Portable blood vessel puncture device and method based on AI identification type-B ultrasonic image

Abstract

The invention relates to the technical field of intelligent medical treatment, and discloses a portable blood vessel puncture device and method based on an AI identification B-ultrasonic image, and the method comprises the steps: determining a puncture body position and a to-be-scanned surface; constructing a coordinate system; based on the two-dimensional grid layout, driving the B-ultrasonic puncture integrated equipment to perform matrix scanning in the to-be-scanned surface to obtain and visualize a plurality of static blood vessel sub-images arranged in a matrix and corresponding coordinates of the static blood vessel sub-images; performing image processing by using the plurality of static blood vessel sub-images and the corresponding coordinates thereof to obtain and visualize continuous blood vessel structure image data; an AI recognition model is constructed, the blood vessel structure image data, the multiple static blood vessel sub-images, the corresponding coordinates and preset puncture data serve as input, a puncture strategy is output and visualized, and the puncture strategy comprises target puncture data and a puncture path which are determined based on a coordinate system; and driving the B-ultrasonic puncture integrated equipment to puncture along the puncture path. According to the invention, full automation of blood vessel puncture to establish deep and shallow artery and vein channels is realized, and the puncture success rate and safety are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and particularly to a portable vascular puncture device and method based on AI recognition of B-ultrasound images. Background Art

[0002] In clinical medicine, vascular puncture is a common and crucial operation technique, widely used in multiple scenarios such as infusion therapy, blood collection, interventional surgery, and blood pressure monitoring. Common puncture sites include the radial artery, femoral artery, internal jugular vein, and peripheral veins, etc., and the puncture purposes cover various medical needs such as catheter placement, blood sample acquisition, and angiography examination.

[0003] Currently, traditional puncture methods mainly rely on the operator's experience and surface anatomical landmarks for blind puncture, or use ultrasonic equipment for real-time guidance. However, these methods still have many deficiencies. On the one hand, the blind puncture method is prone to puncture failure or complications due to individual anatomical differences or different operator technical levels; on the other hand, although ultrasonic guidance improves the visualization of puncture, in actual operation, due to the dynamic characteristics of blood vessels, such as pulsation with the pulse, movement with respiration, and deformation caused by the contact force of the needle tip, the position and shape of blood vessels are constantly changing, resulting in a deviation between the image and the actual position. In addition, the existing image acquisition systems still have certain limitations in frame rate, resolution, and penetration depth, making it difficult to achieve high-precision real-time capture of the dynamic changes of blood vessels during the puncture process, thus affecting the accurate judgment and adjustment of the puncture path. Summary of the Invention

[0004] The present invention aims to provide a portable vascular puncture device and method based on AI recognition of B-ultrasound images. By combining medical imaging technology, artificial intelligence algorithms, and robot control, optimizing the image acquisition strategy and puncture strategy, it realizes the full-process automated operation from blood vessel recognition, path planning to automatic puncture execution, improving the puncture success rate and safety.

[0005] The basic solution provided by the present invention is: a portable vascular puncture method based on AI recognition of B-ultrasound images, including the following steps: S1, determining the puncture position and the surface to be scanned; S2, constructing a coordinate system; driving a B-ultrasound puncture integrated device to perform matrix scanning within the surface to be scanned based on a two-dimensional grid layout, obtaining and visualizing a number of static blood vessel sub-images arranged in a matrix and their corresponding coordinates; S3, using a number of static blood vessel sub-images and their corresponding coordinates for image processing, obtaining and visualizing continuous blood vessel structure image data; S4. Construct an AI recognition model, which takes vascular structure image data, a number of static vascular sub-images and their corresponding coordinates, and preset puncture data as inputs, outputs and visualizes the puncture strategy, and the puncture strategy includes target puncture data and a puncture path determined based on a coordinate system; S5. Drive the B-ultrasound puncture integrated device to perform puncture along the puncture path.

[0006] The present invention also provides a portable vascular puncture device based on AI recognition of B-ultrasound images, which executes the above-mentioned portable vascular puncture method based on AI recognition of B-ultrasound images. The device includes a puncture assist robot and a host computer that are electrically connected; The puncture assist robot includes a control unit, a B-ultrasound puncture integrated device, a motor drive unit, and a pose adjustment unit that are electrically connected to the control unit; wherein the B-ultrasound puncture integrated device is arranged on the pose adjustment unit; The host computer includes a visualization unit, a data processing unit, an AI recognition unit, and a human-computer interaction unit that are electrically connected.

