Puncture entrance site intelligent identification method and system based on ultrasonic image

Abstract

The invention discloses a puncture entrance site intelligent identification method and system based on an ultrasonic image, and relates to the technical field of ultrasonic image identification. According to the method, image edge identification and verification are carried out for a real-time dynamic image flow of an ultrasonic image to obtain spatial form parameters corresponding to a target puncture area, and then spatial position quantification processing is carried out based on the obtained spatial form parameters to obtain the shortest distance between positions to be punctured in the target area. Synchronously judging whether the distance is a risk distance, generating a risk distance matrix, carrying out adaptive positioning calibration on a puncture path, finally, identifying a probe edge mark point in an ultrasonic image, and determining a potential entry point based on prediction process data of a puncture needle space trajectory. And meanwhile, candidate screening is carried out, and the inlet site after candidate screening is displayed on the ultrasonic image, so that the problem of inaccurate puncture inlet site positioning caused by insufficient accuracy of target area boundary segmentation in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of ultrasound image recognition technology, and in particular to a method and system for intelligent identification of puncture site based on ultrasound images. Background Technology

[0002] The determination of the insertion tip location of a peripherally inserted central catheter (PICC) combines ultrasound imaging technology with artificial intelligence algorithms. This precise puncture site determination aims to improve the safety and efficiency of PICC placement. A suitable detection device performs multi-dimensional scanning of the target area to acquire clear raw image data. The system then performs a series of optimization processes on the raw images to remove interference factors and highlight the boundaries and internal features of the target area, providing high-quality data support for subsequent identification. Next, the intelligent algorithm analyzes the processed image based on a preset feature model, accurately locating the position, range, and surrounding environment of the target site, and plans a reasonable operation path based on spatial coordinate calculations. In the subsequent site confirmation stage, multi-technology collaborative feedback dynamically verifies and fine-tunes the preliminary positioning results, ensuring the accuracy and stability of site positioning, forming a complete process from image acquisition, processing, identification to location verification.

[0003] For example, Chinese invention patent application CN115359238A discloses a method for identifying and locating puncture target points for intravenous puncture, including: acquiring images of the left and right arms using a binocular camera, saving them as images captured by the left camera and the right camera respectively, and sequentially performing support vector machine detection, convolutional neural network feature matching, and grayscale histogram analysis to obtain the rectangular envelopes, matching feature points, and grayscale information of the left and right arms; selecting the optimal matching feature points for intravenous puncture on these images, and calculating the three-dimensional coordinates of the optimal puncture target point using the calibration parameters of the binocular camera.

[0004] For example, Chinese invention patent CN113011333B discloses a system and method for obtaining the optimal vein puncture point and direction based on near-infrared images. This includes: using a bilateral filtering method to denoise the acquired infrared image of the back of the hand, and dynamically selecting the Region of Interest (ROI) based on the contour features of the back of the hand; designing a vein feature enhancement algorithm based on the Hessian matrix, using a second-order differential operator to differentiate with the ROI image to obtain the Hessian matrix, constructing a vein enhancement filtering function using its eigenvalues ​​and eigenvectors, obtaining local vein features, extracting the complete vein network structure, and developing a vein width feature visualization algorithm to screen the optimal vein puncture point and puncture orientation.

[0005] The above-mentioned technology has at least the following technical problems: When acquiring images of the puncture area, due to structural similarity, the obtained ultrasound images are prone to problems such as speckle noise, artifacts, and insufficient contrast. This leads to overlapping boundary contours of similar structures, reduced recognizability, and difficulty in accurately delineating their independent ranges. Simultaneously, the dynamic spatial morphological parameters of the target area and the differentiated characteristics of similar structures (such as subtle texture differences, gray-level gradient distribution patterns, and dynamic deformation characteristics) result in inaccurate responses to changes in structural features in real-time dynamic images. This makes it impossible to promptly capture the dynamic differences in similar structures caused by puncture operations or tissue movement, further exacerbating the confusion in identifying similar structures. All these factors contribute to insufficient accuracy in target area boundary segmentation, ultimately leading to inaccurate localization of the puncture entry point. Summary of the Invention

[0006] To address the problem of inaccurate puncture site localization in existing technologies, this invention provides a method and system for intelligent identification of puncture site based on ultrasound images. The technical solution is as follows:

[0007] On the one hand, an intelligent identification method for puncture entry points based on ultrasound images is provided. This method includes: Step 1, performing image edge recognition and verification on the real-time dynamic image stream of ultrasound images to obtain spatial morphological parameters corresponding to the target puncture area. The spatial morphological parameters are used to quantify the relative positional relationship and dynamic deformation law of the three-dimensional geometric features of the puncture sites in the target puncture area; Step 2, performing spatial position quantization processing based on the obtained spatial morphological parameters to obtain the shortest distance between the puncture sites in the target area, simultaneously determining whether it is a risk distance, generating a risk distance matrix, and performing adaptive positioning calibration of the puncture path to improve the accuracy of puncture path positioning in the target puncture area; Step 3, identifying probe edge markers in the ultrasound image, determining potential entry points based on the prediction process data of the puncture needle spatial trajectory, and simultaneously performing candidate screening and displaying the candidate entry points on the ultrasound image.

