Blood transfusion puncture navigation method and system based on three-dimensional image

Abstract

The present application relates to the technical field of image processing, in particular to a blood transfusion puncture navigation method and system based on three-dimensional images. The system takes the line connecting the venous target point coordinates and the real-time puncture needle tip coordinates as the center, and constructs the region of interest based on the puncture needle tip radius size. The puncture needle tip radius size is used as the lower limit to construct multiple scales. The maximum initial tubular tissue feature similarity is extracted combined with the curvature information, and the comprehensive tissue blocking feature value is obtained by fusing the gradient amplitude. The comprehensive tissue blocking feature value of the feature significant section is statistically screened to obtain the penetration resistance value of the real-time needle insertion path, which is used for state judgment. According to the judgment result, the warning signal and the vascular three-dimensional model visualization adjustment are fed back. The present application constructs a risk investigation and cross-section extreme value analysis mechanism based on the actual size constraint of the instrument, accurately quantifies the penetration resistance limit in front of the needle, and provides a highly reliable anti-misplacement safety guarantee for clinical ultrasound-guided puncture.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a transfusion puncture navigation method and system based on three-dimensional images. Background Technology

[0002] Ultrasound-guided deep vein puncture is a routine clinical procedure in emergency and critical care, massive blood transfusions, and deep vein catheterization. Currently, medical institutions typically use a navigation system combining three-dimensional ultrasound and spatial positioning tracking to establish deep vein access.

[0003] In real-world scenarios involving obese patients or lengthy catheter placement procedures, the target vessel is often surrounded by complex tubular tissues (such as accompanying arteries and dense nerve bundles). Due to fluctuations in the patient's respiratory rhythm or subtle body movements, the target vein and surrounding tissues experience continuous spatial displacement. When the target vessel deviates, existing navigation systems typically calculate the displacement directly based on the offset of the vein's centroid in the ultrasound image and forcefully translate and refresh the 3D navigation model to guide the puncture needle to follow the target.

[0004] However, this method of blindly updating coordinates based on target point displacement has serious technical risks: First, because ultrasound beams often generate anechoic dark areas when scanning parallel tubular tissues, existing systems rely solely on single brightness gradients or single fixed-scale feature extraction algorithms, making it extremely easy to miss tubular barriers of varying thicknesses. Second, in in-plane needle insertion operations, existing systems often use a global area averaging algorithm of regional pixels to assess the safety of the needle insertion path. Since extremely high-risk dense nerve cords often have very small cross-sections, their high-risk data features are diluted by the low values ​​of the surrounding vast safe soft tissue, causing the system to fail to identify local penetration obstacles ahead. These defects can lead the system to provide the operator with an incorrect 3D update trajectory that crosses dense nerves or accompanying arteries when the target point shifts. Summary of the Invention

[0005] To address the shortcomings of existing transfusion puncture navigation technologies, such as incomplete feature extraction leading to missed detection of tubular barrier features and the dilution of key salient features by global averaging algorithms, this invention aims to provide a transfusion puncture navigation method and system based on three-dimensional images. The specific technical solution adopted is as follows: This invention proposes a transfusion puncture navigation system based on three-dimensional images, the system comprising: A two-dimensional ultrasound information acquisition module is used to acquire real-time two-dimensional ultrasound images, which include the coordinates of the vein target point and the coordinates of the real-time puncture needle tip; a region of interest is constructed based on the radius of the puncture needle tip, with the line connecting the coordinates of the vein target point and the coordinates of the real-time puncture needle tip as the center. The tissue feature extraction module is used to construct multiple scales with the radius of the puncture needle tip as the lower limit; obtain the initial tubular tissue feature similarity of each pixel in the region of interest at each scale based on curvature information in different directions, and select the largest initial tubular tissue feature similarity as the final tubular tissue feature similarity; obtain the comprehensive tissue blocking feature value of each pixel based on the final tubular tissue feature similarity and gradient magnitude. The orthogonal scanning section analysis module is used to construct multiple orthogonal scanning sections along the centerline in the region of interest, statistically analyze the comprehensive tissue blocking feature values ​​on the orthogonal scanning sections to filter out sections with significant features, and obtain the real-time needle insertion path penetration blocking value based on the comprehensive tissue blocking feature values ​​on the sections with significant features. The puncture warning and navigation visualization module is used to determine whether a dangerous state of tissue obstruction has occurred based on the real-time needle insertion path penetration obstruction value, and to provide early warning signals and visual adjustments to the 3D vascular model based on the judgment results.

[0006] Furthermore, the method for constructing the region of interest includes: Using the connecting line as the center, the radius of the puncture needle tip is increased to both sides along the normal of the connecting line, and the resulting rectangular region is the region of interest.