[0007] The working principle and advantages of the present invention are as follows: The puncture assist robot is installed at a position corresponding to the puncture body position and the surface to be scanned; human-computer interaction is carried out based on the host computer, and the puncture assist robot is controlled through the host computer to execute the portable vascular puncture method based on AI recognition of B-ultrasound images, and the image data is displayed on the host computer.

[0008] Compared with the prior art, the present solution realizes the full automation of vascular puncture to establish deep and shallow arteriovenous channels. By combining medical imaging technology, artificial intelligence algorithms and robot control, the image acquisition strategy and puncture strategy are optimized, and the full-process automated operation from vascular recognition, path planning to automatic puncture execution is realized, improving the puncture success rate and safety.

[0009] In S1, the puncture operator combines clinical experience and artificial intelligence technology to jointly determine the puncture site and scanning plane. Through the human-machine collaboration method, the accuracy of the system in actual applications is effectively improved, the errors that may be brought about by solely relying on model recognition are compensated, the adaptability to individual differences of different patients is enhanced, and a reliable foundation is laid for subsequent image acquisition and intelligent analysis. In the S2 stage, by innovatively constructing a three-dimensional coordinate system and adopting a matrix scanning strategy, the portable B-ultrasound device is driven to perform multi-point static image acquisition in the target area according to a predetermined grid, effectively overcoming the motion blur problem common in traditional dynamic B-ultrasound images, and significantly improving the clarity and spatial positioning accuracy of vascular sub-images. Compared with the existing technology that relies on the processing of the overall vascular image, this solution is based on the stitching and analysis of static sub-images in sub-regions with high density, and can better adapt to the dynamic changes of the local morphology of blood vessels, so as to obtain more accurate and detailed vascular structure information, providing a high-quality data basis for subsequent intelligent recognition and path planning. Subsequently, in S3, by stitching and enhancing multiple static images, continuous and clear vascular structure image data is reconstructed, providing a reliable basis for subsequent intelligent analysis. In S4, a deep learning-driven AI recognition model is introduced, which integrates vascular images, coordinate data, and historical puncture experience to intelligently generate the optimal puncture path and parameter configuration, and realizes visual presentation, greatly improving the decision-making efficiency and accuracy. In the S5 stage, the integrated puncture device is controlled to complete the precise puncture operation according to the planned path, and the whole process does not require manual intervention. Compared with the existing technology, this solution not only significantly improves the puncture success rate and safety, but also has the advantages of small size, controllable cost, and convenient operation, is applicable to a variety of clinical scenarios, and has broad promotion prospects and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is a schematic flow chart of a portable vascular puncture method based on AI recognition of B-ultrasound images provided by Embodiment 1 of the present invention; Figure 2 is a schematic plan view of a matrix scanning provided by Embodiment 1 of the present invention; Figure 3 is a schematic diagram of a puncture strategy provided by Embodiment 1 of the present invention; Figure 4 is a schematic structural diagram of a portable vascular puncture device based on AI recognition of B-ultrasound images provided by Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] The following is a more detailed description through specific embodiments: The embodiment is basically as shown in the attached Figure 1 figures: A portable vascular puncture method based on AI recognition of B-ultrasound images includes the following steps: S1, determine the puncture position and the surface to be scanned; S2. Construct a coordinate system; drive the B-ultrasound puncture integrated device to perform matrix scanning within the surface to be scanned based on a two-dimensional grid layout, and obtain and visualize a number of static blood vessel sub-images arranged in a matrix and their corresponding coordinates; S3. Use a number of static blood vessel sub-images and their corresponding coordinates for image processing to obtain and visualize continuous blood vessel structure image data; S4. Construct an AI recognition model, with the blood vessel structure image data, a number of static blood vessel sub-images and their corresponding coordinates, and preset puncture data as inputs, output and visualize a puncture strategy, where the puncture strategy includes target puncture data and a puncture path determined based on the coordinate system; S5. Drive the B-ultrasound puncture integrated device to perform puncture along the puncture path.