[0008] On the other hand, an intelligent identification system for puncture entry points based on ultrasound images is provided. This system applies methods such as intelligent identification of puncture entry points based on ultrasound images. The system includes: an image edge recognition and verification module, a spatial position quantization processing module, and an entry point screening module. The image edge recognition and verification module is used to perform image edge recognition and verification on the real-time dynamic image stream of ultrasound images. The spatial position quantization processing module is used to perform spatial position quantization processing based on the acquired spatial morphological parameters, simultaneously determine whether it is a risk distance, and generate a risk distance matrix for adaptive positioning calibration of the puncture path. The entry point screening module is used to identify probe edge markers in ultrasound images, determine potential entry points based on the prediction process data of the puncture needle spatial trajectory, and simultaneously perform candidate screening and display the candidate-screened entry points on the ultrasound image.

[0009] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: 1. By combining image edge recognition and verification with dynamic adjustment of connected components, the problem of low accuracy in target region boundary segmentation is solved. In the image edge recognition stage, a keyframe filtering mechanism is adopted. Frames with violent motion are eliminated by calculating the structural similarity of adjacent sequence frames, and stable frames are selected based on the proportion of pixel displacement, ensuring good stability of the image data used for subsequent analysis. In edge enhancement processing, the scale of the edge detection operator is dynamically adjusted to adapt to edge features of different scales, effectively eliminating edge thickening and improving pixel-level edge positioning accuracy. On this basis, the adjustment of connected components can distinguish effective connected components related to the target region edge and accurately eliminate invalid connected components formed by noise and artifacts. This synergy not only solves the boundary overlap problem but also adapts to different ultrasound images and morphological fluctuations of the target region, making the target region edge segmentation more closely resemble the real morphology, providing accurate basic data support for subsequent spatial location quantification and puncture path planning.

[0010] 2. By quantifying spatial location, a risk distance matrix is ​​constructed, improving the comprehensiveness of risk assessment. In the process of spatial location quantification, the spatial morphological parameters of the target area are used as input to clearly delineate the boundaries between key structures and the target area, avoiding misjudgment of risk areas due to overlapping outlines of similar structures. Based on the edge coordinate sequence, the central axis of the target area is fitted and uniformly sampled to accurately quantify the shortest distance between the puncture site and each avoidance area. Then, through matrix processing, a risk distance matrix is ​​formed, which comprehensively presents the spatial positional relationship between the target area and all risk structures, realizing accurate identification of risk areas and providing comprehensive and quantitative basis for puncture path planning.

[0011] 3. By using risk distance screening and dynamic stability verification, the stability of puncture entry point screening in existing technologies is improved. In the candidate screening process, the shortest risk distance in the risk distance matrix is ​​used for initial screening, and the coordinate changes of the entry point in the ultrasound images of a preset number of frames are tracked based on the extracted texture features of the neighboring target area corresponding to the entry point to obtain the corresponding dynamic stability for secondary screening. This achieves the complementarity of spatial and dynamic stability. Risk distance screening ensures the static safety of the puncture path, while dynamic stability verification considers the impact of dynamic factors on the entry point during puncture positioning. This ensures that the entry point will not touch the risk structure and maintains a stable relative position with the target area during puncture positioning, reducing the risk of puncture deviation caused by dynamic displacement of the site.

[0012] 4. By accurately extracting texture features from neighboring target areas and establishing a mapping relationship between features and puncture parameters, the adaptability of puncture parameters is improved. In the texture feature extraction stage, a circular total region is constructed to ensure that the extracted texture features can accurately reflect the actual situation of the tissues adjacent to the target area. By dynamically adjusting the adaptive filtering intensity based on the neighborhood variance deviation, the authenticity and reliability of texture features are guaranteed. By combining the operator scale gradient to generate a gradient map, representative texture pixels are selected to achieve accurate extraction of texture features. On this basis, the puncture parameters are dynamically adjusted according to the characteristics of neighboring target areas, reducing the probability of puncture positioning errors caused by the mismatch between puncture parameters and regional characteristics, and improving the smoothness and safety of puncture. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0014] Figure 1 A flowchart illustrating the intelligent identification method for puncture site based on ultrasound images provided in this embodiment of the invention; Figure 2 This is a flowchart illustrating the image edge recognition and verification process provided in an embodiment of the present invention. Figure 3 The flowchart corresponding to the spatial position quantization process provided in the embodiments of the present invention; Figure 4 A flowchart corresponding to the entry site screening provided in this embodiment of the invention; Figure 5 This is a schematic diagram of the structure of the intelligent identification system for puncture site based on ultrasound images provided in an embodiment of the present invention. Detailed Implementation

[0015] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0016] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0017] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] This invention provides a method for intelligent identification of puncture site based on ultrasound images. For example... Figure 1 The flowchart shown is for an intelligent identification method for puncture entry points based on ultrasound images. The processing flow of this method can include the following steps: Step 1: For the real-time dynamic image stream of the ultrasound image, perform image edge recognition and verification to obtain the spatial morphological parameters corresponding to the target puncture area. The spatial morphological parameters are used to quantify the relative positional relationship and dynamic deformation law of the three-dimensional geometric features of the puncture sites in the target puncture area; Step 2: Based on the obtained spatial morphological parameters, perform spatial position quantization processing to obtain the shortest distance between the puncture sites in the target area, simultaneously determine whether it is a risk distance that needs to be avoided in puncture positioning, generate a risk distance matrix, and perform adaptive positioning calibration of the puncture path to improve the accuracy of puncture path positioning in the target puncture area; Step 3: Identify probe edge markers in the ultrasound image, determine potential entry points based on the prediction process data of the puncture needle spatial trajectory, and simultaneously perform candidate screening and display the candidate entry points on the ultrasound image.

