The human-machine interface 2 is connected to the central processing and control unit and includes a touch screen and operating components. The touch screen is used to display real-time ultrasound images, blood vessel contours, puncture paths and system status. The operating components include physical buttons and a foot switch, and have one-button blood vessel recognition, path confirmation, puncture start and emergency stop functions.

[0062] The safety monitoring and emergency unit is signal-connected to the central processing and control unit, and includes a force sensing feedback module, an optical / electromagnetic positioning module, and a dual confirmation module. The force sensing feedback module is used to monitor the resistance to needle insertion, the optical / electromagnetic positioning module is used to track the displacement of the patient's puncture site, and the dual confirmation module is used to reconfirm the position of the blood vessel before the puncture operation is performed.

[0063] Specifically, the probe-controlling robotic arm 11 is designed as a 5-DOF serial robotic arm, with joint drives using high-precision stepper motors and zero-backlash harmonic reducers. Its end effector features a quick-change interface, compatible with various ultrasound probe types, including linear and convex arrays. The motorized slide rail at the base has a travel of 200mm, allowing for coarse positioning of the probe along the long axis of the patient's forearm. During fine adjustment, the robotic arm can rotate (±45°) and tilt (±30°) the probe around the vessel axis to obtain the optimal acoustic window.

[0064] The puncture execution robotic arm 12 is designed as a 6-DOF robotic arm to provide greater flexibility. Its end effector integrates force sensors (such as a six-dimensional force / torque sensor) and features a quick-loading slot for the indwelling needle. The feed axis (Z-axis) of the puncture needle is driven by a miniature linear module, with a repeatability of up to ±0.02mm, ensuring absolute precision in needle insertion depth.

[0065] The high-definition ultrasound imaging module includes a customized ultrasound host and an image acquisition card. The image acquisition card communicates with the central processing unit (CPU) via a PCIe interface, transmitting lossless RF radio frequency streams or beamformed B-mode images in real time. The CPU converts the ultrasound video stream acquired by the image acquisition card at a frame rate of at least 25fps into digital image frames. Then, using AI core algorithms, it sequentially performs inter-frame denoising algorithms (Gaussian filtering or median filtering), histogram equalization, or edge enhancement algorithms on the digital image frames to improve the contrast between blood vessels and surrounding tissues. The human-machine interface 2 uses an industrial-grade capacitive touchscreen with a resolution of 1920×1080. The screen layout is divided into an image display area (left side, occupying 70%) and a control parameter area (right side, occupying 30%).

[0066] Example 2:

[0067] This embodiment details the core AI algorithm in the central processing and control unit. Digital image frames (1280×720, 25fps) acquired from the acquisition card first enter the preprocessing pipeline. A Gaussian filter is used to smooth the image and remove high-frequency thermal noise; subsequently, a median filter is applied to remove speckle noise specific to ultrasound while preserving tissue edge features.

[0068] Enhancement: Histogram equalization is applied to expand the grayscale dynamic range; edge enhancement is performed by combining Sobel or Canny edge detection operators, which significantly improves the contrast between the vessel wall (strong echo) and the vessel lumen (anechoic).

[0069] The AI ​​recognition core uses the U-Net++ deep neural network. Compared with the traditional U-Net, U-Net++ introduces nested, dense skip paths, which effectively reduces the semantic gap between the encoder and decoder feature maps and greatly improves the segmentation accuracy of small targets (small blood vessels).

[0070] Training dataset: A dataset of no less than 10,000 finely annotated clinical ultrasound vascular images was constructed, covering vascular morphology in different ages, genders, body types, and pathological conditions.

[0071] Loss function optimization: To address the issue of small proportions of blood vessel pixels in the overall image (class imbalance), the loss function... Defined as Dice coefficient loss ( ) and binary cross-entropy loss ( The weighted sum of )

[0072]

[0073] Among them, the weighting coefficient Set it between 0.6 and 0.8. This setting forces the model to focus more on the overlap of vascular regions rather than the classification accuracy of the background.

[0074] Performance metrics: Tests show that the trained model achieves a precision of no less than 95% in identifying target vessels such as the basilar vein and cephalic vein, a Dice coefficient of no less than 0.95 for the segmentation results, and an edge geometric localization error controlled within 0.1mm. It can accurately identify target vessels including the basilar vein and cephalic vein.

[0075] Based on the binary segmentation mask output by the model, perform the following geometric calculations:

[0076] Calculating vessel depth: Based on depth calibration parameters from ultrasound images, a pre-defined actual depth value is used for each pixel. The pixel coordinates of the upper boundary of the vessel in the segmentation mask are extracted and converted into actual depth values ​​based on the skin surface. Specifically, the ordinate of the pixels at the upper boundary of the mask is extracted. Combined with depth calibration parameters (mm / pixel) Calculate vessel depth .