[0012] Specifically: In S1, the puncture operator preliminarily determines a suitable puncture position according to the puncture purpose and actual clinical experience. For example, for venous puncture, on the arm; then determines an empirical puncture point according to actual clinical experience, such as Figure 2 shown as the small black square; take the range with the empirical puncture point inside or centered on the empirical puncture point as the surface to be scanned, such as Figure 2 shown as the large gray frame. Thus, the surface to be scanned is the range where the optimal puncture point is most likely to appear judged by clinical experience. In the way of integrating clinical experience and subsequent artificial intelligence, the accuracy of subsequent puncture-related data can be improved.

[0013] In S2, construct a coordinate system, that is, establish a Cartesian coordinate system, such as Figure 2 shown. The initial scanning starting point of the B-ultrasound puncture integrated device is within the surface to be scanned. Take the initial scanning starting point of the B-ultrasound puncture integrated device as the coordinate origin 0(0, 0), and define the scanning surface as the two-dimensional plane OXY; determine the blood vessel cross-sectional diameter according to the X-axis coordinate, that is, the direction of the X-axis is the direction to determine the blood vessel cross-sectional diameter. X-axis: the horizontal direction parallel to the scanning surface, with the right direction as the positive direction; determine the blood vessel trend according to the Y-axis coordinate, that is, the direction of the Y-axis is the direction to determine the blood vessel trend. Y-axis: perpendicular to the X-axis, with the forward direction as the positive direction. Of course, the specific direction can be adaptively determined according to the mechanical structure of the device.

[0014] In addition, such as Figure 3 shown, take the direction perpendicular to the two-dimensional plane OXY as the Z-axis direction. The initial position of the B-ultrasound puncture integrated device is in the positive direction of the Z-axis, and the blood vessel position is in the negative direction of the Z-axis. Determine the puncture depth according to the Z-axis coordinate.

[0015] Such as Figure 2 shown, for matrix scanning, the coordinate position of each acquisition point of the B-ultrasound puncture integrated device is expressed as P i，j =(x i , y j), i = 1, 2, 3, …, M; j = 1, 2, 3, …, N; where, x i = x0 + (i - 1) × Δx; y j = y0 + (j - 1) × Δy; x0 and y0 are the abscissa and ordinate of the origin of the coordinate system respectively, and Δx and Δy respectively represent the fixed distances between adjacent acquisition points in the X-axis and Y-axis directions, that is, the device movement step size; M and N are the number of grid rows and columns respectively.

[0016] The matrix scanning path is to adopt a row-first traversal strategy: starting from the first row and the first column, scan along the x-axis to the first row and the Nth column; move along the Y-axis to the second row and the Nth column, and scan in the reverse direction to the second row and the first column; repeat the above process until the scanning of the Mth row is completed. The snake-shaped scanning path can reduce the moving distance of the robotic arm.

[0017] Taking the coordinates P i，j of each scanning point in the matrix scanning as the coordinates of the corresponding static blood vessel sub-image, drive the B-ultrasound puncture integrated device to move to the scanning point coordinates P i，j At the corresponding position and after stabilization, start B-ultrasound imaging by triggering a signal (such as an encoder pulse) to obtain the corresponding static blood vessel sub-image I i，j .

[0018] In S3, perform image registration on all the collected static blood vessel sub-images; and use an image stitching algorithm for image stitching to obtain a continuous blood vessel structure image, and at the same time extract the blood vessel centerline and visualize it.

[0019] First, the Frangi filter can be used for image enhancement to highlight the blood vessel structure, in this way, the image quality can be improved.

[0020] Image registration can adopt a feature-based registration method. Extract stable image feature points from each sub-image, such as edges, corners or texture structures, and then use corresponding algorithms for processing. For example, when choosing edges as image feature points, the Canny edge detection algorithm can be used. Taking the first blood vessel sub-image as the reference image, match the feature points in other sub-images with the feature points in the reference image respectively to find the corresponding relationship. According to the set of matching points, estimate the geometric transformation relationship between the images, determine the transformation model, and resample the sub-image according to the estimated transformation model to align it with the reference image. Resampling can adopt bilinear interpolation or cubic spline interpolation. Repeat the above process to complete the one-to-one image registration of the first blood vessel sub-image and other blood vessel sub-images. Since the first blood vessel sub-image is obtained by the first scan and is least affected by external interference, taking the first blood vessel sub-image as the reference image can ensure the accurate relative positions between the images and improve the accuracy of the overall image registration.