[0020] In this embodiment, during actual puncture operations, the application of this method allows for accurate capture of the spatial morphological parameters of the target puncture area through image edge recognition and verification, clearly quantifying the three-dimensional geometric relationship and dynamic deformation patterns of the puncture site. This avoids feature extraction deviations caused by image fluctuations or interference. Based on these parameters, spatial position quantification processing accurately obtains the distance between puncture sites and determines risk distances. The generated risk distance matrix assists in calibrating the puncture path, ensuring that the path avoids anatomical structures that need to be avoided. Simultaneously, by identifying probe edge markers in the ultrasound image and combining them with puncture needle spatial trajectory prediction, potential entry points are determined. After candidate screening, suitable entry points are clearly displayed on the image, providing intuitive and reliable positioning references for the personnel, helping to reduce deviations in puncture entry point positioning, and improving the overall safety and efficiency of the puncture operation.

[0021] like Figure 2The diagram shows a flowchart of image edge recognition and verification provided in an embodiment of the present invention. The effectiveness of deviation calibration is verified by determining key frames, and then operator scale gradient coverage is performed to ensure that the ultrasound image can achieve full coverage of the operator scale. At the same time, the pixel with the strongest pixel edge intensity in the pixel group is retained. After operator scale gradient coverage, the connected component is dynamically adjusted. The threshold of the connected component parameter is increased, decreased or kept unchanged according to the obtained connected component parameters. The spatial morphology parameters of the connected component are determined after the threshold is adjusted.

[0022] Further, image edge recognition and verification are performed. The specific process is as follows: acquire the sequence frames corresponding to the real-time dynamic image stream and calculate the structural similarity between adjacent sequence frames; if the structural similarity is greater than the preset structural similarity, key frames are determined; otherwise, they are judged as frames with violent motion, and adjacent sequence frames are reselected; key frame determination uses a simplified version of a lightweight network to calculate the proportion of regions with statistical pixel displacement greater than the preset pixel value, and selects adjacent sequence frames with the proportion of regions greater than the preset proportion of regions as stable frames; based on the acquired stable frames, retain the stable frames corresponding to the preset number of consecutive frames greater than the preset proportion of regions as key frames; if there are no consecutive stable frames that meet the standard, adjust the inter-frame pixel fusion ratio according to the deviation value between the current number of stable frames and the preset number of consecutive frames, compensate for the motion deviation of a single frame by correcting the mean of the corresponding pixels, and verify the effectiveness of the deviation calibration at the same time.

[0023] The specific process for verifying the effectiveness of deviation calibration is as follows: Based on the ultrasound image after motion deviation correction, the edges of the ultrasound image are covered with gradient regions using a preset operator scale to obtain the pixel edge intensity used to quantify the degree of change of a single pixel along the gradient direction; if the obtained pixel edge intensity is less than the preset pixel edge intensity, the operator scale of the edge detection operator is dynamically adjusted to cover edge features of different scales; the edge direction in the ultrasound image is determined according to the preset gradient direction, and the corresponding adjacent edge pixels in the same direction are divided into edge pixel groups; the edge pixel with the largest pixel edge intensity in the edge pixel group is obtained, while the remaining edge pixels in the group are set to 0 to improve edge positioning accuracy and eliminate edge thickening phenomenon; the connected components of the image edges obtained after image edge detail enhancement are dynamically adjusted.

[0024] Specifically, the dynamic adjustment process for connected components is as follows: The connected component parameters displayed at the image edges are compared with their corresponding reference connected component parameters. The connected component parameters include connected component length and curvature, while the reference connected component parameters include reference connected component length and curvature. Regarding the connected component length, if the obtained connected component length is greater than the reference connected component length, an increase in the connected component length threshold is obtained based on the length deviation to determine whether the connected component is an edge of the target region. If the obtained connected component length is less than the reference connected component length, a decrease in the connected component length threshold is obtained. This process distinguishes effective connected components related to the target region edge, eliminates invalid connected components formed by noise and artifacts, and improves the performance of different ultrasound images. The target's adaptability to fluctuations; for connected region curvature, if the obtained connected region curvature is greater than the reference connected region curvature, an increase in the connected region curvature threshold is obtained based on the connected region curvature deviation; if the obtained connected region curvature is less than the reference connected region curvature, a decrease in the connected region curvature threshold is obtained, so as to accurately screen the morphological trend of the connected regions; if the obtained connected region length is equal to the reference connected region length, or the obtained connected region curvature is equal to the reference connected region curvature, no adjustment is made to the connected region length threshold and the connected region curvature threshold; based on the adjusted connected region parameter thresholds, the effective connected regions in the ultrasound image are obtained, and the spatial morphological parameters in the effective connected regions are obtained. The effective connected regions represent connected regions whose morphology and location are related to the target region.

[0025] The process of obtaining spatial morphological parameters in the effective connected domain is as follows: extract the edge pixel coordinates in the effective connected domain, take the gradient direction of each edge pixel as the starting direction, and form a continuous edge coordinate sequence in a series form; during the formation of the edge coordinate sequence, obtain the path length corresponding to the edge coordinate sequence and update it to the target region length; use the edge coordinate sequence and the target region length together as spatial morphological parameters that characterize the spatial morphology of the target region.

