Puncture auxiliary system for central venous catheterization through peripheral vein puncture

Abstract

The invention relates to the technical field of venous indwelling catheter puncture navigation, and discloses a central venous indwelling catheter puncture auxiliary system through peripheral vein puncture. According to the system, an ultrasonic image and a dissection description text are acquired through a data acquisition and verification module; a three-dimensional blood vessel modeling module is used for carrying out spatial registration on the two blood vessels, and a three-dimensional blood vessel model containing semantic information is constructed; the anatomical variation analysis module is compared with a standard map, a variation region is automatically marked, and the puncture feasibility is evaluated; the puncture path planning module generates a target path in the feasible blood vessel model in combination with the instrument parameters, and decomposes the target path into executable steps associated with real-time ultrasonic features and specific operation instructions; and finally, the path conflict detection and navigation control module carries out calibration and sends an instruction to control the puncture auxiliary device to align. According to the method, a closed loop from objective evaluation of multi-source information fusion to programmed step navigation is realized, and the dependence on personal experience and spatial imagination of an operator is reduced.

Description

Technical Field

[0001] This invention relates to the field of venous catheterization navigation technology, specifically to a peripheral venous puncture central venous catheterization auxiliary system. Background Technology

[0002] Peripherally inserted central venous catheterization is a commonly used medium- to long-term venous therapy technique in clinical practice. Currently, this procedure mainly relies on real-time ultrasound imaging for guidance. The operator needs to identify the target vessel on the ultrasound image and, based on personal experience, determine the vessel's spatial location, course, and anatomical relationship with surrounding nerves and tendons, thereby planning the puncture path and manually performing the puncture. Existing techniques heavily depend on the operator's subjective experience and spatial imagination, resulting in incomplete assessment and inaccurate planning.

[0003] Current technologies have shortcomings. During the preoperative assessment phase, clinical information is fragmented. Ultrasound images only provide cross-sectional information, while crucial information such as abnormal vascular course and anatomical variations is often recorded in written medical records. Current technologies lack effective means to integrate and unify the interpretation of images and textual descriptions, resulting in preoperative assessments relying heavily on two-dimensional images and experience, insufficient identification of complex anatomical variations, and a high risk of inappropriate vessel selection or puncture failure. During the path execution phase, current ultrasound guidance primarily provides static needle insertion points and angle references, or only displays a virtual path. This path lacks a precise and executable mapping relationship with specific instrument manipulation actions and real-time ultrasound image characteristics, leading to a disconnect between planning and execution, and limited navigation accuracy.

[0004] This invention aims to address the insufficient utilization of multimodal information fusion, realizing the transformation from separated image and text information into a three-dimensional anatomical model that can be used for automatic analysis. This invention also aims to address the disconnect between virtual paths and actual operations, realizing the transformation from geometric path planning to sequential operation instructions that can be executed step-by-step by the driving device. Summary of the Invention

[0005] The purpose of this invention is to provide a peripherally inserted central venous catheterization puncture assistance system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a peripherally inserted central venous catheterization puncture assistance system, the system comprising: The data acquisition and verification module is configured to acquire and verify the patient's identity, ultrasound images of the target blood vessel, and anatomical description text. The three-dimensional vascular modeling module is configured to spatially register the ultrasound image with the anatomical description text to generate a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues. The anatomical variation analysis and feasibility assessment module is configured to compare the three-dimensional vascular structure model with a standard vascular atlas library to mark anatomical variation areas, and to assess the puncture feasibility of the target blood vessel based on the anatomical variation areas and a preset contraindication rule library. The puncture path planning module is configured to define a candidate space in the three-dimensional vascular structure model based on the assessed feasible target blood vessel, combined with the physical size parameters of the puncture needle and the specification parameters of the catheter, calculate the initial puncture point set that meets the requirements of avoiding nerves and tendons, and simulate the catheter travel process to generate a target path from the skin to the central vein, and then decompose it into an executable path plan that associates ultrasound image recognition features with puncture instrument operation instructions. The path conflict detection and navigation control module is configured to perform conflict detection and calibration on the executable path plan, and send the calibrated path plan to the navigation execution terminal to control the alignment of the puncture auxiliary device.

[0007] Preferably, the data acquisition and verification module performs verification in the following manner: Receive patient identification codes from the medical information system and retrieve historical medical records and image files associated with the patient identification codes from local storage; Receive a real-time video stream from an ultrasound device and extract one or more key images from the real-time video stream; Receive text annotations about the location of the target blood vessel input by the operator; The historical medical records, image archives, key images, and text annotations are timestamped and their content is verified to be consistent, generating a unified pre-puncture preparation information package.

[0008] Preferably, the three-dimensional blood vessel modeling module generates the three-dimensional blood vessel structure model through the following steps: Apply filtering to key images to reduce speckle noise; Identify the edges of the vascular cavity region in the filtered image and delineate the contour of the vascular wall based on the gray-level gradient changes; The names, relative depths, and keywords of adjacent tissues of blood vessels were extracted from the text annotation information. By combining the blood vessel wall contour with data parsed from text annotation information, preliminary two-dimensional layer data is constructed using contour depth information. Multiple two-dimensional data are stacked and interpolated along the scanning direction of the ultrasound probe to form a three-dimensional vascular structure model with depth dimension, including vascular depth, diameter, and positional relationship with surrounding tissues.

[0009] Preferably, the anatomical variation analysis and feasibility assessment module performs comparison, labeling, and evaluation through the following steps: Call the standard 3D model of the target vessel from the standard vascular atlas library; The three-dimensional vascular structure model, which includes vascular depth, diameter, and positional relationship with surrounding tissues, is spatially aligned and its differences are measured with the standard three-dimensional model. Calculate the geometric deviation between the three-dimensional vascular structure model, which includes vascular depth, diameter, and positional relationship with surrounding tissues, and the standard three-dimensional model, and mark the areas with deviation values ​​exceeding a preset threshold as anatomical variation areas; The taboo rule base is queried, which defines the risk levels and operational taboo clauses corresponding to different types of anatomical variations; Determine the risk level corresponding to the marked anatomical variation area. If the risk level exceeds the allowable range, output the assessment status as infeasible; otherwise, output it as feasible.