[0077] Determining the vessel diameter: In the segmentation mask, sampling is performed along a direction perpendicular to the vessel's direction. The pixel distances between the two edges of the vessel at each sampling point are calculated, converted to actual dimensions, and the average value is taken as the vessel diameter. Specifically, the mask area is scanned along the normal direction, the maximum cross-sectional width is calculated, and the average value of multiple samples is taken. Determining the centerline coordinates: The skeleton of the segmentation mask is extracted, and a morphological thinning algorithm is used to obtain the pixel-level centerline. This centerline is mapped to the world coordinate system. Combined with the planar coordinate calibration parameters of the ultrasound image, the actual planar distance corresponding to each pixel is preset and converted into actual three-dimensional coordinates. The actual three-dimensional coordinates consist of planar coordinates and depth values.

[0078] Example 3:

[0079] The intelligent path planning module, based on the three-dimensional spatial information of blood vessels and the parameters output by AI, plans the optimal puncture path that avoids dangerous areas such as nerves and tendons and conforms to clinical puncture standards.

[0080] Needle insertion point calculation: The target point is the point on the blood vessel centerline closest to the probe. To ensure sufficient indwelling length of the puncture needle within the blood vessel, the insertion point... Set behind the target, at a horizontal distance Set as blood vessel depth 1.5-2 times that.

[0081] Angle Adaptive: The system automatically recommends the needle insertion angle based on the depth of the blood vessel. :

[0082] like (Superficial blood vessels) This shortens the subcutaneous pathway and reduces pain.

[0083] like (Deep blood vessels) This facilitates the insertion of the catheter.

[0084] Simultaneously constrain the horizontal angle between the puncture needle axis and the long axis of the blood vessel. This is to prevent the needle tip from puncturing the sidewall of the blood vessel.

[0085] Obstacle avoidance verification: The planned path is compared with the coordinates of nerve and skeletal restricted areas in the anatomical structure database. If the path passes through a danger zone (such as the superficial branch of the radial nerve), the system automatically fine-tunes the needle insertion point or angle until the path is safe.

[0086] Coordinate transformation: Using inverse kinematics, the planned Cartesian space path points are transformed. Converted into target angular displacement of each joint motor .

[0087] Closed-loop control: The servo motor has a built-in photoelectric encoder with a resolution of ≥1000 lines / revolution, which provides real-time feedback on the joint position, forming a fully closed-loop control.

[0088] Speed ​​control: The needle insertion speed is controlled between 1-3 mm / s via PWM signal modulation. The system has an adaptive speed adjustment function: a faster speed (3 mm / s) is used in the subcutaneous tissue layer, and the speed is automatically reduced to 1 mm / s when approaching the blood vessel wall to reduce impact.

[0089] Example 4:

[0090] It also includes a positioning fixation component and a storage module; the positioning fixation component is used to fix the patient's puncture site and prevent limb movement during the puncture; the storage module is used to record all data of the puncture process in real time, supporting data traceability and postoperative analysis.

[0091] The body positioning fixation component includes a flexible strap and a body positioning support pad; the flexible strap has a Velcro adjustment structure to fit different body types of adults and children, such as the forearm and back of the hand, for puncture sites; the body positioning support pad has a built-in pressure sensor that sends a displacement warning signal to the central processing and control unit when the pressure change caused by the patient's limb movement exceeds 5N.

[0092] The storage module uses a solid-state drive with a capacity of no less than 512GB and a data storage time of no less than 30 days; it supports data export via USB 3.0 interface, and the exported data includes structured data such as ultrasound image frames, puncture path coordinates, robotic arm motion parameters, and force sensing curves.

[0093] The following describes the complete workflow of this system using adult cephalic vein puncture on the back of the hand as an example:

[0094] Step 1: System initialization and body position fixation

[0095] The robot is activated, and after a successful self-check, it enters standby mode. Medical staff place the patient's hand on the support pad of the positioning fixation component and secure it with flexible straps. When the pressure sensor within the support pad detects stable pressure (>5N), the system displays "Positioning complete," allowing the user to proceed to the next step.

[0096] Step 2: Ultrasonic probe positioning and image acquisition

[0097] The operator selects the "probe positioning" mode on the human-machine interface. The probe-controlled robotic arm 11 moves the ultrasound probe along the skin of the back of the hand at a speed of 5 mm / s for pre-scanning. The interface displays ultrasound images in real time. When the operator or the AI ​​algorithm identifies a clear target blood vessel (such as the basilic vein), the robotic arm stops moving and locks its posture, and the image acquisition module begins to stably output the video stream.

[0098] Step 3: AI Recognition and Path Planning

[0099] The operator presses the "One-Click Vessel Recognition" button. The central processing unit calls the U-Net++ model to complete vessel segmentation within 1 second, and displays the vessel outline in red on the screen, simultaneously annotating the vessel depth (e.g., 6mm) and diameter (e.g., 2.5mm). The system automatically generates a recommended path (green dashed line), displaying a needle entry angle of 20° and a needle entry point 10mm from the target. After confirming that the path avoids the danger zone, the operator clicks "Path Confirmation".