[0007] Furthermore, the method for obtaining the initial tubular tissue feature similarity includes: For each pixel in the region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. At each scale, the scale is substituted into the exponential model of the Franji filter algorithm and the region of interest is smoothed. For each pixel in the smoothed region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. The initial tubular tissue feature similarity is calculated by comparing the second-order partial derivative eigenvalues.

[0008] Furthermore, the method for obtaining the comprehensive tissue blocking feature value includes: The final tubular tissue feature similarity and the gradient magnitude are weighted and summed according to a preset weight to obtain the comprehensive tissue blocking feature value; wherein the weight of the final tubular tissue feature similarity is greater than the weight of the gradient magnitude.

[0009] Furthermore, before constructing the region of interest, the following steps are also included: If the number of pixels on the connection line is less than a preset threshold, the real-time needle insertion path penetration blockage value is set to a preset minimum value, and the system directly enters the puncture warning and navigation visualization module.

[0010] Furthermore, the method for screening the salient cross-sections includes: The maximum comprehensive tissue blocking feature value on the orthogonal scanning section is used as a significant index. Based on the significant index, the sections are arranged in descending order, and a preset number of orthogonal scanning sections at the top of the arrangement are selected as the feature significant sections.

[0011] Furthermore, the method for obtaining the real-time needle insertion path penetration resistance value includes: The average significance index of the characteristic significant cross section is used as the real-time needle insertion path penetration blocking value.

[0012] Furthermore, the step of determining whether a tissue obstruction hazard has occurred based on the real-time needle insertion path penetration resistance value includes: If the real-time needle insertion path penetration resistance value is less than the preset safety resistance threshold, it is determined that there is no risk of tissue obstruction; otherwise, it is determined that there is a risk of tissue obstruction.

[0013] Furthermore, the step of feeding back warning signals based on the judgment results and adjusting the visualization of the vascular 3D model includes: If it is determined that there is no risk of tissue obstruction, a three-dimensional translation vector is constructed based on the offset between the initial vein target coordinates and the vein target coordinates in the real-time two-dimensional ultrasound image; the three-dimensional vascular model is then translated and updated based on the three-dimensional translation vector. If a risk of tissue obstruction is detected, a warning signal is fed back. The warning signal includes a cutoff command, which is used to stop the translation and updating of the three-dimensional vascular model.

[0014] The present invention also proposes a transfusion puncture navigation method based on three-dimensional images, which implements the workflow of any of the three-dimensional image-based transfusion puncture navigation systems described in the present invention.

[0015] The present invention has the following beneficial effects: This invention centers on the line connecting the vein target coordinates and the real-time puncture needle tip coordinates, and constructs a region of interest based on the puncture needle tip radius. This strictly limits the image processing range to the physical domain of actual instrument advancement, eliminating signal interference from surrounding irrelevant soft tissues. Multiple scales are constructed using the puncture needle tip radius as a lower limit. The maximum initial tubular tissue feature similarity is extracted by combining curvature information, and gradient amplitude is fused to obtain a comprehensive tissue blocking feature value. By introducing the actual instrument size as a detection lower limit, missed detection of high-risk tubular barriers of varying thicknesses is effectively avoided. The gradient amplitude fusion mechanism further suppresses background noise and accurately highlights the boundaries of real tissues. This invention constructs multiple orthogonal scanning sections along the centerline and statistically selects the comprehensive tissue blocking feature value of sections with significant characteristics to obtain the real-time needle insertion path penetration blocking value. It objectively recreates the physical process of cross-sectional resistance during rigid needle advancement, effectively overcoming the data dilution problem caused by traditional regional global averaging algorithms, preserving and amplifying the extreme signals of local fatal barriers. Finally, it achieves accurate tissue blocking hazard judgment and links warning signals with the visualization adjustment of the 3D vascular model. This invention establishes a risk screening and cross-sectional extreme value analysis mechanism based on the actual size constraints of the instrument, accurately quantifies the penetration resistance limit in front of the needle insertion, and provides a highly reliable safety guarantee against accidental puncture for clinical ultrasound-guided puncture. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a structural block diagram of a blood transfusion puncture navigation system based on three-dimensional images, provided as an embodiment of the present invention. Detailed Implementation

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of a three-dimensional image-based blood transfusion puncture navigation method and system provided by the present invention.

[0020] Please see Figure 1The diagram illustrates a structural block diagram of a three-dimensional image-based blood transfusion puncture navigation system according to an embodiment of the present invention. The system includes a two-dimensional ultrasound information acquisition module 101, a tissue feature extraction module 102, an orthogonal scanning section analysis module 103, and a puncture warning and navigation visualization module 104.