[0026] In this embodiment, the preset structural similarity is the result of summing and averaging the historical structural similarities in the historical image edge recognition and verification process. The preset pixel values ​​are pre-set based on the resolution of the current ultrasound image and the motion amplitude of the target area. The preset area accounts for 80%, and the preset consecutive number is usually 3, but it is not fixed. The preset personnel can fine-tune it according to the current target area. The inter-frame pixel fusion ratio is based on a simplified version of a lightweight network and a sample library containing stable frame deviations and corresponding adjustment values. Through training, the network learns the mapping relationship between deviation features and adjustment values, so that it takes the consecutive number of deviation values ​​as input and the fusion ratio adjustment value as output. In practical applications, the verification results are fed back to the network to relearn the adjustment values, so as to achieve accurate compensation for single-frame motion deviations. The preset operator scale and its corresponding adjustment value are typically selected within the range of 3x3 to 9x9, dynamically adapted according to the actual situation of the specific ultrasound image; the connected component length deviation is represented by the difference between the acquired connected component length and the reference connected component length, the connected component curvature deviation is represented by the difference between the acquired connected component curvature and the reference connected component curvature, the reference connected component parameter is determined by statistical analysis of effective parameters under the same scene, combined with the characteristics of the target region, and the connected component threshold adjustment is based on the acquired connected component deviation value as the threshold adjustment value. For example, if the current connected component length data is 2mm and the reference connected component length data is 3mm, then the connected component length data threshold adjustment value is 1mm.

[0027] By optimizing the image edge recognition and verification process, the problem of inaccurate edge recognition in real-time dynamic ultrasound image streams is solved, improving the stability and accuracy of edge features in the target region. Firstly, by judging the structural similarity of sequence frames and filtering keyframes, image frames with violent motion and unstable features can be eliminated, ensuring that subsequent analysis is based only on stable images with reliable features, and avoiding interference from blurred or deformed edges in unstable frames.

[0028] When no consecutive stable frames meet the standards, dynamically adjusting the inter-frame pixel fusion ratio and correcting the pixel mean to compensate for single-frame motion deviation can maximize the offset of image shift caused by motion, restore the true edge shape of the target area, and avoid edge misalignment or breakage caused by motion deviation. In the verification of the effectiveness of deviation calibration, the dynamic adjustment of the edge detection operator scale can flexibly adapt to edge features of different scales; whether it is a fine branch edge of the target area or a thicker trunk edge, it can be accurately captured. At the same time, the screening and zeroing of edge pixel groups can eliminate the edge thickening phenomenon commonly found in traditional edge detection, so that the edge positioning accuracy reaches the pixel level, avoids feature misjudgment caused by edge blurring and overlap, and provides a clear, complete and accurately positioned edge contour for subsequent connected component analysis, reducing the error of target area feature extraction from the source and providing accurate target area boundaries for positioning.

[0029] Optimizing the dynamic adjustment of connected components improves the accuracy and adaptability of effective connected component selection, ensuring that the extracted spatial morphological parameters truly reflect the actual characteristics of the target area and providing reliable data support for subsequent puncture-related analyses. By comparing the length and curvature of connected components with corresponding reference parameters and dynamically adjusting the threshold based on the deviation, it can flexibly adapt to different scenarios. It avoids missing subtle edge connected components that match the target characteristics due to excessively high thresholds, and also avoids including invalid connected components formed by noise and artifacts due to excessively low thresholds, thus achieving accurate selection of edge connected components of the target area. It effectively avoids parameter distortion caused by interference from invalid connected components, providing accurate target area parameters for subsequent spatial location quantification, risk distance calculation, and puncture path planning. From a data perspective, it reduces the deviation in puncture entry point positioning and improves the reliability of the overall puncture planning.

[0030] like Figure 3 The flowchart shown is a flowchart of the spatial location quantization process provided in the embodiment of the present invention. By determining the spatial morphological parameters of the connected domain, the segmentation mask corresponding to the target area and the preset avoidance area is obtained. At the same time, the Euclidean distance and echo intensity are obtained to calibrate the risk distance to obtain the calibrated risk distance and construct the risk distance matrix. Then, the texture features of the adjacent target area are obtained and adaptive filtering is performed to determine the integrity of the texture features. The puncture parameters are adjusted and fed back to the preset personnel.

[0031] Further, the shortest distance between puncture sites in the target region is obtained. The specific steps are as follows: Using the obtained spatial morphological parameters as input, and based on a spatial morphological region model employing a multi-class semantic segmentation network, a segmentation mask is output for the target puncture region and a preset avoidance region. The preset avoidance region represents the risk structure area that needs to be avoided in the puncture path, including key anatomical structures such as blood vessels, nerves, and the edges of important organs; Based on the edge coordinate sequence, the central axis of the target region is calculated through fitting and uniformly sampled. Simultaneously, the curvature of the target region is input into the spatial morphological region model to obtain an adjustment value for the number of sampling points in that region, thereby improving the coverage of the avoidance region; the preset avoidance region is then obtained. The physical coordinates of the corresponding edge pixels in the avoidance area are combined with the coordinates of the sampling points in the target area to obtain the Euclidean distance between the physical coordinates and the sampling point coordinates. The obtained Euclidean distance and the echo intensity corresponding to the preset avoidance area are input into the spatial morphology region model and, after inference correction, the shortest distance between the puncture site and the preset avoidance area is obtained. The shortest distances of all preset avoidance areas are matrixed to obtain a risk distance matrix used to quantify the spatial positional relationship between the target puncture area and each avoidance area. The target area sampling points are used as rows and each preset avoidance area is used as columns. The corresponding shortest distances are used as matrix elements to fill the matrix, forming a risk distance matrix that quantifies the spatial positional relationship between the two.