[0100] Step 4: Puncture Procedure and Double Confirmation

[0101] The operator presses the foot switch to start the puncture. The puncture execution robot (12) adjusts its posture and drives the indwelling needle to advance along the predetermined trajectory.

[0102] Force feedback monitoring: Force sensors monitor resistance in real time. Under normal circumstances, the resistance fluctuates between 0.5-2N.

[0103] Displacement monitoring: The optical positioning system continuously monitors the affected limb. If the displacement exceeds 1mm, the machine will immediately stop.

[0104] Double Confirmation: When the needle tip is only 1mm away from the skin surface, the system automatically pauses, triggering a "double confirmation." The AI ​​scans again to confirm that the blood vessel position has not shifted even slightly (error ≤ 0.1mm). A pop-up window appears on the screen prompting "Please confirm blood vessel position." After the operator clicks "Continue," the system performs the final insertion.

[0105] Step 5: Puncture completion and data storage

[0106] When the needle tip pierces the blood vessel, the force sensor detects a sudden drop in resistance (a breakthrough sensation), and the ultrasound image shows the needle tip echo centered in the lumen. The system then determines success and stops needle insertion. The robotic arm stabilizes the needle core and prompts the operator or automated mechanism to withdraw the core.

[0107] After the puncture is completed, the storage module automatically saves the ultrasound video, path parameters, and force sensing curves of this operation, generating a structured data package.

[0108] Example 5:

[0109] The safety design of this system goes beyond just hardware stacking; it also incorporates robust logical interlocks. The motion control module, based on forward and inverse kinematics algorithms, utilizes a photoelectric encoder with a resolution of at least 1000 lines / revolution to achieve closed-loop position control. A servo motor is driven by pulse-width modulation signals to control the needle insertion angle at 15°-30° and the insertion speed at 1-3 mm / s, with the speed adaptively adjusted according to vessel depth. The force feedback module integrates miniature force sensors at the needle tip and robotic arm joints, with a preset normal resistance range of 0.5-2 N. When resistance exceeds this range or a sudden change occurs, needle insertion is immediately stopped, and an audible and visual alarm is triggered. The optical / electromagnetic positioning module has a displacement monitoring threshold of 1 mm. When the patient's puncture site displacement exceeds this threshold, the system pauses and requests recalibration. Details are as follows:

[0110] Force-position hybrid control: During needle insertion, the motion control module does not simply execute position commands, but integrates force sensor data in real time. If the position command indicates that the blood vessel wall has not been reached, but the force sensor reading has exceeded the bone contact threshold (2N), the system prioritizes the force signal and stops immediately to prevent hard impact.

[0111] Dynamic replanning: If a slight shift in the position of the blood vessel is detected during the double confirmation stage (e.g., lateral movement of 0.5mm), and it is within the safety threshold (1mm), the intelligent path planning module can calculate the compensation amount in real time and fine-tune the coordinates of the robotic arm end effector to achieve "dynamic correction".

[0112] System-level emergency stop: Whether it is a software crash, sensor disconnection, or emergency stop button being pressed, the underlying hardware circuitry (FPGA or safety relay) of the system will directly cut off the motor drive power and lock the robotic arm with a brake to ensure that the patient will not be harmed in any fault mode.

[0113] In one embodiment, the central processing and control unit further includes a biomechanical compensation module; the end effector of the probe-controlled robotic arm 11 integrates a high-sensitivity triaxial force sensor for real-time monitoring of the normal force applied to the skin surface by the ultrasound probe. ;

[0114] The biomechanical compensation module is configured to perform the following steps:

[0115] Before the puncture procedure begins, an in-situ tissue elasticity calibration procedure is performed: a series of small, known, preset force increments ΔF are applied by controlling the probe to manipulate the robotic arm 11, and the effective Young's modulus of the soft tissue in the target vascular region is calculated using the least squares method based on the corresponding tissue displacement ΔZ captured by the high-definition ultrasound imaging module. ;

[0116] During puncture guidance, the predicted depth offset of the target vessel centerline caused by probe pressure is calculated in real time. The target puncture point depth coordinates output by the intelligent path planning module are dynamically corrected based on this offset; The calculation formula is:

[0117]

[0118] in, The predicted depth offset (m) of the vessel centerline; The probe normal force (N) is monitored in real time. The Poisson's ratio for soft tissues is 0.45-0.49 (dimensionless), with a preset empirical value for mixed subcutaneous fat and muscle tissue. The effective Young's modulus (Pa) calculated by the in-situ calibration procedure; Width of the contact surface of the ultrasonic probe (m); Based on the initial depth H and initial diameter of the blood vessel The calculated geometric influence factor (dimensionless) is used to characterize the non-uniformity of stress distribution.