[0021] The system of this invention is applied in ultrasound-guided puncture clinical procedures such as emergency and critical care resuscitation or deep vein catheterization. Patients experience displacement of the target vein due to respiratory fluctuations, and the target vein is often surrounded by high-risk tubular tissues such as arteries or dense nerve bundles. The system monitors obstacles in front of the needle in real time during the insertion of the metal puncture needle into superficial tissue to prevent accidental puncture of barrier organs. The two-dimensional ultrasound information acquisition module 101 is linked with an ultrasound diagnostic instrument and an external spatial positioning and tracking system. The ultrasound probe scans the human body and outputs a real-time two-dimensional ultrasound image sequence to the navigation system. The spatial positioning and tracking system simultaneously acquires the pose matrix of the ultrasound probe and the medical puncture needle in three-dimensional physical space. The navigation system performs matrix multiplication mapping to determine the coordinates of the vein target point and the real-time puncture needle tip image in the two-dimensional image. The system retrieves the physical size parameters of the physical puncture needle and performs pixel conversion to construct the region of interest. The tissue feature extraction module 102 runs in the image computing unit of the navigation workstation. This module takes over the two-dimensional image grayscale matrix within the region of interest, calls internal filtering algorithms and gradient operators, and extracts pixel features using the previously input radius of the physical puncture needle. The orthogonal scanning section analysis module 103 runs in the computing unit of the navigation workstation. This module uses geometric matrix operations to generate discrete scan lines on a two-dimensional image feature base map, simulating the cross-sectional arrangement action of a physical puncture needle as it advances through physical tissue, and calculating the real-time needle penetration resistance value. The puncture warning and navigation visualization module 104 is linked with the underlying 3D rendering engine and the external display screen. When the system determines it is safe, this module sends a spatial translation vector to the 3D rendering engine, driving the 3D vascular model on the display screen to update its coordinates according to the patient's breathing displacement. When the system determines it is dangerous, this module sends a truncation command to the 3D rendering engine to freeze the vascular model mesh and calls the user interface to overlay a warning layer on the front of the display screen.

[0022] Specifically, the two-dimensional ultrasound information acquisition module 101 is used to acquire real-time two-dimensional ultrasound images and various positional information within the images. During actual puncture operations, the ultrasound probe pose matrix can be acquired simultaneously. This matrix characterizes the pose features of the probe when acquiring real-time two-dimensional ultrasound images. Similarly, the puncture needle pose matrix can also be obtained. These two pose matrices can be combined to derive the needle tip-to-ultrasound image projection matrix, representing their relative spatial transformation. Specifically, the ultrasound probe pose matrix can be inverted first, and the inverse result multiplied by the puncture needle pose matrix to obtain the needle tip-to-ultrasound image projection matrix.

[0023] This invention can utilize conventional texture recognition or image segmentation models to identify the coordinates of vein target points in real-time two-dimensional ultrasound images. Alternatively, these points can be directly annotated by medical personnel in the image; details will not be elaborated further. During the acquisition of real-time two-dimensional ultrasound images, the scale of the ultrasound image can be further obtained, which represents how much real-world physical distance each pixel can represent.

[0024] The actual coordinates of the puncture needle are spatial coordinates. In this embodiment of the invention, these coordinates need to be projected onto a real-time two-dimensional ultrasound image. Therefore, the spatial coordinates of the puncture needle tip can be multiplied by the projection matrix from the needle tip to the ultrasound image to complete the dimensionality mapping, generating a dimensionally reduced two-dimensional real-time puncture needle tip coordinate. Consequently, both the vein target coordinates and the real-time puncture needle tip coordinates exist simultaneously in the real-time two-dimensional ultrasound image.

[0025] Because the medical puncture needle used in practice is a rigid cylinder with a fixed outer diameter, when it punctures the superficial tissue of the patient, it will only exclude the anatomical structures directly in front of it and within its own width. To avoid interference from surrounding irrelevant muscle tissue in subsequent feature calculations, the system must construct a precise selection area based on the physical width and geometric orientation of the needle. Therefore, the region of interest can be constructed by using the line connecting the coordinates of the vein target point and the real-time coordinates of the puncture needle tip as the center, combined with the radius of the puncture needle tip. It should be noted that the radius of the puncture needle tip can be mapped to the two-dimensional ultrasound image dimension using the scale obtained above. That is, the information in the image processing process described in this embodiment of the invention is all information in the two-dimensional ultrasound image. The dimensional conversion process of information is a well-known technique to those skilled in the art and will not be elaborated or specifically limited here.