[0032] The adaptive positioning calibration of the puncture path involves the following steps: First, using the target region and its spatial morphological parameters as the center, obtain a total annular region encompassing the target region and neighboring target regions. The neighboring target regions represent areas planned according to a preset spatial range, centered on the target region. Second, based on the neighborhood variance deviation within the total annular region, obtain an adaptive filter intensity adjustment value. This dynamically optimizes the filtering effect to compensate for speckle noise generated by ultrasound artifacts. Third, based on the adjusted total annular region and the operator scale gradient, generate a corresponding gradient map to filter out pixels with statistical gradient values ​​greater than a preset statistical gradient value, thus obtaining the texture features of the neighboring target regions. Fourth, based on the constructed risk distance matrix, quantize and convert the obtained texture features to obtain a feature-parameter adjustment mapping relationship. This mapping is then performed to obtain the puncture parameter adjustment value, which is then fed back. The puncture parameters include the puncture angle and puncture pressure.

[0033] In this embodiment, the spatial morphology region model is obtained by using a pre-established dataset of ultrasound images, avoidance areas, and spatial morphology parameters corresponding to different puncture areas. A multi-class semantic segmentation architecture is selected, and the output layer is modified to match the number of categories of the target and risk structures. After incorporating the consistency constraints of the target region morphology parameters and morphology, it can achieve multi-class output (such as taking the region curvature as input and outputting the adjusted number of sampling points; taking the current echo intensity as input and outputting the corrected shortest distance). The neighborhood variance deviation represents the difference between the neighboring target region and the neighborhood variance corresponding to the target region. The adaptive filter intensity adjustment value increases the filter intensity by the neighborhood variance deviation ratio. For example, for every 10% deviation exceeding the reference variance, the sigma value increases by 0.1.

[0034] In the shortest distance calculation and risk distance matrix construction stages, misjudgment of risk areas caused by overlapping contours of similar structures is avoided, laying a precise foundation for subsequent distance calculations by dividing the region. The central axis is fitted based on the target region edge coordinate sequence and the number of sampling points is adjusted by combining curvature, so that the distribution of sampling points is more in line with the actual shape of the target region, avoiding distance calculation deviations caused by uneven sampling. Furthermore, by correcting the Euclidean distance and echo intensity, the interference of image noise and artifacts on distance calculation is further reduced, ensuring that the shortest distance between the puncture site and the avoidance area is true and reliable. The resulting risk distance matrix can comprehensively and quantitatively present the spatial relationship between the target region and the risk structure, providing a comprehensive safety basis for puncture path planning.

[0035] A ring-shaped overall region is constructed centered on the target area and its spatial morphological parameters to ensure that the extracted texture features always surround the key areas around the target area, avoiding interference from irrelevant textures in the analysis. Adaptive filtering intensity is dynamically adjusted based on neighborhood variance bias to specifically compensate for speckle noise caused by ultrasound artifacts, making the texture features more closely resemble the true characteristics of the tissue. Gradient maps generated by operator scale gradients are then used to filter effective pixels, further improving the representativeness and discriminative power of the texture features and avoiding feature distortion caused by noise contamination. Combining texture features with a risk distance matrix for quantification mapping establishes a precise correlation between tissue texture characteristics and puncture parameters, allowing puncture parameter adjustments to dynamically adapt to the actual conditions of the adjacent target area, rather than relying on fixed standards.

[0036] like Figure 4 The diagram shows a flowchart of the entry point screening process provided in this embodiment of the invention. The process involves determining the effectiveness of the imaging based on the obtained imaging validity index. If the index is not greater than a preset value, the number of imaging attempts is adjusted to ensure imaging effectiveness. Then, candidate screening is performed based on the minimum risk distance in the risk distance matrix. The relative displacement is obtained by using the standard displacement difference with a preset standard. The process determines whether the obtained displacement is greater than a preset value. If it is, the point is removed; otherwise, the entry point is visualized in the ultrasound image.

[0037] Furthermore, potential entry points are determined based on the prediction process data of the puncture needle spatial trajectory. The specific process is as follows: Based on the established puncture coordinate system and probe parameters, an imaging effectiveness index is obtained to quantify the degree of adaptation between the spatial trajectory and the projection in ultrasound imaging; according to the risk distance matrix, the depth and directional offset of the target area are obtained. The depth of the target area reflects the vertical distance from the body surface to the target area and the thickness of the tissue layer distribution, while the directional offset reflects the spatial deviation angle and horizontal displacement of the puncture area relative to the probe reference axis; if the obtained imaging effectiveness index is greater than the preset imaging effectiveness index, the current target area depth and directional offset are used as potential entry points, and candidate screening is performed; if the obtained imaging effectiveness index is not greater than the preset imaging effectiveness index, the number of ultrasound imaging times is adjusted based on the imaging effectiveness index deviation to reduce the degree of single-shot distortion caused by trajectory projection; and the imaging effectiveness index is re-obtained for judgment. If it is still not greater than the preset imaging effectiveness index, the ultrasound imaging range is adjusted to improve the probe depth adaptation, and potential entry points are obtained simultaneously, followed by candidate screening.