[0119] This embodiment provides a highly precise biomechanical compensation mechanism designed to address a key technical challenge in ultrasound-guided puncture: soft tissue deformation caused by the ultrasound probe pressing on the skin surface and its impact on the accuracy of target vessel positioning. In ultrasound-guided procedures, to obtain clear medical images, the probe must be in close contact with the skin and apply pressure. However, human soft tissues (including skin, subcutaneous fat, and muscle) possess complex viscoelastic, nonlinear, and anisotropic biomechanical properties. External forces cause significant tissue deformation, resulting in displacement of the actual spatial position of the target vessel, particularly in the depth direction. If the puncture robot relies solely on ultrasound images acquired under current pressure for positioning, without considering this deformation effect, a systematic deviation will exist between the planned target puncture point coordinates and the actual vessel coordinates. This deviation can easily lead to the puncture needle tip failing to accurately hit the vessel center, or even completely deviating from the target. The core task of the biomechanical compensation module is to establish a personalized biomechanical model by accurately measuring tissue elastic parameters and calculating the vessel displacement caused by probe pressure in real time, thereby dynamically correcting the puncture path and achieving high-precision "deformation sensing" and "force compensation."

[0120] The module's workflow is divided into two closely linked stages: the in-situ tissue elasticity calibration stage before puncture, and the real-time depth compensation stage during puncture guidance.

[0121] In the in-situ tissue elasticity calibration phase, the system aims to obtain the equivalent elastic modulus of the soft tissue in the puncture area, which is a physical quantity that measures the tissue's resistance to deformation (stiffness). This process is fundamental to achieving personalized compensation because there are significant differences in tissue stiffness between different patients and between different sites within the same patient. The calibration procedure is executed automatically before the formal puncture procedure begins.

[0122] First, a robust hardware foundation is crucial. The end effector of the probe-controlled robotic arm 11 integrates a highly sensitive triaxial force sensor. This sensor (e.g., employing strain gauge technology or fiber optic grating sensing technology) can monitor the normal force (i.e., the pressure perpendicular to the skin surface) applied by the ultrasound probe to the skin surface in real time with micro-Newton-level accuracy and extremely high sampling frequency. The system first controls the probe to slowly approach and just touch the skin surface, defining this initial contact state as the zero-stress reference point.

[0123] Subsequently, the central processing and control unit sends a series of precise control commands to the probe-controlled robotic arm, instructing it to apply a series of small, known, preset force increments along the normal direction. This loading process is quasi-static to minimize the effects of tissue viscoelasticity. The magnitudes of these force increments are experimentally designed in advance to ensure that tissue deformation remains within the linear elastic range, avoiding irreversible deformation or patient discomfort.

[0124] Simultaneously with each increment of force, the high-definition ultrasound imaging module acquires ultrasound image frames of the tissue under the current pressure state. Image analysis algorithms in the central processing and control unit are responsible for accurately quantifying tissue displacement from these image frames. This is typically achieved using high-precision image registration techniques, such as speckle tracking based on cross-correlation algorithms or optical flow methods. Speckle tracking utilizes the random speckle patterns formed by tissue microstructures in ultrasound images as natural acoustic markers. By tracking the trajectories of these specks under different pressure states, the displacement field within the tissue can be calculated at high resolution, particularly the change in depth in the target vascular region.

[0125] Through this series of operations, the system obtains a set of discrete force-displacement data pairs. Next, the biomechanical compensation module uses this data to calculate the effective Young's modulus of the soft tissue in the target vascular region. The module employs numerical optimization methods, particularly least squares fitting techniques, to process this data. Least squares aims to find an optimal Young's modulus value that minimizes the sum of errors between the theoretical displacement predicted based on this modulus and the actual observed displacement. In this way, the system completes an in-situ, non-invasive, and personalized quantitative assessment of the tissue elasticity at the current puncture site.

[0126] Once the puncture guidance phase begins, the biomechanical compensation module starts performing real-time depth compensation. At its core is a biomechanical model based on the principles of continuum mechanics, which predicts the depth offset of the target vessel's centerline based on the real-time probe pressure.

[0127] The prediction process involves the comprehensive calculation of several key parameters. First, there is the continuous real-time probe normal force data stream provided by the triaxial force sensor, which is the primary driving input to the model. Second, the model utilizes the personalized effective Young's modulus obtained during the calibration phase to reflect the patient-specific tissue stiffness.

[0128] Furthermore, the model must also consider the Poisson effect of soft tissue. Poisson's ratio is a physical quantity that describes the tendency of a material to expand in the perpendicular direction when compressed in one direction. Since subcutaneous fat and muscle mixed tissue is close to incompressible, its volume remains essentially constant during deformation, resulting in a Poisson's ratio close to 0.5. The system is preset with empirical values ​​for specific anatomical locations, typically set in the range of 0.45 to 0.49, to accurately simulate the volume retention characteristics of tissue.