[0010] Preferably, the puncture path planning module defines the candidate space in the following manner: Read the diameter and bevel length parameters of the puncture needle, and read the length and outer diameter parameters of the catheter; On the surface of a three-dimensional vascular structure model that is deemed feasible in the evaluation state and includes the vessel depth, diameter and positional relationship with surrounding tissues, a virtual puncture and needle insertion space is offset outward along the normal direction of the vessel wall, with the length of the puncture needle as a reference. Inside a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues, a virtual catheter travel channel is formed by contracting inward along the vascular centerline with reference to the outer diameter of the catheter. The candidate space is obtained by performing a Boolean operation on the intersection of the virtual puncture needle insertion space and the virtual catheter travel channel.

[0011] Preferably, the puncture path planning module calculates the initial puncture point set through the following steps: Obtain the physical dimensions of the ultrasound probe and the angle parameters of the scanning plane; On the outer surface of the candidate space, the placement and orientation of the simulated ultrasound probe are arranged so that the scanning plane can completely cover a section of the candidate space. Calculate the theoretical resolution and sharpness of the ultrasound image under the scanning plane to generate image quality simulation parameters; Within the scanning area that meets the image quality requirements, all surface points whose distance to the nerve and tendon structures marked in the distance model is greater than the safe distance are further screened out. The set of these surface points constitutes the initial puncture point set.

[0012] Preferably, the puncture path planning module generates a target path from the skin to the central vein, including: In the initial set of puncture points, select the point closest to the skin surface as the planning starting point; In a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues, a path search is performed from the planning starting point along the vascular centerline toward the heart chamber. During the path search process, the catheter flexibility parameters are read in real time, which limit the allowable curvature of the path. When the search path encounters an area of ​​anatomical variation, the detour path is calculated based on the catheter's flexibility parameters to ensure that the curvature of the detour path is within the catheter's allowable range. The search path stops when it reaches the preset endpoint of the model's central vein, and the recorded continuous search trajectory is the target path.

[0013] Preferably, the puncture path planning module forms the executable path plan through the following steps: Sampling is performed along the target path at fixed intervals to obtain a series of ordered spatial location coordinates; For each spatial location coordinate, the key image generated in step three is matched, and the typical texture features of the blood vessels corresponding to the spatial location coordinates in the ultrasound image are extracted as ultrasound image recognition features. For the path segment between two adjacent spatial sites, the basic operations required for the puncture needle or catheter are defined. These basic operations include needle insertion angle, rotation direction and advance distance. These basic operations are encoded as puncture instrument operation instructions. By sequentially binding spatial location coordinates, ultrasound image recognition features, and puncture instrument operation instructions, an executable path plan is formed.

[0014] Preferably, the path conflict detection and navigation control module performs conflict detection and calibration through the following steps: Retrieve the spatial coordinates of all regions marked as absolutely forbidden from the taboo rule library; The coordinates of each spatial point in the executable path plan are compared with the spatial coordinates of the absolutely forbidden area to determine whether spatial location intrusion has occurred. If a spatial intrusion occurs, a local alternative path that avoids the absolutely forbidden area will be replanned on the path segment near the intrusion point, based on the path search method in step seven. Replace the corresponding part of the original path plan with a local alternative path to generate a calibrated path plan.

[0015] Preferably, the path conflict detection and navigation control module sends the path plan to the navigation execution terminal through the following steps: Convert all data in the calibrated path plan into a data stream format that conforms to the communication protocol of the navigation execution end; The data stream format includes spatial coordinate instructions, image feature comparison instructions, and instrument action instructions arranged in sequence. The route plan in data stream format is transmitted to the navigation execution end in real time via the data interface; The navigation execution end parses the data stream, drives its robotic arm to adjust the position of the ultrasound probe to match the image feature comparison instructions, and guides the puncture needle to move along the spatial coordinate instructions according to the instrument action instructions.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By spatially registering ultrasound images with anatomical descriptions, a three-dimensional vascular structure model incorporating semantic information is constructed. This model transcends mere geometric representation, integrating key descriptions of vascular characteristics and surrounding tissues from clinical texts. Based on this, the system automatically compares this individualized model with standard vascular atlases, objectively and quantitatively identifying areas of anatomical variation. This method overcomes the limitations of relying solely on two-dimensional ultrasound images and physician experience, shifting preoperative assessment from subjective judgment to objective analysis based on multi-source data fusion. This improves the accuracy of identifying complex or variant vessels, providing a more reliable anatomical basis for puncture feasibility.

[0017] After planning the path in a 3D model, the geometric path is decomposed into an executable plan that associates ultrasound image recognition features with puncture instrument operation instructions. The system pre-defines specific ultrasound image features that should appear at key points along the path and translates the actions to reach those points into specific instrument control commands. This process establishes a precise correspondence between "path points - image features - operation actions." The abstract spatial path is translated into a series of step-by-step operation commands that rely on real-time image feedback for verification. The accuracy of puncture navigation no longer depends solely on the doctor's understanding of the virtual path and hand-eye coordination; instead, the system breaks down the path into closed-loop control commands that can be executed step-by-step by the auxiliary device, reducing the operational difficulty and uncertainty under complex paths and achieving a leap from path display to programmed guidance. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the working principle of the peripherally inserted central venous catheterization auxiliary system described in this invention. Figure 2 A flowchart illustrating the operation of the data acquisition and verification module; Figure 3 A flowchart illustrating the operation of the 3D blood vessel modeling module; Figure 4 This is a diagram illustrating the anatomical variations of a three-dimensional vascular model. Figure 5 For puncture path planning and navigation map. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides a peripherally inserted central venous catheterization assistance system. The system includes: an integrated data acquisition and verification module, a three-dimensional vascular modeling module, an anatomical variation analysis and feasibility assessment module, a puncture path planning module, and a path conflict detection and navigation control module. The data acquisition and verification module acquires and verifies the patient's identification, ultrasound images of the target vessel, and anatomical description text. The three-dimensional vascular modeling module spatially registers the ultrasound images and anatomical description text to generate a three-dimensional vascular structure model including vessel depth, diameter, and positional relationship with surrounding tissues. The anatomical variation analysis and feasibility assessment module compares the three-dimensional vascular structure model with a standard vascular atlas to label solutions. The system analyzes the anatomical variation area and assesses the feasibility of puncturing the target blood vessel based on the anatomical variation area and a pre-defined contraindication rule base. The puncture path planning module, based on the assessed feasible target blood vessel, combines the physical size parameters of the puncture needle and the specification parameters of the catheter to delineate the candidate space in the three-dimensional blood vessel structure model, calculates the initial puncture point set that meets the requirements of avoiding nerves and tendons, simulates the catheter travel process to generate the target path from the skin to the central vein, and decomposes it into an executable path plan that associates ultrasound image recognition features and puncture instrument operation instructions. The path conflict detection and navigation control module performs conflict detection and calibration on the executable path plan, and sends the calibrated path plan to the navigation execution end to control the alignment of the puncture auxiliary device.