[0038] Specifically, the candidate selection process is as follows: Based on the minimum value of the risk distance in the risk distance matrix, entry points in the ultrasound image whose shortest distance through the preset avoidance area is greater than the preset shortest risk distance are retained; at the same time, the texture features of the neighboring target area corresponding to the entry point are extracted for dynamic stability verification: the coordinate changes of the tracked entry point are obtained in a preset number of ultrasound images to obtain the displacement standard deviation used to quantify the degree of positional fluctuation of the entry point in the ultrasound image; for entry points whose displacement standard deviation exceeds the preset range, their relative displacement in the ultrasound image is further obtained. If the relative displacement is greater than the preset relative displacement, the point is removed; if the relative displacement is not greater than the preset relative displacement, it is visualized on the ultrasound image.

[0039] In this embodiment, the imaging effectiveness index deviation is represented by the difference between the acquired imaging effectiveness index and the preset imaging effectiveness index. The preset imaging effectiveness index is represented by the summation and average of historical imaging effectiveness indices for determining potential entry points. The adjustment of the number of ultrasound imaging sessions is based on the currently acquired deviation and equipment performance. When the deviation is small (e.g., the ratio corresponding to the deviation is less than the preset deviation ratio, for example, currently set to 40% based on equipment performance, but can be adjusted according to the current equipment performance in actual application), a preset range (1-2 times) is used as the adjustment value. When the deviation is large (the ratio corresponding to the deviation is not less than the preset deviation ratio), the deviation ratio is used as the corresponding adjustment range for the number of imaging sessions. The preset shortest risk distance is preset based on the risk distance matrix and the current puncture trajectory. Different puncture trajectories have different risk distances. The preset range of the displacement standard deviation is also preset based on the different entry points. Furthermore, the corresponding entry points during the screening process are fed back to the preset personnel for further judgment.

[0040] The specific expression for the imaging effectiveness index E is: ; In the formula, L represents the projected length of the predicted puncture trajectory on the ultrasound imaging plane, W is the effective imaging width of the probe, θ represents the angle between the predicted puncture trajectory and the central axis of the probe, and its value ranges from 0 ≤ θ ≤ π / 2. Z1 represents the depth of the puncture target area, Z2 represents the preset imaging depth (device inherent parameter), and Z3 represents the maximum imaging depth (device inherent parameter). The predicted ratio reflects the coverage ratio of the trajectory projection within the transverse imaging range of the probe. Based on the geometric projection principle, the ratio of the trajectory projection length to the effective imaging width of the probe quantifies the coverage ratio of the trajectory within the transverse imaging range of the probe, reflecting the effectiveness of the visible length of the trajectory in the ultrasound image. At the same time, the influence on imaging is quantified using the characteristics of the cosine function. The monotonicity of the cosine function is used to transform the angle between the trajectory and the ultrasound beam into a quantitative index. The fractional part measures the deviation between the target depth and the preset probe imaging. The closer E is to 1, the higher the surface imaging effectiveness.

[0041] In the potential entry point identification stage, the degree of adaptation of the puncture trajectory projection in ultrasound imaging is quantified to ensure that the trajectory projection corresponding to the potential entry point can be clearly presented, avoiding positioning deviations caused by poor trajectory-image adaptation. Simultaneously, by combining the target area depth and direction offset obtained from the risk distance matrix, the entry point location is set to fully consider the risk structure avoidance requirements, ensuring the safety of subsequent puncture paths from a spatial perspective. By adjusting the number of ultrasound imaging sessions to reduce trajectory projection distortion or adjusting the imaging range to improve probe depth adaptation, the imaging effect of the entry point can be further optimized, ensuring that the potential entry point not only meets the spatial location requirements of the target area but also has good ultrasound visibility, providing a high-quality candidate basis for subsequent screening.

[0042] In the candidate screening stage, it is ensured that the puncture trajectory corresponding to the retained entry point maintains a safe distance from the preset avoidance area; the coordinate changes of the entry point are tracked and the displacement standard deviation is calculated to quantify the dynamic stability of the point in continuous ultrasound frames, avoiding puncture errors caused by fluctuations in the point due to probe jitter, etc.; for points with displacement standard deviation exceeding the standard, a second verification is performed through relative displacement to carefully distinguish whether the fluctuation is within the controllable range. This avoids blindly eliminating controllable points with slight fluctuations, and also avoids retaining unreliable points with excessive fluctuations. The final selected entry point has both a safe distance guarantee and dynamic stability, providing accurate and reliable positioning reference for puncture operation and reducing the probability of puncture error.

[0043] like Figure 5The diagram shows a schematic of the intelligent puncture entry point identification system based on ultrasound images provided in this embodiment of the invention. The system includes the following modules: an image edge recognition and verification module, a spatial position quantization processing module, and an entry point screening module. The image edge recognition and verification module performs image edge recognition and verification on the real-time dynamic image stream of the ultrasound image. The spatial position quantization processing module performs spatial position quantization processing based on the acquired spatial morphology parameters, simultaneously determines whether it is a risk distance, generates a risk distance matrix, and performs adaptive positioning calibration of the puncture path. The entry point screening module identifies probe edge markers in the ultrasound image, determines potential entry points based on the prediction process data of the puncture needle's spatial trajectory, performs candidate screening, and displays the selected entry points on the ultrasound image.

