An intelligent analysis method and system for ultrasonic blood vessels

Abstract

The present application relates to an ultrasonic blood vessel intelligent analysis method and system, the method comprises: based on the real-time image frame stream input and preprocessing of ultrasonic operation request; based on the blood vessel boundary box coordinates and confidence, determine whether the current image frame is a valid image frame; the operation suggestion module inserts the current image frame into the graph structure for the detection effective image frame; using the graph structure to save the detection effective image frame; determine whether to carry out the withdrawal operation; determine the withdrawal position and withdrawal operation mode when necessary. The present application combines big data and artificial intelligence technology, so that individual intelligent ultrasonic operation guidance can be carried out for different operators, the module parameters are customized based on different ultrasonic operation plans, and the intelligent suggestion model is loaded according to the stage of different operators, each different module can be independently upgraded, and the ultrasonic operation efficiency is significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of smart medical technology, and in particular relates to an intelligent ultrasound vascular analysis method and system. Background Technology

[0002] How to leverage rapidly developing big data and artificial intelligence technologies to improve the efficiency of ultrasound procedures and create personalized ultrasound procedure advice systems is a key technical challenge. To address this, this invention integrates big data and artificial intelligence technologies, enabling personalized intelligent ultrasound procedure guidance for different operators. It customizes module parameters based on different ultrasound procedure plans and loads intelligent advice models according to the operator's stage of development, providing personalized ultrasound guidance paths tailored to the operator's proficiency level, thus significantly improving ultrasound procedure efficiency. Summary of the Invention

[0003] To address the aforementioned problems in the prior art, this invention proposes an intelligent ultrasound vascular analysis method and system, the method comprising:

[0004] Step S1: Real-time image frame stream input and preprocessing based on ultrasound operation request;

[0005] Step S2: The vascular quality assessment module determines whether the current image frame is a valid image frame based on the vascular bounding box coordinates and confidence level;

[0006] The operation suggestion module determines whether a valid image frame is standard based on the clarity and morphology of the blood vessel imaging, inserts the current image frame into the graph structure, and provides operation suggestions.

[0007] The image frame management module uses a graph structure to store the detected valid image frames. For a valid current image frame, if it is a standard image frame, it is made a child node of all leaf nodes of the node corresponding to the first standard image frame in its preorder nodes. Otherwise, the prediction adjustment number from the current image frame to the standard image frame is determined, and the current image frame is made a child node of the current node. The current node is then updated to the node where the current image frame is located.

[0008] Step S3: The image frame management module determines whether to perform a withdrawal operation; if so, it determines the withdrawal position; based on the difference between the operation method at the withdrawal position and the operation method of the current image frame, it determines the withdrawal operation method; it returns to step S1 until the planned operation time length; wherein:

[0009] Upon receiving the request to insert a graph structure, the path length between the node containing the nearest standard image frame and the current node is determined. If the path length is greater than a preset path length, a withdrawal operation is performed.

[0010] When the withdrawal method is maximum withdrawal, the node containing the nearest standard image frame to the current node is taken as the withdrawal position; otherwise, the node with the fewest prediction adjustments between the nearest standard image frame to the current node and the current node is taken as the withdrawal position; the node at the withdrawal position is taken as the current node.

[0011] Furthermore, the medical-specific interface uses the DICOM Streaming or RTSP protocol and is encoded in H.264.

[0012] Furthermore, when using a graph structure to save image frames, the standard score of the image frame, the acquisition method, the location of blood vessels, the acquisition timestamp, etc. are saved as node attributes.

[0013] Furthermore, determining the number of prediction adjustments from the current image frame to the standard image frame specifically involves: analyzing the historical image frame sequence, taking historical image frames similar to the current image frame as target image frames, determining the standard image frame closest to the target image frame in the historical image frame sequence, and taking the number of historical image frames between them as the number of adjustments; repeating the above steps until the historical image frame sequence is completely processed to obtain one or more adjustment numbers; and taking the feature value of the adjustment number as the number of prediction adjustments.

[0014] Furthermore, step S1 specifically involves: the ultrasound device pushing image frame streams in real time through a medical-specific interface; the API gateway performing identity verification and format validation on the image frame streams, filtering out invalid pushed image frame streams, and pushing the filtered current image frame to the vascular quality assessment module in real time.