[0021] Example 1: See Figure 2In its implementation, the data acquisition and verification module performs the verification function, which is achieved through a series of ordered data receiving, extraction, comparison, and integration operations. The module receives patient identification codes from the medical information system. Each patient identification code is a unique code corresponding to a patient. Based on these codes, the module retrieves all historical medical records and historical image archives associated with the patient identification code from the system's local storage. Historical medical records include the patient's past diagnostic and surgical records, while historical image archives include the patient's past CT or MRI images. Simultaneously, the module receives real-time video streams from the ultrasound equipment. These streams are continuous image sequences generated when the ultrasound probe scans the target vascular region of the patient. One or more key images meeting the requirements are extracted from the real-time video stream based on predefined image clarity thresholds and vascular structure integrity rules. The module also receives text annotations about the target vascular location manually entered by the operator via an input device. These text annotations describe the surface projection location of the vascular vessel and surrounding anatomical landmarks in string format.

[0022] In some embodiments, the data acquisition and verification module performs timestamp alignment on the received multi-source information. The timestamp information originates from the medical information system, the ultrasound equipment clock, and the input device system time. The data acquisition and verification module compares the creation time of historical medical records, the capture time of historical image archives, the generation time of key images extracted from the real-time video stream, and the entry time of text annotation information to check whether all time information is within a reasonable time window of the same medical session.

[0023] In practical implementation, the data acquisition and verification module performs content consistency verification, which involves comparing the consistency of anatomical structures described by different types of data. The module parses text descriptions of blood vessel names and locations from historical medical records and text annotations, identifies the morphology and orientation of blood vessels from historical image archives and key images, and establishes a mapping relationship between text descriptions and image features. The module then calculates a spatial consistency score between the anatomical location indicated by the text description and the actual location of the blood vessel identified in the image. One formula for calculating the spatial consistency score is as follows:

[0024] in: Indicates the spatial consistency score. This represents the location description vector parsed from the text annotation information. This represents a vector describing the location of blood vessels identified from a key image; the function... Calculate the semantic similarity between two description vectors. This represents the theoretical bounding box of blood vessels in the image, inferred from the text description. The function represents the actual segmented blood vessel bounding box from the key image. Calculate the intersection-union ratio (CUP) of the two bounding boxes. and It is a weighting coefficient used to balance semantic similarity and spatial overlap.

[0025] It is understandable that only when the spatial consistency score is... Only when the preset consistency threshold is exceeded can the data acquisition and verification module determine that the multi-source information content is consistent. The data acquisition and verification module packages all verified historical medical records, historical image archives, key images and text annotations, along with their corresponding timestamps and spatial consistency scores, into a structured and unified pre-puncture preparation information package.

[0026] Optionally, when extracting key images from the real-time video stream, the data acquisition and verification module employs a blood vessel contour stability detection method based on inter-frame difference. The data acquisition and verification module calculates the variance of pixel grayscale changes in the blood vessel contour region between consecutive frames in the video stream. When the variance is lower than a set stability threshold, the current frame's blood vessel image is determined to be clear and stable, and the current frame is included in the candidate key image set. Finally, one or more frames are selected from the candidate key image set based on image sharpness scores as the final key images.

[0027] Optionally, when receiving text annotation information, the data acquisition and verification module provides a structured input interface. This interface includes a predefined dropdown list of vessel names and checkboxes for adjacent tissues to standardize operator input and reduce ambiguity that may arise from free text. The data acquisition and verification module combines the content selected by the operator through the input interface with the manually entered content to generate a complete text annotation information string.

[0028] Example 2: See Figure 3In its implementation, the 3D vascular modeling module operates based on the pre-puncture preparation information package generated by the data acquisition and verification module. This module processes key images and text annotations within the package to generate a 3D vascular structure model that includes vessel depth, diameter, and its positional relationship with surrounding tissues. The module applies filtering to the key images to reduce speckle noise. An adaptive median filter is used, dynamically adjusting the filter window size based on the pixel grayscale statistical characteristics within a local window of the key image. This suppresses inherent speckle noise in the ultrasound image while preserving as much edge detail as possible in the vessel wall. The module then identifies the edges of the vascular lumen region in the filtered image. This identification process leverages the characteristic that the vascular lumen appears as an anechoic or hypoechoic dark area in ultrasound images. It calculates the grayscale gradient value of each pixel in the image and uses gradient thresholding to initially determine the approximate region of the vascular lumen. Finally, it delineates the inner and outer boundaries of the vessel wall based on the location of the largest grayscale gradient change.

[0029] In some embodiments, the 3D vascular modeling module parses the vessel name, relative depth, and adjacent tissue keywords from text annotation information. The parsing process utilizes a pre-trained natural language processing model that identifies named entities related to anatomical structures in the text annotation string. For example, "basal vein" and "brachial artery" are identified as vessel name entities, "approximately 1.5 cm deep" is identified as a relative depth entity, and "adjacent to the median nerve" is identified as an adjacent tissue keyword entity. The 3D vascular modeling module combines the vessel wall contour with the data parsed from the text annotation information. This combination process involves spatial location mapping. The relative depth data parsed from the text annotation information provides the absolute depth information of the vessel wall contour within the human tissue, while adjacent tissue keywords are used to annotate the surrounding environmental structures in subsequent modeling.