[0026] Preferably, in this embodiment of the invention, the method for constructing the region of interest includes: Centered on the connecting line, the radius of the puncture needle tip is expanded outwards along the normal of the connecting line, resulting in a rectangular region, which is the region of interest. The two longer sides of the final rectangular region are always parallel to the line connecting the needle insertion centers, while the two shorter sides pass through the two endpoints of the line connecting the needle insertion centers and are strictly perpendicular to the longer sides. Subsequent image processing is performed within the region of interest, avoiding the influence of irrelevant backgrounds.

[0027] Preferably, in some implementations of the embodiments of the present invention, considering that the essential purpose of the embodiments of the present invention is to assign a real-time needle path penetration resistance value in the area to be punctured between the needle tip and the target point, in order to assess the puncture risk of the area to be punctured, but if the puncture is in the later stage, it proves that the puncture is almost completed, and the distance between the needle tip and the target point is too small, thus losing the meaning of further analysis, therefore, before constructing the region of interest, the following is also included: If the number of pixels on the connecting line is less than a preset threshold, the real-time needle insertion path penetration obstruction value is set to a preset minimum value, and the system directly enters the puncture warning and navigation visualization module. This threshold can be set to 5 pixels.

[0028] Existing feature extraction algorithms based solely on a single brightness gradient or fixed scale in an image cannot effectively extract the true features of tubular tissue. Therefore, the tissue feature extraction module 102, based on gradient amplitude information, further determines the final tubular tissue feature similarity between each pixel location and the tubular tissue through multi-scale curvature analysis. Because tubular anatomical barriers such as arteries and nerve cords vary in thickness within the human body, using only a single fixed scale to detect image curvature may miss high-risk tubular tissues that are thicker or thinner than the puncture needle. To comprehensively prevent potential accidental puncture risks, the system must construct a set of detection scales covering the upper and lower safety limits, based on the size of the currently used instrument. Therefore, this embodiment of the invention first constructs multiple scales with the radius of the puncture needle tip as the lower limit. The construction of these scale sets enables subsequent tubular feature analysis to focus only on structures with radii greater than or equal to the thickness of the puncture needle. This is because in clinical physical puncture, tiny blood vessels or peripheral nerves that are thinner than rigid puncture needles are directly excluded and do not constitute substantial penetration resistance, nor are they high-risk or fatal barriers. By setting the lower limit of the scale in this way, the computational interference of these tiny tubular structures can be filtered out, allowing the system to focus on higher-scale information and avoid missed detections.

[0029] It should be noted that, in one specific implementation of this invention, the radius of the puncture needle tip is used as the lower limit, and the increment is set to 1 pixel until a preset upper limit is reached, thus obtaining all scales. This preset upper limit can be set according to the radius of the expected maximum high-risk tubular tissue, for example, 50 pixels, or half the width of the rectangular region of interest along the needle insertion path.

[0030] After obtaining the scale set, the initial tubular tissue feature similarity of each pixel in the region of interest can be obtained at each scale based on curvature information in different directions. This is because the pixel grayscale distribution of a two-dimensional ultrasound image constitutes a data surface. Blood vessels and nerve bundles appear as continuous band-like regions on this atlas surface. The geometric characteristics of tubular tissue determine its local grayscale anisotropy: along the direction of the tube's extension, the grayscale change inside or at the edge of the tissue is small, corresponding to a curvature feature value close to zero; along the cross-sectional direction perpendicular to the tube, there is a difference in grayscale between the inside and outside of the tissue, corresponding to a curvature feature value with a large absolute value. Therefore, by comparing curvature information in different directions, pixels with unidirectional high curvature characteristics can be identified as tubular structures, distinguishing them from patchy tissue with bidirectional high curvature and homogeneous soft tissue with bidirectional low curvature. In other words, the higher the initial tubular tissue feature similarity, the more likely the pixel at that scale corresponds to tubular tissue that obstructs the puncture needle.

[0031] For any given pixel, after obtaining the initial tubular tissue feature similarity at various scales, multi-scale screening can be performed. That is, the maximum initial tubular tissue feature similarity is selected as the final tubular tissue feature similarity, indicating that the pixel has more significant tubular features at the scale corresponding to the final tubular tissue feature similarity. If the final tubular tissue feature similarity is also greater, it means that the probability of the pixel belonging to tubular tissue is higher.

[0032] By further combining the final tubular tissue feature similarity with the gradient magnitude, the comprehensive tissue blocking feature value of each pixel can be obtained. That is, the larger the gradient magnitude of a pixel, the closer it is to the edge of the image, and the more likely it is to create resistance to needle insertion. At the same time, the higher the probability of belonging to tubular tissue, the greater the obstruction to needle insertion, and the larger the final comprehensive tissue blocking feature value.