[0044] In this embodiment, the image edge recognition and verification module is the fundamental support module of the entire system. After performing edge recognition and verification on the real-time dynamic image stream of the ultrasound image, it outputs the spatial morphological parameters corresponding to the target puncture area. These parameters serve as core data and directly provide input to the spatial position quantization processing module. Upon receiving the spatial morphological parameters, the spatial position quantization processing module performs spatial position quantization processing, determines the risk distance and generates a risk distance matrix, and simultaneously completes the adaptive positioning calibration of the puncture path. These processing results are then transmitted to the entry point screening module, providing it with crucial safety and path reference information. Based on the identification of probe edge markers in the ultrasound image, the entry point screening module combines the puncture needle spatial trajectory prediction data and the risk distance matrix output by the spatial position quantization processing module to determine potential entry points and complete candidate screening. Finally, the screened entry points are displayed on the ultrasound image.

[0045] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0046] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0047] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0048] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0049] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0050] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0051] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0052] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0054] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent identification of puncture site based on ultrasound images, characterized in that, Includes the following steps: Step 1: For the real-time dynamic image stream of ultrasound images, perform image edge recognition and verification to obtain the spatial morphological parameters corresponding to the target puncture area. The spatial morphological parameters are used to quantify the relative positional relationship and dynamic deformation law of the three-dimensional geometric features of the puncture site in the target puncture area. Step 2: Based on the acquired spatial morphological parameters, perform spatial location quantification to obtain the shortest distance between the puncture sites in the target area, simultaneously determine whether it is a risk distance, generate a risk distance matrix, and perform adaptive positioning calibration of the puncture path. Step 3: Identify probe edge markers in the ultrasound image, determine potential entry points based on the predicted process data of the puncture needle spatial trajectory, and simultaneously perform candidate screening and display the selected entry sites on the ultrasound image.

2. The intelligent identification method for puncture site based on ultrasound images as described in claim 1, characterized in that, The specific process for image edge recognition and verification is as follows: Obtain the sequence frames corresponding to the real-time dynamic image stream and calculate the structural similarity between adjacent sequence frames; If the structural similarity is greater than the preset structural similarity, keyframes are determined; otherwise, they are judged as frames with violent motion, and adjacent sequence frames are reselected. The key frame is determined by statistically analyzing the proportion of regions where the pixel displacement is greater than a preset pixel value, and selecting adjacent sequence frames whose region proportion is greater than a preset region proportion as stable frames. Based on the acquired stable frames, retain stable frames with a preset number of consecutive frames greater than a preset region percentage and use them as key frames. If there are no consecutive stable frames that meet the standard, the inter-frame pixel fusion ratio is adjusted according to the deviation between the current number of stable frames and the preset number of consecutive frames. The motion deviation of a single frame is compensated by correcting the mean value of the corresponding pixels, and the effectiveness of the deviation calibration is verified at the same time.

3. The intelligent identification method for puncture site based on ultrasound images as described in claim 2, characterized in that, The specific process for verifying the effectiveness of the deviation calibration is as follows: Based on the ultrasound image data after motion deviation correction, the degree of change of a single pixel along the gradient direction in the ultrasound image data is quantified by using preset operator scale parameters to obtain the pixel edge intensity. If the obtained pixel edge intensity is less than the preset pixel edge intensity, the operator scale parameter of the edge detection operator is adjusted to obtain edge feature data at different scales. The edge direction data is determined based on the preset gradient direction, and the edge pixel groups are divided by combining the corresponding one edge pixel above and below in the same direction. In the edge pixel group, the edge pixel with the strongest edge intensity is retained, while the remaining edge pixels in the group are set to 0 to improve edge positioning accuracy and eliminate edge thickening. Dynamic adjustment of connected components is performed based on edge data obtained after verification of the effectiveness of deviation calibration.

4. The intelligent identification method for puncture site based on ultrasound images as described in claim 3, characterized in that, The dynamic adjustment of the connected components is specifically carried out as follows: The connected component parameter data contained in the edge data are compared with the corresponding reference connected component parameter data. The connected component parameter data includes connected component length data and connected component curvature data, and the reference connected component parameter data includes reference connected component length data and reference connected component curvature data. If the obtained connected component length data is greater than the reference connected component length data, then an increase in the connected component length data threshold is obtained based on the connected component length data deviation to determine whether the connected component data belongs to the edge data of the target region. If the obtained connected component length data is less than the reference connected component length data, then a decrease in the connected component length data threshold is obtained to filter out the valid connected component data related to the edge data of the target region. If the obtained connected component curvature data is greater than the reference connected component curvature data, the threshold value of the connected component curvature data is increased based on the deviation of the connected component curvature data. If the obtained connected component curvature data is less than the reference connected component curvature data, the threshold value of the connected component curvature data is decreased, so as to accurately filter the shape trend of the connected component data.

5. The intelligent identification method for puncture site based on ultrasound images as described in claim 4, characterized in that, The dynamic adjustment of the connected components also includes: If the obtained connected component length data is equal to the reference connected component length data, or the obtained connected component curvature data is equal to the reference connected component curvature data, then the connected component parameter threshold data is maintained. Based on the adjusted connected component parameter threshold data, effective connected component data in ultrasound images are obtained by filtering, and spatial morphological parameters in the effective connected component data are acquired. The effective connected component data represents the data corresponding to connected components whose shape and location are related to the target region. The process of obtaining the spatial morphological parameters in the effective connected components is as follows: Extract the edge pixel coordinates within the effective connected components, and form a continuous edge coordinate sequence by taking the gradient direction of each edge pixel as the starting direction data and concatenating them. During the formation of the edge coordinate sequence, the path length corresponding to the edge coordinate sequence is obtained and updated to the target region length data; The edge coordinate sequence and the length of the target region are used together as spatial morphological parameters to characterize the spatial morphology of the target region.