[0030] In practical implementation, the 3D vascular modeling module utilizes contour depth information to construct preliminary two-dimensional layer data. Each key image and its corresponding text parsing result generate a two-dimensional layer data set. This two-dimensional layer data set is a structure containing the pixel coordinates of the vessel wall contour, the absolute depth value corresponding to that contour, and the neighboring tissue labels parsed from the text. It can be understood that the two-dimensional layer data set is a discrete representation of the 3D model on a single scanning plane. The 3D vascular modeling module stacks multiple two-dimensional layer data sets along the ultrasound probe scanning direction, sorting them according to the absolute depth value carried by each two-dimensional layer data set. Since the images obtained from ultrasound scanning are discrete, the 3D vascular modeling module performs interpolation operations between adjacent two-dimensional layer data sets to generate intermediate-level vessel contours, thereby forming a continuous and complete 3D geometric model with depth dimension. A formula for interpolation between adjacent contours is expressed as follows:

[0031] in: This represents the coordinates of a point on the new contour line generated by interpolation. and These represent the corresponding parameter points on the blood vessel wall contour lines in two adjacent two-dimensional data layers. coordinates, parameters It is a normalized quantity along the length of the contour line, parameter Identify a specific sequence of points on the outline. It is an interpolation weighting factor between 0 and 1, whose value is proportional to the depth difference between the point to be interpolated and layer A and layer B.

[0032] Optionally, when outlining the blood vessel wall, the 3D blood vessel modeling module employs an active contour model algorithm. The algorithm initializes a closed curve near the edge of the dark area of ​​the blood vessel in the image. This curve evolves under the combined influence of internal energy and image edge energy. The internal energy controls the continuity and smoothness of the curve, while the image edge energy attracts the curve to the position with the largest gray-level gradient, i.e., the blood vessel wall boundary. When the curve evolution energy function reaches its minimum, the shape of the curve is the final outline of the blood vessel wall.

[0033] It is understandable that through the above stacking and interpolation process, the continuous tubular structure of blood vessels in three-dimensional space can be reconstructed. The three-dimensional blood vessel modeling module further calculates the diameter of each point on the blood vessel path in the reconstructed three-dimensional geometric model, and records the spatial orientation and distance relationship between the outer wall of the blood vessel and the structures represented by the keywords of the adjacent tissues parsed from the text annotation information. Finally, it outputs a three-dimensional blood vessel structure model that includes the blood vessel depth, diameter, and positional relationship with the surrounding tissues.

[0034] In some embodiments, when there is a conflict between the depth information parsed from the text annotation information and the depth estimated from the image grayscale information, the 3D blood vessel modeling module sets a confidence fusion rule. The rule assigns a higher confidence score to the depth estimated from the image grayscale information, but retains the depth parsed from the text annotation information as a verification reference. The inconsistency is recorded in the log of the 3D blood vessel modeling module, and the modeling is based primarily on the depth estimated from the image, while also annotating any uncertainties that may exist in the text information.

[0035] Example 3: In specific implementation, the anatomical variation analysis and feasibility assessment module receives a 3D vascular structure model generated by the 3D vascular modeling module, which includes vascular depth, diameter, and positional relationship with surrounding tissues. The module then retrieves a standard 3D model of the target vascular vessel from a standard vascular atlas library. This standard vascular atlas library is a pre-built anatomical structure database storing standard geometric shapes and spatial relationship models of various vascular types, derived from statistical analysis of extensive population anatomical data. The retrieval process involves searching the standard vascular atlas library by matching the target vascular vessel's name keywords and extracting the corresponding standard 3D model data. The module then spatially aligns the 3D vascular structure model (containing vascular depth, diameter, and positional relationship with surrounding tissues) with the standard 3D model. This spatial alignment operation uses an iterative nearest-point algorithm to find the optimal rigid body transformation between the two models, ensuring a match in their overall position and orientation in 3D space. After spatial alignment, the module performs a difference metric, which quantifies the distance between corresponding point sets of the two models in the spatially aligned state.

[0036] In some embodiments, the geometric deviation between a three-dimensional vascular structure model, including vessel depth, diameter, and positional relationship with surrounding tissues, and a standard three-dimensional model is calculated. The anatomical variation analysis and feasibility assessment module samples a series of points on the surface of the three-dimensional vascular structure model, and for each sampled point, finds its nearest neighbor on the surface of the standard three-dimensional model, calculating the Euclidean distance between each pair of corresponding points. The deviation value is a statistical summation of all these Euclidean distances, and a formula for calculating the summative deviation value is expressed as:

[0037] in: This represents the root mean square deviation value. This represents the total number of points sampled from the surface of a three-dimensional vascular structure model, which includes the vessel depth, diameter, and positional relationship with surrounding tissues. Indicates the first The three-dimensional coordinates of each sampling point on a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues. Represents the surface of a standard 3D model and The corresponding three-dimensional coordinates of the nearest neighbor, symbol This indicates the calculation of the Euclidean distance between two points. The anatomical variation analysis and feasibility assessment module will calculate the deviation value. Compare with a preset threshold.

[0038] In practice, the anatomical variation analysis and feasibility assessment module marks areas with deviation values ​​exceeding preset thresholds as anatomical variation regions. This marking process not only targets overall deviations, but also divides the 3D vascular structure model (including vessel depth, diameter, and positional relationship with surrounding tissues) into local regions, calculating the local deviation value between each sub-region and its corresponding region in the standard 3D model. When the local deviation value of a sub-region exceeds its corresponding local preset threshold, the module marks that sub-region as an anatomical variation region in the model data and records its spatial coordinate range and deviation type, which includes abnormal vessel diameter, tortuous course, or positional offset.