[0033] Preferably, in some implementations of the present invention, considering that the two eigenvalues ​​obtained in the method of calculating curvature using the Hessian matrix can effectively characterize the curvature features in two directions—along the pipeline extension direction and along the cross-sectional direction perpendicular to the pipeline—the method for obtaining the initial tubular tissue feature similarity includes: For each pixel in the region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. At each scale, the scale is substituted into the exponential model of the Franji filter algorithm and the region of interest is smoothed. For each pixel in the smoothed region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. The initial tubular tissue feature similarity is calculated by comparing the second-order partial derivative eigenvalues.

[0034] In the Franji filter algorithm, each scale is used as the standard deviation parameter of the Gaussian kernel. Before calculating the second-order partial derivative of the Hessian matrix, the region of interest is first smoothed by convolution using the Gaussian kernel. The scale can determine the physical detection window of the Gaussian kernel, enabling the algorithm to filter out irrelevant frequencies and specifically match and enhance tubular structures with the same size as that scale. The specific algorithm is a well-known technique in the art and will not be described or limited here.

[0035] In one specific implementation of this invention, for the two second-order partial derivative eigenvalues ​​of the Hessian matrix, the morphology ratio factor is calculated by dividing the eigenvalue with the smaller absolute value by the eigenvalue with the larger absolute value. This factor is used to distinguish between linear and tubular structures; a smaller value indicates a higher likelihood of a tubular structure. The square root of the sum of the squares of the two second-order partial derivative eigenvalues ​​is used as the structure strength factor. This factor is used to distinguish between real anatomical structures and low-frequency background noise; a larger value indicates a higher likelihood of a real anatomical structure. The initial tubular tissue feature similarity can be obtained by further substituting into the following formula:

[0036] in Let denoted as , where 'r' is the morphological ratio factor and 's' is the normalized structural strength factor. Normalization can be achieved using range standardization, which is done by referencing the maximum and minimum values ​​from a historical database. If the data exceeds the maximum or minimum value from the historical data, it can be forcibly set to 1 or 0. In this formula, the morphological ratio factor is between 0 and 1. Therefore, a negative correlation is achieved by subtracting the morphological ratio factor from the positive integer 1. A smaller morphological ratio factor indicates that the curvature is more concentrated in a certain direction, making it more likely to be a tubular tissue. Simultaneously, a larger structural strength factor indicates that it is more likely to be a true anatomical structure, resulting in higher data reliability and a greater initial similarity of tubular tissue features. Furthermore, the final initial similarity of tubular tissue features is also between 0 and 1.

[0037] Preferably, in some implementations of the embodiments of the present invention, the method for obtaining comprehensive tissue blocking feature values ​​includes: The final tubular tissue feature similarity and the gradient magnitude are weighted and summed according to a preset weight to obtain the comprehensive tissue blocking feature value; wherein the weight of the final tubular tissue feature similarity is greater than the weight of the gradient magnitude, the purpose of which is to ensure that dense structures are filtered out and the comprehensive tissue blocking feature value focuses more on tubular features.

[0038] In one specific implementation of this invention, the gradient magnitude can be obtained using the Sobel operator, a technique well-known to those skilled in the art, and will not be elaborated upon here. Furthermore, considering that the final tubular tissue feature similarity value ranges between 0 and 1 in one specific implementation of this invention, the gradient magnitude can be further normalized. Normalization can also be performed using maximum and minimum value normalization, which will not be elaborated upon here. The weight of the final tubular tissue feature similarity can be set to 0.75, and the weight of the gradient magnitude to 0.25. Finally, through weighted summation, when the target point simultaneously possesses a clear tubular structure tendency and a strong edge drop (e.g., a typical arterial wall), its fusion value will approach 1; while in soft tissue gaps with only grayscale noise, a lower morphological weight can suppress gradient interference, keeping its value at a low level.

[0039] Because a medical metal puncture needle is a rigid cylinder with a fixed outer diameter, during its insertion into deep tissue, the needle tip only needs to push through and penetrate the tissue directly in front of its cross-section, without being interfered with or blocked by tissue in the more distant areas on either side of the needle path. To address this objective propagation law, the system objectively simulates this spatial penetration process by discretizing and segmenting the needle path. Therefore, the orthogonal scanning section analysis module 103 first constructs multiple orthogonal scanning sections along the centerline in the region of interest. An orthogonal section is defined as several sections in the region of interest perpendicular to the centerline (the line connecting the coordinates of the vein target and the real-time puncture needle tip). These sections can be formed by drawing perpendicular lines from each pixel on the centerline; the pixels through which these perpendicular lines pass in the region of interest constitute the orthogonal sections.