[0021] The sub-images are merged into a complete and continuous vascular structure image by using an image stitching algorithm. That is, after the transformation models between all images and the reference image are established, each image is resampled into the coordinate system of the reference image; the positions of each image in the final stitched image are determined; the overlapping regions are fused, for example, bilinear interpolation is used to process the overlapping regions to eliminate stitching gaps, or the weighted average method is used; a complete vascular structure image is output, which contains all vascular cross-sectional information within the scanning range and the contour shape data of the surrounding tissues, including: continuous display of vascular tomographic images within all acquisition regions; structural information such as vascular orientation, branches, and diameters; contour information of the surrounding tissues; and spatial consistency and geometric accuracy are ensured. The main direction of the blood vessels is obtained by combining PCA (Principal Component Analysis), and the centerline of the blood vessels is extracted by using a centerline extraction algorithm (such as VMTK or a graph search-based method), including the centerline segments of the blood vessels divided according to the two-dimensional grid layout, the orientation of the centerline segments of the blood vessels, and their spatial positions in the coordinate system, which are used as one of the important features for inputting the vascular structure image data of the subsequent AI recognition model.

[0022] In S4, for the construction of the AI recognition model, the clinical database can be used to prepare the data for model training, including but not limited to vascular images, puncture points, puncture angles, puncture paths, puncture forces, etc., and the model is trained and verified to be qualified.

[0023] The process of taking the vascular structure image data, several static vascular sub-images and their corresponding coordinates, and the puncture preset data as inputs and outputting the puncture strategy includes: S41, determining the optimal puncture vascular sub-image among several static vascular sub-images.

[0024] Specifically, by comparing the curvature of the vascular orientation and the diameter of the vascular cross-section in each vascular sub-image, the vascular sub-image with the smallest curvature of the orientation and the largest cross-section diameter is found as the optimal puncture vascular sub-image.

[0025] Based on the centerline segments of the blood vessels (such as Figure 2 the schematic diagram of the curved line segment) in each vascular sub-image, the curvature of the vascular orientation is determined, that is, the direction vector is obtained by determining the two end points of the centerline segment of the blood vessel, and the offset angle of the direction vector from the Y-axis (such as Figure 2 the schematic diagram of the angle ɑ) is used to represent the curvature of the vascular orientation. The larger the offset angle, the larger the curvature of the orientation, and the smaller the offset angle, the smaller the curvature of the orientation. The diameter of the vascular cross-section is determined by the width of the vascular edge in the X-axis in each vascular sub-image.

[0026] Based on the clinical experience data, an evaluation model is constructed to determine the mapping relationship between the optimal curvature of the vascular orientation and the optimal diameter of the vascular cross-section, so as to select the optimal puncture vascular sub-image according to the comparison of the curvature of the vascular orientation and the diameter of the vascular cross-section in each vascular sub-image.

[0027] S42. On the optimal puncture blood vessel sub-image, determine the puncture target point and its corresponding coordinates based on the blood vessel structure image data.

[0028] Specifically, the blood vessel structure image data includes the blood vessel central line segment, the trend of the blood vessel central line segment, and its three-dimensional coordinates in the coordinate system; use the midpoint of the blood vessel central line segment corresponding to the optimal puncture blood vessel sub-image (e.g., Figure 3 the line segment indicates the blood vessel central line segment, and the black dot on the line segment indicates the midpoint) as the puncture target point, and obtain the corresponding coordinates (x, y, z).

[0029] S43. Determine the puncture depth, the optimal needle insertion point and its corresponding coordinates based on the puncture target point and its corresponding coordinates, the puncture preset data, and the blood vessel structure image data.

[0030] Specifically, as Figure 3 shown, the puncture preset data includes the puncture angle β, usually 45 degrees, or other angles; starting from the puncture target point coordinate point, along the trend of the blood vessel central line segment as the extension direction, with the puncture angle β as the extension angle, perform a straight-line extension (e.g., Figure 3 the dotted line indicates), and the intersection point of the extension to the plane OXY and its corresponding coordinates are the optimal needle insertion point and its corresponding coordinates; calculate the puncture depth based on the coordinates of the puncture target point and the optimal needle insertion point using the principle of triangulation.