6. The intelligent identification method for puncture site based on ultrasound images as described in claim 5, characterized in that, The specific steps for obtaining the shortest distance between the puncture sites in the target area are as follows: Using the obtained spatial morphological parameters as input, and based on the spatial morphological region model, the output is a segmentation mask between the target puncture area and the preset avoidance area; Based on the edge coordinate sequence, the central axis of the target area is obtained by fitting and uniform sampling. At the same time, the curvature of the target area is input into the spatial morphology area model to obtain the sampling point adjustment value of the area, so as to improve the coverage of the avoidance area. Obtain the physical coordinates of the corresponding edge pixels in the preset avoidance area, and combine them with the sampling point coordinates of the target area to obtain the Euclidean distance between the physical coordinates and the sampling point coordinates; The obtained Euclidean distance and the echo intensity corresponding to the preset avoidance area are input into the spatial morphology region model. After inference correction, the shortest distance between the puncture site and the preset avoidance area is obtained. The shortest distances of all preset avoidance zones are matrixed to obtain a risk distance matrix used to quantify the spatial relationship between the target puncture area and each avoidance zone.

7. The intelligent identification method for puncture site based on ultrasound images as described in claim 6, characterized in that, The specific process for the adaptation and positioning calibration of the puncture path is as follows: Using the target area and its spatial morphological parameters as the center, obtain a total annular area that includes the target area and neighboring target areas. The neighboring target areas refer to the areas planned according to a preset spatial range with the target area as the center. Based on the neighborhood variance deviation within the total annular region, an adaptive filter intensity adjustment value is obtained, and the speckle noise generated by ultrasound artifacts is compensated by dynamically optimizing the filter effect. Based on the adjusted annular total area and combined with the operator scale gradient, a corresponding gradient map is generated to filter out pixels with statistical gradient values ​​greater than the preset statistical gradient value, thereby obtaining the texture features of the neighboring target area. Based on the constructed risk distance matrix, the acquired texture features are quantized and transformed to obtain the feature-parameter adjustment mapping relationship. The mapping is then performed to obtain the puncture parameter adjustment value and feedback is provided. The puncture parameters include the puncture angle and puncture pressure.

8. The intelligent identification method for puncture site based on ultrasound images as described in claim 1, characterized in that, The process of determining potential entry points based on the prediction data of the puncture needle spatial trajectory is as follows: Based on the established puncture coordinate system and probe parameters, an imaging effectiveness index is obtained to quantify the degree of fit between the spatial trajectory and the projection in ultrasound imaging. Based on the risk distance matrix, the depth and directional offset of the target area are obtained. The depth of the target area is used to reflect the vertical distance from the body surface to the target area and the thickness of the tissue layer distribution. The directional offset is used to reflect the spatial deviation angle and horizontal displacement of the puncture site relative to the probe reference axis. If the acquired imaging effectiveness index is greater than the preset imaging effectiveness index, then the current target area depth and orientation offset are used as potential entry points, and candidate screening is performed. If the obtained imaging effectiveness index is not greater than the preset imaging effectiveness index, the number of ultrasound imaging times is adjusted based on the deviation of the imaging effectiveness index to reduce the degree of distortion per imaging caused by trajectory projection. The imaging effectiveness index is re-acquired for evaluation. If it is still not greater than the preset imaging effectiveness index, the ultrasound imaging range is adjusted to improve the probe depth adaptation. At the same time, potential entry points are acquired, and then candidate screening is carried out.

9. The intelligent identification method for puncture site based on ultrasound images as described in claim 8, characterized in that, The specific process of candidate screening is as follows: Based on the minimum value of the risk distance in the risk distance matrix, the entry point in the ultrasound image is retained where the shortest distance of the puncture trajectory through the preset avoidance area is greater than the preset shortest risk distance; Simultaneously, extract the texture features of the neighboring target region corresponding to the entry point and perform dynamic stability verification: obtain the coordinate changes of the tracking entry point in the ultrasound images of a preset number of frames, and obtain the displacement standard deviation used to quantify the degree of positional fluctuation of the entry point in the ultrasound images. For entry points whose displacement standard deviation exceeds the preset range, their relative displacement in the ultrasound image is further obtained. If the relative displacement is greater than the preset relative displacement, the point is removed. If the relative displacement is not greater than the preset relative displacement, it is visualized on the ultrasound image.

10. A puncture site intelligent identification system based on ultrasound images, using the puncture site intelligent identification method based on ultrasound images as described in any one of claims 1-9, comprising the following modules: an image edge recognition and verification module, a spatial location quantization processing module, and an entry site screening module; in, The image edge recognition and verification module is used to perform image edge recognition and verification on the real-time dynamic image stream of ultrasound images. The spatial location quantization processing module is used to perform spatial location quantization processing based on the acquired spatial morphology parameters, simultaneously determine whether it is a risk distance, generate a risk distance matrix, and perform adaptive positioning calibration of the puncture path. The entry site screening module is used to identify probe edge markers in ultrasound images, determine potential entry points based on the prediction process data of the puncture needle spatial trajectory, and simultaneously perform candidate screening and display the selected entry sites on the ultrasound images.

Images