[0129] The geometry of the probe is also an important factor affecting stress distribution. The model needs to know the width of the ultrasonic probe's contact surface. The wider the contact surface, the lower the pressure per unit area when the same force is applied, and the more dispersed the stress distribution.

[0130] Finally, to more accurately describe the non-uniformity of stress transmission and distribution within tissue, the model introduces a geometric influence factor. This factor is calculated based on the initial depth and diameter of the blood vessel. It reflects the phenomenon that stress gradually decreases with increasing depth (consistent with Saint-Venant's principle). Under the same surface pressure, blood vessels of different depths and sizes experience different local stresses and strains. By considering the specific geometry and location of the blood vessel, the geometric influence factor allows the model to more accurately predict the offset of a specific target blood vessel.

[0131] The biomechanical compensation module substitutes all the above parameters into a preset biomechanical model for real-time solution, outputting the predicted depth offset of the target blood vessel centerline. This calculation process is continuous and real-time, capable of rapidly responding to any changes in probe pressure.

[0132] The calculated predicted depth offset is then fed back to the intelligent path planning module in real time. When determining the three-dimensional coordinates of the target puncture point based on ultrasound image information, the intelligent path planning module uses this depth offset to dynamically correct the depth coordinates of the target puncture point. The system fine-tunes the original path plan based on the predicted offset, ensuring that the endpoint of the puncture path accurately corresponds to the actual spatial position of the blood vessel centerline, even when the tissue is under pressure and deformation. This dynamic compensation mechanism greatly improves the robot's operational accuracy in complex biological environments, achieving a leap from simple "image guidance" to "image and mechanical fusion guidance," significantly increasing the success rate of the first puncture.

[0133] The central processing and control unit further includes an acoustic coupling adaptive optimization module; the acoustic coupling adaptive optimization module is configured as follows:

[0134] The image frames of the ultrasound video stream are analyzed in real time to calculate the acoustic coupling quality index (ACQ). The ACQ index is comprehensively evaluated by analyzing the reverberation artifact intensity in the near-field region (NFR) of the image and the spatial frequency energy distribution in the target vascular region (ROI).

[0135] Set the ACQ optimization threshold T_opt and tolerance threshold T_min;

[0136] When ACQ is lower than T_opt, the module activates the probe attitude fine-tuning protocol: controlling the probe manipulator 11 to perform a sequence of micro-movements, including micro-angle deflections (≤0.5°) along the probe's long and short axes and periodic fine-tuning of normal pressure (amplitude ≤0.1N, frequency 1-2Hz); these micro-movements are designed to redistribute the ultrasound gel between the probe and the skin and to expel microbubbles from the contact surface;

[0137] If ACQ remains below T_min after executing the probe posture fine-tuning protocol, the system will pause the puncture procedure and prompt the operator to intervene manually through the human-machine interface 2.

[0138] This embodiment further provides an innovative acoustic coupling adaptive optimization module, designed to address the technical bottleneck of maintaining stable and high-quality ultrasound imaging in ultrasound-guided puncture robotic systems. The performance of ultrasound-guided systems, particularly the accuracy of the AI ​​vessel recognition and segmentation module, is highly dependent on the quality of the input ultrasound image. Image quality, in turn, directly depends on the acoustic coupling effect between the ultrasound probe and the patient's skin—that is, the efficiency of sound energy transmission between the probe and the tissue. In clinical practice, acoustic coupling is often suboptimal and dynamically changing due to factors such as uneven application of ultrasound gel, residual micro-air bubbles in the gel, improper probe-skin contact angle, or insufficient contact pressure. Poor acoustic coupling leads to ultrasound signal attenuation, reduced signal-to-noise ratio, and increased image artifacts, severely impacting subsequent image processing and intelligent decision-making. The acoustic coupling adaptive optimization module achieves autonomous monitoring and optimization of acoustic coupling by quantitatively evaluating coupling quality in real time and actively controlling the robotic arm for fine-tuning, ensuring that image quality remains optimal throughout the entire puncture process.

[0139] First, real-time image quality quantification is a prerequisite for adaptive optimization. This module receives video stream image frames from the high-definition ultrasound imaging module in real time and performs in-depth analysis on each frame to calculate a quantified Acoustic Coupling Quality Index (ACQ). This index is designed to sensitively reflect subtle changes in acoustic coupling and possesses good robustness. This invention employs a multi-feature fusion method to calculate the ACQ, comprehensively considering the feature performance of different regions of the image.

[0140] The ACQ index is calculated primarily based on the analysis of two key regions: the near field region (NFR) and the target vessel region (ROI).