[0031] S44. Determine the puncture path with the optimal route based on the initial position coordinates of the B-ultrasound puncture integrated device, the optimal needle insertion point and its corresponding coordinates, the puncture target point and its corresponding coordinates, as well as the puncture angle and the puncture depth. As Figure 3 shown, the intersection point of the dotted line and OXZ is the initial position of the B-ultrasound puncture integrated device, and the dotted line indicates the puncture path. Of course, the Dijkstra algorithm can also be used to find the optimal puncture path in three-dimensional space; and use B-spline interpolation or moving average to smooth the path.

[0032] As Figure 4 shown, this solution also provides a portable blood vessel puncture device based on AI recognition of B-ultrasound images, which executes the portable blood vessel puncture method based on AI recognition of B-ultrasound images. The device includes a puncture assist robot and a host computer that are electrically connected; The puncture assist robot includes a control unit, a B-ultrasound puncture integrated device, a motor drive unit, and a pose adjustment unit that are electrically connected to the control unit; among them, the B-ultrasound puncture integrated device is arranged on the pose adjustment unit; The host computer includes a visualization unit, a data processing unit, an AI recognition unit, and a human-computer interaction unit that are electrically connected.

[0033] Specifically in application, the puncture assistance robot is fixedly installed at a position corresponding to the puncture body position and the surface to be scanned; human-machine interaction is carried out based on the host computer, and the puncture assistance robot is controlled through the host computer to execute the portable blood vessel puncture method based on AI recognition of B-ultrasound images, and the image data is displayed on the host computer.

[0034] Select a position far from the experienced puncture point and with a suitable distance according to the device operation mode to fix the puncture assistance robot, so that the experienced puncture point is within the movement limit range of the device, ensuring sufficient exposure of this area to fix the puncture device of this solution at a suitable body surface position of the selected puncture body position, facilitating subsequent B-ultrasound scanning and puncture operations, and improving the accuracy of subsequent process data acquisition and puncture.

[0035] The human-machine interaction unit is used for basic settings, including the operating parameters of the puncture assistance robot, coordinate system establishment, and sending instructions such as start to the control unit.

[0036] The control unit is used to receive instructions from the host computer, link and control the B-ultrasound puncture integrated device, the motor drive unit and the pose adjustment unit, and execute the portable blood vessel puncture method based on AI recognition of B-ultrasound images for puncture.

[0037] The visualization unit uses the visualization VTK technology to realize the visualization and rendering of blood vessels and puncture strategies on the host computer. The rendered content includes: blood vessel structure, puncture target points, puncture angles, depth marks, puncture path lines, etc., and different colors are used for visual distinction.

[0038] It can be understood that this device can fully execute the above method and effects, which will not be elaborated here.

[0039] The portable blood vessel puncture device and method based on AI recognition of B-ultrasound images provided in this embodiment, compared with the prior art, this solution realizes the full automation of blood vessel puncture to establish deep and shallow arteriovenous channels. By combining medical imaging technology, artificial intelligence algorithms and robot control, it realizes the full-process automated operation from blood vessel recognition, path planning to automatic puncture execution, improving the puncture success rate and safety. It also has the advantages of small size, controllable cost, convenient operation, etc., is applicable to a variety of clinical scenarios, and has broad promotion prospects and application value.

[0040] Embodiment 2 Different from Embodiment 1, the input of the AI recognition model also includes patient information and puncture scenario information.

[0041] Patient information includes but is not limited to age, gender, weight, medical history, coagulation function, etc.; puncture scenario information includes but is not limited to puncture site (such as radial artery, femoral artery, deep vein (such as subclavian vein, internal jugular vein), superficial vein (at a relatively shallow position on the body surface)), puncture purpose (such as catheter placement, sampling), equipment limitations, etc. Specifically, information is input through the human-computer interaction unit of the host computer, and after being processed by the data processing unit, it is input into the AI recognition unit for processing.

[0042] The AI recognition model also includes a multi-modal data fusion sub-model, which fuses the features of patient information, puncture scenario information, images, and puncture preset data, introduces an attention mechanism and an adaptive learning module, and can fine-tune the matrix scanning area accordingly according to different patient information and puncture scenario information, and then continue to determine the subsequent puncture strategy.

[0043] The portable vascular puncture device and method based on AI recognition of B-ultrasound images provided in this embodiment introduce more-dimensional data input during the AI recognition process. Through the multi-modal data fusion of patient information, puncture site information, and puncture scenario information, it can analyze individual differences and operating environments more comprehensively, so as to collect more appropriate image source data for different patient conditions from the source, generate more accurate and personalized puncture strategies, and improve the adaptability and clinical practicability of the plan.