[0039] Understandably, the anatomical variation analysis and feasibility assessment module queries the contraindication rule base, a dataset storing risk assessment rules. This base defines the risk levels and procedural contraindications corresponding to different types of anatomical variations. For example, the contraindication rule base might include rules such as "If a blood vessel runs within 0.3 cm of a critical nerve bundle, it is marked as high-risk and associated with the procedural contraindication 'puncture prohibited'," or "If a blood vessel has aneurysmal dilatation, it is marked as medium-risk and associated with the procedural contraindication 'assess vessel wall stability'." The anatomical variation analysis and feasibility assessment module then matches the type, location, and size information of the marked anatomical variation region with the rule entries in the contraindication rule base.

[0040] In some embodiments, the risk level corresponding to the marked anatomical variation areas is determined. The anatomical variation analysis and feasibility assessment module assigns a risk level label to each anatomical variation area based on the matched rule entries. Risk level labels are categorized as high risk, medium risk, and low risk. The anatomical variation analysis and feasibility assessment module integrates the risk levels of all anatomical variation areas. If the risk level of any anatomical variation area exceeds the allowable range (which is set to only allow low risk levels and below), the module outputs an assessment status of "infeasible." If the risk levels of all anatomical variation areas are within the allowable range, the module outputs an assessment status of "feasible." The conclusion of "infeasible" or "feasible," along with the information of the marked anatomical variation areas, is passed as output to the subsequent puncture path planning module.

[0041] Optionally, when calculating the deviation value, the anatomical variation analysis and feasibility assessment module calculates not only the deviation of surface geometry but also the difference in curvature variation of the vessel centerline. This module extracts the vessel centerline from a three-dimensional vessel structure model containing vessel depth, diameter, and positional relationship with surrounding tissues. Simultaneously, it extracts the corresponding standard vessel centerline from a standard three-dimensional model. The similarity between the two centerlines is quantified using the Fraser distance method, and this similarity metric is incorporated as a supplementary criterion for geometric deviation calculation. The module applies the Fraser distance method to quantify the similarity between the two centerlines. The Fraser distance calculates the overall morphological difference by simulating the shortest path between points on two curves. Specifically, it traverses all point sequences on the centerline, searching for a matching path that minimizes the maximum distance between corresponding points on the two curves. This maximum distance value is the Fraser distance result, reflecting the similarity of the centerlines in their direction and curvature.

[0042] Optionally, the rules in the taboo rule base support hierarchical and logical combinations. When matching rules, the anatomical variation analysis and feasibility assessment module can handle composite risk levels corresponding to multiple combinations of anatomical variation features. For example, if an anatomical variation region simultaneously satisfies both the conditions of "proximity to a nerve" and "narrow vessel diameter," the anatomical variation analysis and feasibility assessment module will, based on the logic defined in the taboo rule base, determine this combination of conditions to have a higher risk level than a single condition.

[0043] See Figure 4 This diagram demonstrates the anatomical variation analysis process of a 3D vascular model. The image clearly compares the patient's actual vascular structure with a standard vascular model using blue and green point clouds, making the spatial morphological differences between the two models immediately apparent. Red, orange, and yellow clusters are specifically marked as sites of anatomical variation. These areas are automatically identified by the system after calculating geometric deviations and comparing them with preset thresholds, representing anatomical variations in diameter, course, or location. The diagram also includes simulated distributions of surrounding tissues and a comparison of the vascular centerline, presenting a comprehensive and three-dimensional view of the complex spatial relationship between the blood vessel and its surrounding environment. This intuitively demonstrates how the system uses quantitative analysis to assess the feasibility and risks of puncture.

[0044] Example 4: In specific implementation, the puncture path planning module receives a three-dimensional vascular structure model, including vessel depth, diameter, and positional relationship with surrounding tissues, from the anatomical variation analysis and feasibility assessment module, which is deemed feasible. The module reads the diameter and bevel length parameters of the puncture needle, and the length and outer diameter parameters of the catheter. The puncture needle and catheter parameters are obtained from a predefined instrument specification database, which stores the physical dimensions of different models of puncture needles and catheters. On the surface of the three-dimensional vascular structure model (which is deemed feasible) and includes vessel depth, diameter, and positional relationship with surrounding tissues, the puncture path planning module offsets outward along the normal direction of the vessel wall's outer surface, using the puncture needle length as a reference, to create a virtual puncture insertion space. This virtual space is a three-dimensional region extending along the normal direction from the vessel wall's outer surface as the starting point, with a distance equal to the puncture needle length. Inside the three-dimensional vascular structure model, the puncture path planning module contracts inward along the vessel's centerline, using the catheter's outer diameter as a reference, to form a virtual catheter travel channel. The virtual catheter travel channel is a tubular space region with the vessel centerline as its central axis and a cross-sectional radius slightly larger than the catheter's outer diameter. Inward contraction ensures a tiny gap between the catheter and the vessel wall as it travels within the vessel. The puncture path planning module performs a Boolean intersection operation on the virtual puncture needle insertion space and the virtual catheter travel channel. This intersection operation yields a continuous three-dimensional space that simultaneously satisfies the requirements of both the puncture needle insertion space and the catheter travel channel; this continuous three-dimensional space is the candidate space.

[0045] In some embodiments, the puncture path planning module acquires the physical dimensions and scanning plane angle parameters of the ultrasound probe. The physical dimensions of the ultrasound probe include the length and width of the probe's bottom surface, and the scanning plane angle parameter refers to the angle between the two-dimensional sectional plane formed by the emitted ultrasound beam and the normal to the probe's bottom surface. The puncture path planning module simulates the placement position and orientation of the ultrasound probe on the outer surface of the candidate space. The simulation process calculates the contact point and contact angle between the probe's bottom surface and the outer surface of the candidate space, ensuring that the scanning plane of the ultrasound probe can completely cover a continuous area of ​​the candidate space. The puncture path planning module calculates the theoretical resolution and sharpness of the ultrasound image under the scanning plane. The calculation is based on the ultrasound frequency, probe aperture, and simulated tissue acoustic characteristics between the candidate space and the probe, generating image quality simulation parameters, including spatial resolution and signal-to-noise ratio (SNR) values. A formula for calculating the SNR of an ultrasound image is expressed as follows:

[0046] in: Signal-to-noise ratio, expressed in decibels (dB). This represents the power of the ultrasonic echo signal received from the target area. This represents the combined noise power of system noise and tissue noise. Within the scanning area that meets image quality requirements, the puncture path planning module further filters out all surface points whose distance to nerve and tendon structures, as indicated in the distance model, is greater than the safe distance. These nerve and tendon structures are labeled in a three-dimensional vascular structure model that includes vessel depth, diameter, and positional relationship with surrounding tissues. The safe distance is a preset minimum interval value to avoid damage. The set of these surface points that meet all conditions constitutes the initial puncture point set, as shown in Table 1.