[0141] In the near-field region (i.e., the superficial tissue area near the probe surface in the image), the most typical manifestation of poor acoustic coupling is the appearance of reverberation artifacts. When an air layer or bubble exists between the probe and the skin, ultrasound waves will repeatedly reflect between the probe surface and the air interface, forming a series of strong echoing bright lines parallel to the probe surface. The acoustic coupling adaptive optimization module quantifies the intensity of these reverberation artifacts through image processing algorithms. For example, the module can use texture analysis techniques (such as gray-level co-occurrence matrix) to extract parameters reflecting the characteristics of periodic bright lines, or analyze the gray-level histogram distribution and local variance of the near-field region. The higher the artifact intensity, the worse the acoustic coupling.

[0142] Within the target vascular area (ROI), acoustic coupling quality directly impacts image sharpness and detail. This module assesses this by analyzing the spatial frequency energy distribution within the ROI. Spatial frequency reflects the rate of grayscale changes in the image; high spatial frequencies correspond to image edges and details. The module typically employs frequency domain analysis techniques, such as Fast Fourier Transform (FFT) or wavelet transform, to convert the ROI image from the spatial domain to the frequency domain, and then analyzes its spectrum. A higher proportion of high-frequency components in the spectrum indicates a sharper image, richer details, and better acoustic coupling. Conversely, severe attenuation of high-frequency energy indicates a blurry image, suggesting potential coupling problems.

[0143] The acoustic coupling adaptive optimization module calculates the final comprehensive ACQ index by weighted fusion of these two key features.

[0144] Secondly, the threshold-based intelligent decision-making logic determines the system's response strategy. The module internally sets two key ACQ thresholds: the optimization threshold (T_opt) and the tolerance threshold (T_min). T_opt represents the optimal acoustic coupling level required for the system's ideal operating state. T_min represents the minimum acceptable acoustic coupling level to ensure the system's safe and reliable operation.

[0145] When the module detects that ACQ is lower than T_opt but still higher than T_min, it indicates that the acoustic coupling is in a suboptimal state and intervention is needed to restore optimal image quality. At this time, the module will activate an automated probe attitude fine-tuning protocol.

[0146] Third, the execution of the probe posture fine-tuning protocol is a key step in achieving autonomous optimization. This protocol is a sophisticated robotic control strategy designed to physically improve the contact between the probe and the skin by controlling the probe to manipulate the robotic arm 11 to perform a series of tiny, sequential movements. The goal of these micro-movements is to redistribute the ultrasound gel and facilitate the expulsion of microbubbles from the contact surface without losing the target vascular field of view.

[0147] The protocol mainly consists of a combination of two core actions: small-angle deflection and periodic fine-tuning of normal pressure.

[0148] Micro-angle deflection refers to controlling the robotic arm to make slight tilting or swaying movements of the ultrasound probe along its long and short axes. These angular deflections are very small (e.g., typically controlled at the sub-degree level, such as no more than 0.5 degrees). This slight angular change has a dual effect. On one hand, it helps the probe contact surface better adapt to the local curvature of the skin surface, ensuring uniform contact. On the other hand, the tilting motion generates shear stress within the gel layer. This shearing action promotes gel flow, filling any tiny gaps that may be present, and helps to "scrape away" or push air bubbles adhering to the probe surface or skin to the edges of the contact surface.

[0149] Periodic fine-tuning of normal pressure refers to controlling the robotic arm to apply a tiny, periodic pressure change in a direction perpendicular to the skin surface. This can manifest as a slight vibration or a press-and-release motion. Its amplitude is very small (e.g., the force change does not exceed 0.1 Newtons) and the frequency is moderate (e.g., between 1 and 2 Hz). This periodic pressure change has a "pumping effect," helping to break the surface tension of microbubbles, causing them to migrate from the center of the contact surface to the edge and eventually be expelled. Simultaneously, this tiny pressure fluctuation also helps the gel better penetrate the skin's microstructure.

[0150] During the execution of the probe attitude fine-tuning protocol, the acoustic coupling adaptive optimization module continuously monitors changes in the ACQ index. If the ACQ index rises above the optimization threshold T_opt, the module stops the fine-tuning protocol, and the system resumes normal boot operations.

[0151] Finally, the module also includes a safety mechanism. If, after executing the probe posture fine-tuning protocol, the ACQ index remains below the tolerance threshold T_min, this indicates extremely poor acoustic coupling, and the image quality has deteriorated to a level that cannot guarantee AI recognition accuracy and puncture safety. At this point, the system determines that its automatic optimization capabilities have reached their limit, and manual intervention is necessary. The module will immediately instruct the central processing and control unit to pause the entire puncture procedure, lock all robotic arm movements to ensure safety, and issue clear alarms and prompts to the operator through the human-machine interface 2 (e.g., prompting to check the ultrasound gel and manually adjust the probe). The operator needs to intervene manually. After the problem is resolved, the system reassesses the ACQ index, and the puncture procedure can only continue after confirming that the image quality is acceptable.