[0044] Embodiment Three Different from Embodiments One and Two, during the process of obtaining and visualizing a number of static vascular sub-images arranged in a matrix and their corresponding coordinates, coordinate conversion is performed. The coordinate system constructed by this solution is the physical coordinate system.

[0045] The physical coordinate unit uses millimeters (mm), and is converted with the B-ultrasound image pixel coordinates through a calibration coefficient k (such as 1mm = k pixels). Let the displacement accuracy of the device robotic arm in the X-axis and Y-axis directions be δx and δy, and the position coordinates are recorded in real time through an encoder. The relationship between the actual scanning point coordinates (Xphys, Yphys) of the device and the physical dimensions (dx, dy) corresponding to the unit pixel of the robotic arm control parameter is: Xphys =dx×δx Yphys=dy×δy For the mapping between image coordinates and physical coordinates, assuming the pixel coordinates of the vascular sub-image are (u, v), the physical coordinate conversion formula is:

[0046]

[0047] Among them, 、 are respectively the coordinates of the pixel coordinates (u, v) of the vascular sub-image in the physical coordinate system; ( , ), which are the central pixel coordinates of the vascular sub-image, × is the pixel size of the vascular sub-image.

[0048] Use a dictionary or two-dimensional array to store the mapping between the sub-image and the coordinates:

[0049] Mark the physical coordinates for the key vascular structures, and the formula used is: The marked coordinates = (Xphys′, Yphys′) = ( ) The human-computer interaction unit of the host computer supports interactive coordinate query, and the physical position corresponding to the pixel can be obtained by clicking the mouse.

[0050] If there is a difference between the marked coordinates (Xphys′, Yphys′) of each vascular sub-image and the actual scanning point coordinates (Xphys, Yphys) of the device, it indicates that there is an offset in the coordinate system and calibration is required.

[0051] The portable vascular puncture device and method based on AI recognition of B-ultrasound images provided in this embodiment achieve high-precision mapping of spatial positioning by establishing an accurate conversion relationship between the physical coordinate system and the B-ultrasound image pixel coordinate system, effectively improving the accuracy of target recognition and path planning in the puncture operation, providing an important guarantee for the overall stability and reliability of the system; at the same time, it can also detect errors in the coordinate system. During long-term use, it is beneficial to calibrate the coordinate system based on the detected errors to improve the puncture success rate.

[0052] The above are only the embodiments of the present invention. Specific structures and common knowledge such as characteristics known to the public are not described in detail here. Those of ordinary skill in the art know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can know all the existing technologies in this field, and have the ability to apply the conventional experimental means before this date. Those of ordinary skill in the art can, under the inspiration given in this application and combined with their own abilities, improve and implement this solution. Some typical well-known structures or well-known methods should not become an obstacle for those of ordinary skill in the art to implement this application. It should be noted that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can still be made, and these should also be regarded as the protection scope of the present invention, and these will not affect the implementation effect of the present invention and the practicality of the patent.

Claims

1. A portable blood vessel puncture method based on AI recognition of B-ultrasound images, characterized in that The method includes the following steps: S1. Determine the puncture position and the surface to be scanned; S2. Construct a coordinate system; drive the B-ultrasound puncture integrated device to perform matrix scanning within the surface to be scanned based on a two-dimensional grid layout, and obtain and visualize a plurality of static blood vessel sub-images arranged in a matrix and their corresponding coordinates; S3. Process the images using the plurality of static blood vessel sub-images and their corresponding coordinates to obtain and visualize continuous blood vessel structure image data; S4. Construct an AI recognition model, take the blood vessel structure image data, a plurality of static blood vessel sub-images and their corresponding coordinates, and preset puncture data as inputs, output and visualize a puncture strategy, where the puncture strategy includes target puncture data and a puncture path determined based on the coordinate system; S5. Drive the B-ultrasound puncture integrated device to perform puncture along the puncture path.

2. The portable blood vessel puncture method based on AI recognition of B-ultrasound images according to claim 1, wherein In S1, determine the puncture position according to the puncture purpose and determine the empirical puncture point; use the range with the empirical puncture point inside or centered on the empirical puncture point as the surface to be scanned.