[0040] The orthogonal cross-section passes through several pixels, each corresponding to a comprehensive tissue blocking feature value. If the comprehensive tissue blocking feature values ​​of all orthogonal cross-sections are extracted and globally averaged, in cases with long puncture paths, a large number of low-segment cross-sections generated due to penetration of the safe fat layer will significantly lower the high-segment cross-sections generated when occasionally penetrating tough nerve cords, resulting in a lower final average score and causing the system to miss intervention alarms. To completely eliminate the dilution effect of large areas of safe soft tissue on local high-risk interference extreme values, the orthogonal scanning cross-section analysis module 103 further sets up a screening step, statistically analyzing the comprehensive tissue blocking feature values ​​on the orthogonal scanning cross-sections to screen for feature-significant cross-sections; these feature-significant cross-sections are considered to be the most realistic and influential cross-sections that can represent the current region of interest. Therefore, by further statistically analyzing the comprehensive tissue blocking feature values ​​on the feature-significant cross-sections, the real-time needle insertion path penetration blocking value can be obtained. This screening operation ensures that as long as there is any local high-risk interference in the needle insertion path that is sufficient to sever nerves or puncture hardened blood vessel walls, the real-time needle insertion path penetration resistance value can directly and explicitly increase the final average needle insertion path penetration resistance value, completely eliminating the side effect of long path area mean dilution.

[0041] Preferably, in some implementations of the present invention, the method for screening the feature-significant cross-section includes: The maximum comprehensive tissue blocking characteristic value on the orthogonal scanning cross-section is used as a salient index. Based on the salient index, the cross-sections are arranged in descending order, and a preset number of orthogonal scanning cross-sections at the top of the sorting results are selected as the salient cross-sections. This preset number can be set to 3.

[0042] Furthermore, the method for obtaining the real-time needle path penetration resistance value includes: The average significance index of the characteristic significant cross section is used as the real-time needle insertion path penetration blocking value.

[0043] The real-time needle insertion path penetration resistance value represents the puncture resistance in the current region of interest. If the real-time needle insertion path penetration resistance value is large, it indicates that further puncture may pose a risk, and a timely warning signal is needed to alert medical staff. Therefore, the puncture warning and navigation visualization module 104 can determine whether a tissue obstruction danger has occurred based on the real-time needle insertion path penetration resistance value, and provide a warning signal and adjust the 3D vascular model visualization based on the judgment result.

[0044] Preferably, in some implementations of the present invention, determining whether a tissue obstruction danger state occurs based on the real-time needle insertion path penetration resistance value includes: If the real-time needle insertion path penetration resistance value is less than the preset safety resistance threshold, it is determined that there is no risk of tissue obstruction; otherwise, it is determined that there is a risk of tissue obstruction.

[0045] It should be noted that the preset safety barrier threshold can be obtained based on the two-dimensional ultrasound image in the initial state, which is the initial moment immediately after puncture. For the two-dimensional ultrasound image at the initial moment, the initial vein target image coordinates can be extracted using the same method. After extracting the initial vein target coordinates, the skin contour line at the top of the image is located in the initial ultrasound image. After locating the skin contour line, the system establishes the shortest vertical connection between the initial vein target image coordinates and the skin contour line. After establishing this vertical connection, the system uses this vertical connection as the central axis and expands to both sides by a preset sampling pixel width (e.g., 20 pixels), thereby defining a safe sampling area in the initial ultrasound image. This area represents the normal distribution area of ​​fat and muscle soft tissue between the probe directly below and the target blood vessel. After defining the safe sampling area, the system extracts the comprehensive tissue blocking feature value of all pixels within the selected area and calculates the mean. The obtained mean is multiplied by a preset fluctuation coefficient for fine-tuning to obtain the safety barrier threshold. This fluctuation coefficient can be set to 1.2.

[0046] It should be noted that, based on the initial state feature extraction described above, a puncture state verification mechanism can also be set up. Since a two-dimensional ultrasound image is merely a cross-section of the three-dimensional physical world, if the external metal puncture needle is not within this cross-section (i.e., out-of-plane needle insertion), forcibly projecting the three-dimensional needle tip coordinates onto the two-dimensional image will cause the subsequently extracted tissue features to fail to represent the actual tissue ahead of the puncture. Therefore, three-dimensional angle verification must be performed before dimensionality reduction mapping. The system can acquire the initial ultrasound probe pose matrix and the puncture needle pose matrix from the initial ultrasound image synchronously acquired by the external spatial positioning sensor. After acquiring these two pose matrices, the three-dimensional normal vector perpendicular to the direction of the probe piezoelectric crystal array is extracted from the initial ultrasound probe pose matrix. Simultaneously, the system extracts the three-dimensional direction vector along the major axis of the puncture needle from the puncture needle pose matrix. The system calculates the spatial physical angle between two vectors using the dot product formula of spatial geometric vectors and an inverse cosine operation. The system then determines whether this angle falls within a preset coplanar tolerance range. This tolerance range can be set between 85 and 95 degrees. When the angle falls within this range, the system determines that the long axis of the puncture needle is almost parallel to the ultrasound scanning plane, confirming that the current operation is in a safe in-plane puncture state. Upon confirmation of this state, the system triggers a release command, allowing operation of the two-dimensional ultrasound information acquisition module and subsequent modules; otherwise, the current navigation calculation is suspended, and a warning signal is sent to prompt the operator to adjust the probe or needle angle.