[0152] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An ultrasound-guided intelligent indwelling needle puncture robot, characterized in that, Including the following: The multi-degree-of-freedom robotic arm system (1) includes a probe control robotic arm (11) and a puncture execution robotic arm (12); the probe control robotic arm (11) is used to clamp and adjust the position and orientation of the ultrasound probe to ensure that the ultrasound imaging plane can completely cover and accurately locate the target blood vessel area; the puncture execution robotic arm (12) is used to clamp and drive the indwelling needle puncture device to perform fine puncture actions including needle insertion, needle withdrawal and multi-angle adjustment; The high-definition ultrasound imaging module is used to acquire ultrasound video streams of cross-sections or longitudinal sections of the target vascular region in real time, providing deep imaging information including blood vessels, nerves and surrounding tissues. The central processing and control unit includes an AI vessel recognition and segmentation module, an intelligent path planning module, and a motion control module. The AI ​​vessel recognition and segmentation module acquires ultrasound video streams and, based on a pre-trained deep learning model, identifies target vessels in the ultrasound video streams in real time, segments the vessel contours, and calculates the vessel depth, diameter, and centerline coordinates. The intelligent path planning module, based on the three-dimensional spatial information of the vessels, plans the optimal puncture path that avoids dangerous areas and conforms to clinical puncture standards. The motion control module converts the optimal puncture path into robotic arm motion commands, controlling the puncture needle to perform the needle insertion operation at a preset angle and speed. The human-computer interaction interface (2) includes a touch screen and an operation component. The touch screen is used to display real-time ultrasound images, blood vessel contours, puncture paths and system status. The operation component includes physical buttons and a foot switch, and has one-click blood vessel recognition, path confirmation, puncture start and emergency stop functions. The safety monitoring and emergency unit includes a force sensing feedback module, an optical / electromagnetic positioning module, and a dual confirmation module; the force sensing feedback module is used to monitor the needle insertion resistance, the optical / electromagnetic positioning module is used to track the displacement of the patient's puncture site, and the dual confirmation module is used to reconfirm the blood vessel position before the puncture operation is performed. The central processing and control unit also includes a biomechanical compensation module; The end effector of the probe-controlled robotic arm (11) integrates a high-sensitivity triaxial force sensor for real-time monitoring of the normal force applied to the skin surface by the ultrasound probe. The biomechanical compensation module is configured to perform the following steps: Before the puncture procedure begins, an in-situ tissue elasticity calibration procedure is performed: a series of small, known preset force increments ΔF are applied by controlling the probe to manipulate the robotic arm (11), and the effective Young's modulus of the soft tissue in the target vascular region is calculated using the least squares method based on the corresponding tissue displacement ΔZ captured by the high-definition ultrasound imaging module. ; During puncture guidance, the predicted depth offset of the target vessel centerline caused by probe pressure is calculated in real time. The predicted depth offset is dynamically corrected based on this offset, adjusting the target puncture point depth coordinates output by the intelligent path planning module. The calculation formula is: in, This represents the predicted depth offset of the vessel centerline. The probe's normal force is monitored in real time; The Poisson's ratio for soft tissues, specifically for mixed subcutaneous fat and muscle tissue, is preset to an empirical value of 0.45-0.

49. The effective Young's modulus calculated by the in-situ calibration procedure; The width of the contact surface of the ultrasonic probe; Based on the initial depth H and initial diameter of the blood vessel The calculated geometric influence factor is used to characterize the non-uniformity of stress distribution.

2. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The high-definition ultrasound imaging module includes an ultrasound host and an image acquisition card. The image acquisition card communicates with the central processing and control unit via a PCIe interface. The central processing and control unit converts the ultrasound video stream acquired by the image acquisition card at a frame rate of not less than 25fps into digital image frames, and sequentially performs inter-frame denoising algorithms such as Gaussian filtering or median filtering, and image enhancement algorithms such as histogram equalization or edge enhancement on the digital image frames to improve the contrast between blood vessels and surrounding tissues.

3. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The AI ​​vessel recognition and segmentation module employs a pre-trained U-Net++ deep learning model. This model has been trained on no fewer than 10,000 clinical ultrasound vessel images, achieving a vessel recognition accuracy of no less than 95%, and can accurately identify target vessels including the basilic vein and cephalic vein. The loss function of the U-Net++ deep learning model is a weighted sum of Dice loss and binary cross-entropy loss, with the Dice loss weight coefficient ranging from 0.6 to 0.

8. After training, the Dice coefficient for vessel segmentation of the U-Net++ deep learning model is no less than 0.95, and the vessel edge recognition error is ≤0.1mm.

4. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The calculation process of the intelligent path planning module includes: Based on the three-dimensional coordinates of blood vessels output by the AI ​​blood vessel recognition and segmentation module, including the planar coordinates and depth values ​​of the blood vessel centerline, and taking the point on the blood vessel centerline closest to the skin as the target puncture point, the coordinates of the needle insertion point are calculated: the horizontal distance between the needle insertion point and the target puncture point is 1.5-2 times the blood vessel depth, and it is located on the extended line of the long axis of the blood vessel. The AI ​​vessel recognition and segmentation module's calculation process for vessel depth, diameter, and centerline coordinates includes: Blood vessel depth: Based on the depth calibration parameters of ultrasound images, the actual depth value corresponding to each pixel is preset, the pixel coordinates of the upper boundary of blood vessels in the segmentation mask are extracted and converted into the actual depth value based on the skin surface; Blood vessel diameter: In the segmentation mask, samples are taken along the direction perpendicular to the direction of the blood vessel, the pixel distance between the two sides of the blood vessel at each sampling point is calculated, and the average value is taken as the blood vessel diameter after being converted into the actual size. Centerline coordinates: The skeleton of the segmentation mask is extracted and a morphological thinning algorithm is used to obtain the pixel-level centerline. Combined with the planar coordinate calibration parameters of the ultrasound image, the actual planar distance corresponding to each pixel is preset and converted into actual three-dimensional coordinates. The actual three-dimensional coordinates are composed of planar coordinates and depth values.

5. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, Determine the needle insertion angle based on the depth and diameter of the blood vessel: when the blood vessel depth is ≤5mm, the needle insertion angle is 25°-30°, and when the depth is >5mm, the needle insertion angle is 15°-20°, ensuring that the angle between the puncture needle path and the long axis of the blood vessel is ≤10°. Based on a pre-set anatomical structure database, the system checks whether the path avoids dangerous areas. If there is overlap, it automatically adjusts the horizontal distance or angle of the needle entry point until the path is safe. The three-dimensional coordinate parameters of the final path are output, including the coordinates of the needle insertion point, the coordinates of the target puncture point, and the needle insertion angle, and are converted into motion commands that can be executed by the robotic arm.

6. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The motion control module is based on forward and inverse kinematics algorithms and combines a photoelectric encoder with a resolution of no less than 1000 lines / revolution to achieve closed-loop position control. The servo motor is driven by a pulse width modulation signal to control the insertion angle of the puncture needle to be 15°-30° and the insertion speed to be 1-3 mm / s. The speed can be adaptively adjusted according to the depth of the blood vessel.

7. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The force sensing feedback module integrates a miniature force sensor at the tip of the puncture needle and the joint of the robotic arm, with a preset normal resistance range of 0.5-2N. When the resistance exceeds the range or changes abruptly, the needle insertion is stopped immediately and an audible and visual alarm is triggered. The displacement monitoring threshold of the optical / electromagnetic positioning module is 1mm. When the displacement of the patient's puncture site exceeds the threshold, the system pauses and requests recalibration.

8. The ultrasound-guided intelligent indwelling needle puncture robot according to claim 1, characterized in that, The central processing and control unit further includes an acoustic coupling adaptive optimization module; the acoustic coupling adaptive optimization module is configured as follows: The image frames of the ultrasound video stream are analyzed in real time to calculate the acoustic coupling quality index (ACQ). The ACQ index is comprehensively evaluated by analyzing the reverberation artifact intensity in the near-field region (NFR) of the image and the spatial frequency energy distribution in the target vascular region (ROI). Set the optimization threshold T_opt and tolerance threshold T_min for ACQ; When ACQ is below T_opt, the module activates the probe attitude fine-tuning protocol: controls the probe manipulator (11) to perform a sequence of micro-movements, including micro-angle deflections along the long and short axes of the probe and periodic fine-tuning of the normal pressure; the micro-movements are designed to redistribute the ultrasound gel between the probe and the skin and to expel micro-bubbles from the contact surface; If ACQ is still lower than T_min after the probe posture fine-tuning protocol is executed, the system will pause the puncture procedure and prompt the operator to make manual intervention through the human-machine interface (2).

9. An ultrasound-guided intelligent indwelling needle puncture system, characterized in that, The ultrasound-guided intelligent indwelling needle puncture robot, as described in any one of claims 1-8, further includes a body positioning fixation component and a storage module; The body positioning fixation component is used to fix the patient's puncture site and prevent limb movement during the puncture process; The storage module is used to record all data during the puncture process in real time, supporting data traceability and postoperative analysis. The body positioning fixation component includes a flexible strap and a body positioning support pad; the flexible strap has a Velcro adjustment structure to adapt to puncture sites of different body types in adults and children; the body positioning support pad has a built-in pressure sensor that sends a displacement warning signal to the central processing and control unit when the pressure change caused by the patient's limb movement exceeds 5N. The storage module uses a solid-state drive with a capacity of no less than 512GB and a data storage time of no less than 30 days; it supports data export via USB 3.0 interface, and the exported data includes structured data such as ultrasound image frames, puncture path coordinates, robotic arm motion parameters, and force sensing curves.

Images