3. The portable blood vessel puncture method based on AI recognition of B-ultrasound images according to claim 1, wherein, In S2, construct a coordinate system. The initial scanning starting point of the B-ultrasound puncture integrated device is within the surface to be scanned. Take the initial scanning starting point of the B-ultrasound puncture integrated device as the coordinate origin O, define the scanning surface as a two-dimensional plane OXY, the direction of the X-axis is the direction for determining the blood vessel cross-sectional diameter, and the direction of the Y-axis is the direction for determining the blood vessel trend.

4. The portable blood vessel puncture method based on AI-recognized B-ultrasound images according to claim 3, wherein In S2, the path of the matrix scanning is as follows: adopt a row-first traversal strategy, start from the first row and the first column, scan along the X-axis to the first row and the Nth column; move along the Y-axis to the second row and the Nth column, and scan in the reverse direction to the second row and the first column; repeat the above process until the scanning of the Mth row is completed.

5. The portable blood vessel puncture method based on AI recognition of B-ultrasound images according to claim 1, wherein, In S2, use the coordinates of each scanning point position in the matrix scanning as the coordinates of the corresponding static blood vessel sub-image. After driving the B-ultrasound puncture integrated device to move to the corresponding position of the scanning point position coordinates and stabilizing, start B-ultrasound imaging by triggering a signal to obtain the corresponding static blood vessel sub-image.

6. The portable blood vessel puncture method based on AI-identified B-ultrasound images according to claim 1, wherein In S3, perform image registration on all the collected static blood vessel sub-images; and use an image stitching algorithm to perform image stitching to obtain a continuous blood vessel structure image, and at the same time extract the blood vessel center line and visualize it.

7. The portable blood vessel puncture method based on AI recognition of B-ultrasound images according to claim 1, characterized in that, In S4, the process of outputting the puncture strategy includes: S41. Determine the optimal puncture blood vessel sub-image among a plurality of static blood vessel sub-images; S42. On the optimal puncture blood vessel sub-image, determine the puncture target point and its corresponding coordinates based on the blood vessel structure image data; S43. Determine the puncture depth and the optimal needle insertion point and its corresponding coordinates based on the puncture target point and its corresponding coordinates, the puncture preset data, and the blood vessel structure image data; S44. Determine the puncture path with the best route based on the initial position coordinates of the B-ultrasound puncture integrated device, the optimal needle insertion point and its corresponding coordinates, the puncture target point and its corresponding coordinates, as well as the puncture angle and the puncture depth.

8. The portable blood vessel puncture method based on AI-identified B-ultrasound images according to claim 7, wherein, In S41, compare the curvature of the blood vessel trend and the blood vessel cross-sectional diameter in each blood vessel sub-image, and find the blood vessel sub-image with the smallest curvature of the trend and the largest cross-sectional diameter as the optimal puncture blood vessel sub-image.

9. The portable blood vessel puncture method based on AI recognition of B-ultrasound images according to claim 7, characterized in that, The blood vessel structure image data includes the blood vessel center line segment, the trend of the blood vessel center line segment, and its three-dimensional coordinates in the coordinate system; the puncture preset data includes the puncture angle; In S42, the midpoint of the central line segment of the blood vessel corresponding to the optimal puncture blood vessel sub-image is used as the puncture target point, and the corresponding coordinates are obtained; In S43, starting from the puncture target point, extending linearly along the direction of the central line segment of the blood vessel as the extension direction and with the puncture angle as the extension angle, and combining with the coordinate system, the optimal needle insertion point and its corresponding coordinates are determined, and at the same time, the puncture depth is calculated based on the coordinates of the puncture target point and the optimal needle insertion point.

10. A portable blood vessel puncture device based on AI recognition of B-ultrasound images, characterized in that, Execute the portable blood vessel puncture method based on AI recognition of B-ultrasound images according to any one of claims 1-9; the device includes a puncture assistance robot and a host computer that are electrically connected; The puncture assistance robot includes a control unit and a B-ultrasound puncture integrated device, a motor drive unit, and a pose adjustment unit that are electrically connected to the control unit; among them, the B-ultrasound puncture integrated device is arranged on the pose adjustment unit; The host computer includes a visualization unit, a data processing unit, an AI recognition unit, and a human-computer interaction unit that are electrically connected.

Images