[0047] Table 1: Specifications of Puncture Needles and Catheters Device type Parameter name Parameter value unit Puncture needle needle outer diameter 1.1 millimeters Puncture needle Needle tip bevel length 3.0 millimeters catheter catheter length 500 millimeters catheter outer diameter of the catheter 1.6 millimeters

[0048] In practice, the puncture path planning module selects the point closest to the skin surface from the initial puncture point set as the planning starting point. The distance from the skin surface is determined by calculating the shortest Euclidean distance from each point in the initial puncture point set to the skin surface model within the 3D vascular structure model, which includes vessel depth, diameter, and positional relationship with surrounding tissues. Starting from the planning starting point, the puncture path planning module searches for a path along the vessel centerline towards the heart chamber within the 3D vascular structure model. The path search process employs a graph-based search algorithm, discretizing the vessel centerline into a series of nodes, with connections between nodes representing possible path segments. During the path search, the puncture path planning module reads the catheter flexibility parameters in real time. These parameters are expressed as minimum allowable bending radius or maximum allowable curvature, limiting the allowable bending curvature of the path. When the search path encounters an anatomical variation region marked by the anatomical variation analysis module, the puncture path planning module calculates a detour path based on the catheter flexibility parameters, ensuring the curvature of the detour path remains within the allowable range of the catheter. The search path stops when it reaches the preset endpoint of the central vein in the model. The preset endpoint of the central vein is the coordinate point in the model where the superior vena cava and the right atrium meet. The recorded continuous search trajectory is the target path from the planning starting point to the preset endpoint of the central vein.

[0049] The puncture path planning module samples along the target path at fixed intervals, with the sampling interval set according to the path length and accuracy requirements, resulting in a series of ordered spatial site coordinates. For each spatial site coordinate, the module matches a key image generated by the 3D vascular modeling module, extracting typical texture features of the blood vessel corresponding to the spatial site coordinate in the ultrasound image as ultrasound image recognition features. This extraction process involves locating the pixel region corresponding to the spatial site coordinate in the key image and calculating the gray-level co-occurrence matrix features or local binary pattern features of that region. The puncture path planning module defines the basic operations required for the puncture needle or catheter for the path segment between two adjacent spatial sites. These basic operations include the needle insertion angle, rotation direction, and advancement distance, which are encoded as puncture instrument operation instructions. These instructions are represented in a machine-readable instruction set, such as "needle insertion angle α, rotation direction clockwise, advancement distance d mm". The puncture path planning module sequentially binds the spatial site coordinates, ultrasound image recognition features, and puncture instrument operation instructions to form an executable path plan. An executable path plan is a data structure in which each entry contains a spatial location coordinate, the corresponding ultrasound image recognition feature, and the puncture instrument operation instructions for moving from the previous location to the current location.

[0050] Optionally, when extracting ultrasound image recognition features from key images, the puncture path planning module employs a feature fusion method. The module extracts not only texture features but also shape features of the vessel boundary, such as contour curvature and Fourier descriptors. It then combines these texture and shape features into a multi-dimensional feature vector as the ultrasound image recognition feature for the spatial location coordinates, thereby improving the robustness of feature matching during subsequent navigation.

[0051] In some embodiments, when defining puncture instrument operation instructions, the puncture path planning module considers the kinematic constraints of the instrument. For puncture from a spatial location... to adjacent spatial sites The path segment, the needle insertion angle is calculated by vector. At the skin surface The angle between the normal vectors at the location is obtained, the rotation direction is determined by comparing the difference between the current catheter tip mark and the target orientation, and the advancement distance is the spatial location. and The straight-line distance between them. The encoding process quantizes the calculated angle, direction, and distance values ​​into discrete instruction codes.

[0052] See Figure 5This diagram illustrates the complete process of puncture path planning. The surface morphology of the blood vessel is outlined by light blue dot clouds, while the solid blue line indicates the direction of the vessel's centerline. Yellow squares represent the candidate puncture space calculated by the system, green dots represent the initial set of puncture points that meet safety conditions, and the most prominent red curve is the target path planned by the system from the skin surface to the central vein. The diagram also clearly marks the locations of key anatomical structures such as nerves and tendons, and uses semi-transparent spheres to indicate safety distance boundaries, effectively highlighting the safety avoidance principle in path planning. Detailed parameters of the puncture needle and catheter are listed on one side of the diagram, emphasizing the specific constraints of instrument specifications on path planning. The overall diagram comprehensively presents the entire process combining intelligent planning with clinical safety considerations.

[0053] Example 5: In specific implementation, the path conflict detection and navigation control module receives an executable path plan generated by the puncture path planning module. The module then retrieves the spatial coordinates of all regions marked as absolutely forbidden from the taboo rule base. The taboo rule base stores spatial constraints corresponding to different anatomical structures and risk levels. The spatial coordinates of regions marked as absolutely forbidden define the enclosed space range in the 3D model where puncture instruments are strictly prohibited from entering or contacting, such as the precise 3D coordinate boundaries of important nerve bundles, arterial walls, or specific tendons. The path conflict detection and navigation control module compares the spatial coordinates of each spatial point in the executable path plan with the spatial coordinates of the absolutely forbidden regions. This comparison is achieved by calculating the shortest distance from the spatial point coordinates to the boundary of the absolutely forbidden region, and determining whether this shortest distance is less than or equal to zero to determine whether spatial invasion has occurred.