[0047] Preferably, in this embodiment of the invention, the feedback of early warning signals and the visualization adjustment of the vascular 3D model based on the judgment result includes: If it is determined that there is no risk of tissue obstruction, a three-dimensional translation vector is constructed based on the offset between the initial vein target coordinates and the vein target coordinates in the real-time two-dimensional ultrasound image; the three-dimensional vascular model is then translated and updated based on the three-dimensional translation vector.

[0048] Specifically, in one specific implementation of this invention, the three-dimensional translation vector takes the form of: ,in In the dimension of a two-dimensional ultrasound image, the x-coordinate offset between the initial state coordinates of the vein target point and the real-time coordinates of the vein target point in the two-dimensional ultrasound image. This represents the ordinate offset between the initial vein target coordinates and the vein target coordinates in the real-time 2D ultrasound image. Both 2D coordinate offsets can be obtained by subtracting the initial vein target coordinates from the vein target coordinates in the real-time 2D ultrasound image. The first 0 in the 3D translation vector represents the z-axis offset component. Because the 2D ultrasound section cannot capture anatomical displacement perpendicular to the probe scanning section (i.e., the probe thickness direction), it can be forcibly set to 0 to ensure that the unknown dimension does not produce erroneous coordinate flyaways. The trailing 0 is a standard homogeneous coordinate constant added in computer graphics to ensure that the 3D translation vector conforms to the penalty rules of standard affine transformations. The final 1×4 three-dimensional translation vector is multiplied with the real-time ultrasound probe pose matrix to obtain the target three-dimensional translation vector. This target three-dimensional translation vector is then applied to the underlying three-dimensional engine. The rendering engine, based on the polygon vertex translation algorithm, superimposes the three-dimensional components of the target three-dimensional translation vector onto the coordinates of all vertices in the vascular three-dimensional model mesh set. After the calculation is completed, the updated three-dimensional model can be pushed to the visualization display screen to present the operator with a real-time navigation screen that safely follows the patient's breathing displacement.

[0049] If a risk of tissue obstruction is detected, indicating that due to organ slippage, a dense tubular structure exceeding the safe tolerance range has appeared directly in front of the new target point connection (e.g., the puncture needle will cross an impenetrable accompanying artery wall), a warning signal is issued. This warning signal includes a cutoff command, which stops the translation and updating of the 3D vascular model. After the cutoff command is issued, the system can retrieve the 3D vascular model stored in memory at the previous safe state as the current 3D output source. At this time, the coordinates of the 3D vascular model on the external display screen will freeze, actively blocking the coordinate system's follow-up refresh, visually depriving the operator of the guidance to blindly insert the needle into the wrong target point. Simultaneously, the system extracts pre-recorded warning layer material (e.g., a static image with a red RGB color value and a transparency channel). After extracting this red warning layer, the system calls the rendering engine's user interface (UI) top-level overlay interface. Through this interface, the system directly overlays and renders the red warning layer above the virtual needle insertion ray area at the forefront of the display screen. By combining the image freeze with the overlay of a red layer, the system ultimately completed the application-layer visual interception of tissue penetration interference anomalies.

[0050] In summary, this invention uses the line connecting the vein target coordinates and the real-time puncture needle tip coordinates as the center, and constructs a region of interest based on the puncture needle tip radius. Multiple scales are constructed using the puncture needle tip radius as a lower limit. The maximum initial tubular tissue feature similarity is extracted by combining curvature information, and gradient amplitude is fused to obtain a comprehensive tissue obstruction feature value. Multiple orthogonal scanning sections are constructed along the centerline, and the comprehensive tissue obstruction feature values ​​of sections with significant characteristics are statistically selected to obtain the real-time needle insertion path penetration obstruction value. The real-time needle insertion path penetration obstruction value is used to determine whether a tissue obstruction danger state has occurred, and a warning signal is fed back based on the judgment result, along with visual adjustments to the 3D vascular model. This invention constructs a risk screening and cross-sectional extreme value analysis mechanism based on the actual size constraints of the instrument, accurately quantifying the penetration resistance limit in front of the needle insertion, and providing a highly reliable safety guarantee against accidental puncture for clinical ultrasound-guided puncture.