[0054] In some embodiments, if spatial intrusion occurs, the path conflict detection and navigation control module replans a local alternative path that avoids the absolutely forbidden region on the path segment near the intrusion point using a path search method. The path conflict detection and navigation control module uses the preceding safe spatial point of the intrusion site as the starting point of the new plan and the following safe spatial point of the intrusion site as the ending point, applying a path search algorithm similar to that used in the puncture path planning module between these two points. During replanning, the path conflict detection and navigation control module sets the absolutely forbidden region as an impassable obstacle and searches for a new path segment connecting the starting and ending points within the space defined by a three-dimensional vascular structure model including vessel depth, diameter, and positional relationship with surrounding tissues. This new path segment is the local alternative path. The path conflict detection and navigation control module replaces the portion of the original executable path plan from the newly planned starting point to the newly planned ending point with the local alternative path, generating a calibrated path plan. If no spatial intrusion occurs, the calibrated path plan is the same as the original executable path plan.

[0055] In practical implementation, the path conflict detection and navigation control module converts all data in the calibrated path plan into a data stream format conforming to the communication protocol of the navigation actuator. The conversion process involves data serialization and instruction encoding. Spatial coordinate instructions in the calibrated path plan are converted into floating-point arrays in the navigation actuator coordinate system, image feature comparison instructions are converted into binary representations of feature vectors, and instrument action instructions are converted into operation codes and parameters recognizable by the navigation actuator controller. The data stream format contains sequentially arranged spatial coordinate instructions, image feature comparison instructions, and instrument action instructions. These three types of instructions are interleaved according to the path execution order, forming a continuous instruction stream. The path conflict detection and navigation control module transmits the path plan in data stream format to the navigation actuator in real time through a data interface, which can be an Ethernet, USB, or industrial bus interface. The navigation actuator parses the received data stream, drives its robotic arm to adjust the position of the ultrasound probe to match the image feature comparison instructions, and guides the puncture needle to move along the spatial coordinate instructions according to the instrument action instructions.

[0056] It is understandable that determining spatial intrusion requires geometric calculations. One method for calculating spatial locations... The formula for the distance to the boundary of a certain absolutely forbidden region (modeled as a convex polyhedron) is expressed as:

[0057] in: This represents the calculated shortest distance. This represents the coordinates of a spatial location within an executable path plan. This represents a surface of a convex polyhedron representing an absolutely forbidden region. Let the set of all surfaces of the convex polyhedron be represented by the function. Return to surface The equation of the plane in which it is located, the function Calculation points to plane The signed vertical distance. When At that time, determine the spatial location. Spatial intrusion occurs when a location is situated within or on the surface of this absolutely forbidden area.

[0058] In some embodiments, after the navigation actuator parses the data stream, its control flow is closed-loop. The navigation actuator first moves the puncture instrument to its approximate position according to spatial coordinate instructions. Then, according to image feature comparison instructions, it drives the ultrasound probe to scan and extract features from the real-time ultrasound image. The extracted real-time features are matched with the ultrasound image recognition features embedded in the instructions. If the feature matching degree is higher than a preset threshold, the navigation actuator executes the next instrument movement instruction; if the matching degree is lower than the threshold, the navigation actuator pauses and feeds back a signal to the path conflict detection and navigation control module. The path conflict detection and navigation control module can issue pause or recalibration instructions according to a preset strategy.

[0059] Optionally, when replanning local alternative paths, the path conflict detection and navigation control module not only considers obstacle avoidance but also optimizes the smoothness of the path. The path conflict detection and navigation control module introduces a smoothness cost term into the path search algorithm, so that the generated local alternative path minimizes sharp corners while satisfying obstacle avoidance and safety curvature constraints, thus facilitating the smooth advancement of the catheter or puncture needle.

[0060] Optionally, the data stream format employs an encapsulation structure with check frames. Each set of spatial coordinate commands, image feature comparison commands, and instrument action commands is encapsulated into a data frame. The header of the data frame contains the frame sequence number and command type, while the tail contains a cyclic redundancy check (CRC) code. Upon receiving the data frame, the navigation execution unit first verifies the integrity of the data. Only after the verification passes will it parse and execute the commands within, ensuring the accuracy and reliability of the commands during transmission.

[0061] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0062] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A peripherally inserted central venous catheterization auxiliary system, characterized in that, The system includes: The data acquisition and verification module is configured to acquire and verify the patient's identity, ultrasound images of the target blood vessel, and anatomical description text. The three-dimensional vascular modeling module is configured to spatially register the ultrasound image with the anatomical description text to generate a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues. The anatomical variation analysis and feasibility assessment module is configured to compare the three-dimensional vascular structure model with a standard vascular atlas library to mark anatomical variation areas, and to assess the puncture feasibility of the target blood vessel based on the anatomical variation areas and a preset contraindication rule library. The puncture path planning module is configured to define a candidate space in the three-dimensional vascular structure model based on the assessed feasible target blood vessel, combined with the physical size parameters of the puncture needle and the specification parameters of the catheter, calculate the initial puncture point set that meets the requirements of avoiding nerves and tendons, and simulate the catheter travel process to generate a target path from the skin to the central vein, and then decompose it into an executable path plan that associates ultrasound image recognition features with puncture instrument operation instructions. The path conflict detection and navigation control module is configured to perform conflict detection and calibration on the executable path plan, and send the calibrated path plan to the navigation execution terminal to control the alignment of the puncture auxiliary device.

2. The peripherally inserted central venous catheterization auxiliary system according to claim 1, characterized in that, The data acquisition and verification module performs verification in the following manner: Receive patient identification codes from the medical information system and retrieve historical medical records and image files associated with the patient identification codes from local storage; Receive a real-time video stream from an ultrasound device and extract one or more key images from the real-time video stream; Receive text annotations about the location of the target blood vessel input by the operator; The historical medical records, image archives, key images, and text annotations are timestamped and their content is verified to be consistent, generating a unified pre-puncture preparation information package.

3. The peripherally inserted central venous catheterization auxiliary system according to claim 2, characterized in that, The three-dimensional blood vessel modeling module generates the three-dimensional blood vessel structure model through the following steps: Apply filtering to key images to reduce speckle noise; Identify the edges of the vascular cavity region in the filtered image and delineate the contour of the vascular wall based on the gray-level gradient changes; The names, relative depths, and keywords of adjacent tissues of blood vessels were extracted from the text annotation information. By combining the blood vessel wall contour with data parsed from text annotation information, preliminary two-dimensional layer data is constructed using contour depth information. Multiple two-dimensional data are stacked and interpolated along the scanning direction of the ultrasound probe to form a three-dimensional vascular structure model with depth dimension, including vascular depth, diameter, and positional relationship with surrounding tissues.

4. The peripherally inserted central venous catheterization auxiliary system according to claim 3, characterized in that, The anatomical variation analysis and feasibility assessment module performs comparison, labeling, and evaluation through the following steps: Call the standard 3D model of the target vessel from the standard vascular atlas library; The three-dimensional vascular structure model, which includes vascular depth, diameter, and positional relationship with surrounding tissues, is spatially aligned and its differences are measured with the standard three-dimensional model. Calculate the geometric deviation between the three-dimensional vascular structure model, which includes vascular depth, diameter, and positional relationship with surrounding tissues, and the standard three-dimensional model, and mark the areas with deviation values ​​exceeding a preset threshold as anatomical variation areas; The taboo rule base is queried, which defines the risk levels and operational taboo clauses corresponding to different types of anatomical variations; Determine the risk level corresponding to the marked anatomical variation area. If the risk level exceeds the allowable range, output the assessment status as infeasible; otherwise, output it as feasible.

5. The peripherally inserted central venous catheterization auxiliary system according to claim 4, characterized in that, The puncture path planning module defines the candidate space in the following manner: Read the diameter and bevel length parameters of the puncture needle, and read the length and outer diameter parameters of the catheter; On the surface of a three-dimensional vascular structure model that is deemed feasible in the evaluation state and includes the vessel depth, diameter and positional relationship with surrounding tissues, a virtual puncture and needle insertion space is offset outward along the normal direction of the vessel wall, with the length of the puncture needle as a reference. Inside a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues, a virtual catheter travel channel is formed by contracting inward along the vascular centerline with reference to the outer diameter of the catheter. The candidate space is obtained by performing a Boolean operation on the intersection of the virtual puncture needle insertion space and the virtual catheter travel channel.

6. The peripherally inserted central venous catheterization auxiliary system according to claim 5, characterized in that, The puncture path planning module calculates the initial puncture point set through the following steps: Obtain the physical dimensions of the ultrasound probe and the angle parameters of the scanning plane; On the outer surface of the candidate space, the placement and orientation of the simulated ultrasound probe are arranged so that the scanning plane can completely cover a section of the candidate space. Calculate the theoretical resolution and sharpness of the ultrasound image under the scanning plane to generate image quality simulation parameters; Within the scanning area that meets the image quality requirements, all surface points whose distance to the nerve and tendon structures marked in the distance model is greater than the safe distance are further screened out. The set of these surface points constitutes the initial puncture point set.

7. The peripherally inserted central venous catheterization auxiliary system according to claim 6, characterized in that, The puncture path planning module generates a target path from the skin to the central vein, including: In the initial set of puncture points, select the point closest to the skin surface as the planning starting point; In a three-dimensional vascular structure model that includes vascular depth, diameter, and positional relationship with surrounding tissues, a path search is performed from the planning starting point along the vascular centerline toward the heart chamber. During the path search process, the catheter flexibility parameters are read in real time, which limit the allowable curvature of the path. When the search path encounters an area of ​​anatomical variation, the detour path is calculated based on the catheter's flexibility parameters to ensure that the curvature of the detour path is within the catheter's allowable range. The search path stops when it reaches the preset endpoint of the model's central vein, and the recorded continuous search trajectory is the target path.

8. The peripherally inserted central venous catheterization auxiliary system according to claim 7, characterized in that, The puncture path planning module generates the executable path plan through the following steps: Sampling is performed along the target path at fixed intervals to obtain a series of ordered spatial location coordinates; For each spatial location coordinate, the key image generated in step three is matched, and the typical texture features of the blood vessels corresponding to the spatial location coordinates in the ultrasound image are extracted as ultrasound image recognition features. For the path segment between two adjacent spatial sites, the basic operations required for the puncture needle or catheter are defined. These basic operations include needle insertion angle, rotation direction and advance distance. These basic operations are encoded as puncture instrument operation instructions. By sequentially binding spatial location coordinates, ultrasound image recognition features, and puncture instrument operation instructions, an executable path plan is formed.

9. The peripherally inserted central venous catheterization auxiliary system according to claim 8, characterized in that, The path conflict detection and navigation control module performs conflict detection and calibration through the following steps: Retrieve the spatial coordinates of all regions marked as absolutely forbidden from the taboo rule library; The coordinates of each spatial point in the executable path plan are compared with the spatial coordinates of the absolutely forbidden area to determine whether spatial location intrusion has occurred. If a spatial intrusion occurs, a local alternative path that avoids the absolutely forbidden area will be replanned on the path segment near the intrusion point, based on the path search method in step seven. Replace the corresponding part of the original path plan with a local alternative path to generate a calibrated path plan.

10. The peripherally inserted central venous catheterization auxiliary system according to claim 9, characterized in that, The path conflict detection and navigation control module sends the path plan to the navigation execution terminal through the following steps: Convert all data in the calibrated path plan into a data stream format that conforms to the navigation execution terminal communication protocol; The data stream format includes spatial coordinate instructions, image feature comparison instructions, and instrument action instructions arranged in sequence. The route plan in data stream format is transmitted to the navigation execution end in real time via the data interface; The navigation execution end parses the data stream, drives its robotic arm to adjust the position of the ultrasound probe to match the image feature comparison instructions, and guides the puncture needle to move along the spatial coordinate instructions according to the instrument action instructions.

Images