[0051] Based on the same inventive concept, this invention also proposes a transfusion puncture navigation method based on three-dimensional images, which implements the workflow of any one of the three-dimensional image-based transfusion puncture navigation systems described above.

[0052] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A transfusion puncture navigation system based on three-dimensional images, characterized in that, The system includes: A two-dimensional ultrasound information acquisition module is used to acquire real-time two-dimensional ultrasound images, which include the coordinates of the vein target point and the coordinates of the real-time puncture needle tip; a region of interest is constructed based on the radius of the puncture needle tip, with the line connecting the coordinates of the vein target point and the coordinates of the real-time puncture needle tip as the center. The tissue feature extraction module is used to construct multiple scales with the radius of the puncture needle tip as the lower limit; obtain the initial tubular tissue feature similarity of each pixel in the region of interest at each scale based on curvature information in different directions, and select the largest initial tubular tissue feature similarity as the final tubular tissue feature similarity; obtain the comprehensive tissue blocking feature value of each pixel based on the final tubular tissue feature similarity and gradient magnitude. The orthogonal scanning section analysis module is used to construct multiple orthogonal scanning sections along the centerline in the region of interest, statistically analyze the comprehensive tissue blocking feature values ​​on the orthogonal scanning sections to filter out the sections with significant features, and obtain the real-time needle insertion path penetration blocking value based on the comprehensive tissue blocking feature values ​​on the sections with significant features. The puncture warning and navigation visualization module is used to determine whether a dangerous state of tissue obstruction has occurred based on the real-time needle insertion path penetration obstruction value, and to provide early warning signals and adjust the 3D model of the blood vessel based on the judgment results.

2. The transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The method for constructing the region of interest includes: Using the connecting line as the center, the radius of the puncture needle tip is increased to both sides along the normal of the connecting line, and the resulting rectangular region is the region of interest.

3. The transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The method for obtaining the initial tubular tissue feature similarity includes: For each pixel in the region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. At each scale, the scale is substituted into the exponential model of the Franji filter algorithm and the region of interest is smoothed. For each pixel in the smoothed region of interest, a Hessian matrix centered on the pixel is obtained, and two second-order partial derivative eigenvalues ​​of the Hessian matrix are extracted. The initial tubular tissue feature similarity is calculated by comparing the second-order partial derivative eigenvalues.

4. The transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The method for obtaining the comprehensive tissue blocking feature value includes: The final tubular tissue feature similarity and the gradient magnitude are weighted and summed according to a preset weight to obtain the comprehensive tissue blocking feature value; wherein the weight of the final tubular tissue feature similarity is greater than the weight of the gradient magnitude.

5. The transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, Before constructing the region of interest, the following steps are also included: If the number of pixels on the connection line is less than a preset threshold, the real-time needle insertion path penetration blockage value is set to a preset minimum value, and the system directly enters the puncture warning and navigation visualization module.

6. The transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The screening method for the feature-significant cross-section includes: The maximum comprehensive tissue blocking feature value on the orthogonal scanning section is used as a significant index. Based on the significant index, the sections are arranged in descending order, and a preset number of orthogonal scanning sections at the top of the arrangement are selected as the significant feature sections.

7. A transfusion puncture navigation system based on three-dimensional images according to claim 6, characterized in that, The method for obtaining the real-time needle insertion path penetration resistance value includes: The average significance index of the characteristic significant cross section is used as the real-time needle insertion path penetration blocking value.

8. A transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The step of determining whether a tissue obstruction hazard has occurred based on the real-time needle insertion path penetration resistance value includes: If the real-time needle insertion path penetration resistance value is less than the preset safety resistance threshold, it is determined that there is no risk of tissue obstruction; otherwise, it is determined that there is a risk of tissue obstruction.

9. A transfusion puncture navigation system based on three-dimensional images according to claim 1, characterized in that, The step of feeding back early warning signals and adjusting the visualization of the 3D vascular model based on the judgment result includes: If it is determined that there is no risk of tissue obstruction, a three-dimensional translation vector is constructed based on the offset between the initial vein target coordinates and the vein target coordinates in the real-time two-dimensional ultrasound image; the three-dimensional vascular model is then translated and updated based on the three-dimensional translation vector. If a risk of tissue obstruction is detected, a warning signal is fed back. The warning signal includes a cutoff command, which is used to stop the translation and updating of the three-dimensional vascular model.

10. A transfusion puncture navigation method based on three-dimensional images, characterized in that, The method implements the workflow of a three-dimensional image-based blood transfusion puncture navigation system as described in any one of claims 1 to 9.

Images