[0015] Further, in step S4: the image frame management module determines a personalized path based on the graph structure and provides it to the operator; specifically: the image frame management module traverses the graph structure starting from the root node, determines the personalized graph path from the root node to the leaf node, organizes the acquisition methods of all nodes involved in the personalized graph path in chronological order to obtain the personalized path; and provides the personalized path to the operator; the personalized path includes the shortest operation path, the highest quality path, and / or the fastest achievement path, and the corresponding personalized graph path is the path with the shortest node span, the path with the highest average standard score, and / or the shortest time path.

[0016] Furthermore, after completing the request to insert the graph structure, it is determined whether an operation reversal and / or forward operation is required.

[0017] Furthermore, the vascular quality assessment module is used to detect target blood vessels; the vascular quality assessment model is equipped with a pre-trained vascular detection model; the input image frames are sent to the vascular detection model; the vascular detection model performs inference on each image frame and outputs the coordinates and confidence score of the vascular bounding box; based on the coordinates and confidence score of the vascular bounding box, it is determined whether the current image frame is a valid image frame.

[0018] An intelligent ultrasound vascular analysis system is provided, which is used to implement the aforementioned intelligent ultrasound vascular analysis method.

[0019] An intelligent ultrasound vascular analysis platform is provided, which is used to implement the above-mentioned intelligent ultrasound vascular analysis method.

[0020] The beneficial effects of this invention include:

[0021] (1) Image frame management based on directed graph structure can obtain detection experience for the current image frame from the big data of image frames containing existing monitoring experience, guide the current operator to conduct standardized operation simulation training of ultrasound section and form a directed graph representing personalized operation path; and can provide personalized path based on the directed graph to optimize ultrasound operation.

[0022] (2) The integration of big data and artificial intelligence technologies enables personalized intelligent ultrasound operation guidance for different operators, customized module parameters based on different ultrasound operation plans, and intelligent suggestion models loaded and upgraded according to the stage of different operators, providing personalized ultrasound guidance paths based on the operator's proficiency, which significantly improves the efficiency of ultrasound operation. Attached Figure Description

[0023] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to unduly limit the invention. In the drawings:

[0024] Figure 1 This is a schematic diagram of the ultrasonic intelligent vascular analysis method provided by the present invention.

[0025] Figure 2 This is a schematic diagram of a directed graph provided in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the initialization of the ultrasonic vascular intelligent analysis system provided by the present invention. Detailed Implementation

[0027] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. The illustrative embodiments and descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0028] This invention proposes an intelligent ultrasound vascular analysis method and system, as shown in the attached figure. Figure 1 As shown, the method includes the following steps:

[0029] Step S1: Real-time image frame stream input and preprocessing based on ultrasound operation request; specifically: the ultrasound device pushes image frame stream in real time through a medical-specific interface; the API gateway performs identity verification and format validation on the image frame stream, filters invalid pushed image frame streams, and pushes the filtered current image frame to the vascular quality assessment module in real time.

[0030] Preferably, the medical-specific interface uses the DICOM Streaming or RTSP protocol and is encoded using H.264.

[0031] Preferably, the authentication is a dual authentication based on the operator's identity and / or the device ID;

[0032] The invalid filtering of the real-time pushed image frame stream is specifically as follows: dynamic frame extraction processing of the image frame stream is performed; motion-aware adaptive frame extraction is adopted; initially, 2 frames are extracted per second by default, and when the motion amount calculated by the inter-frame difference method is greater than the preset value of motion amount, it is determined that the probe movement is detected, and the speed is automatically increased to 5 frames per second; consecutive similar frames in the extracted frames are deduplicated to ensure input diversity.

[0033] Preferably, the image frames are organized as image data files, which contain image data and acquisition methods, including probe pose, scanning depth, pressure parameters and / or gain, etc.

[0034] Preferably, the ultrasound operation request includes the operator's identity and device ID, and the ultrasound operation plan; the ultrasound operation plan includes the planned operation time, the targeted vascular region, the scoring criteria, the preset path length, the withdrawal method, the acceptable withdrawal length, the forward feedback time, and the N value; an API gateway, a vascular quality assessment module, an operation suggestion module, and an image frame management module are loaded based on the ultrasound operation request; the ultrasound operation plan is formulated for the current operator.

[0035] Step S2: The vascular quality assessment module is used to detect target blood vessels; the vascular quality assessment model is equipped with a pre-trained vascular detection model; the input image frames are sent to the vascular detection model; the vascular detection model performs inference on each image frame and outputs the coordinates and confidence score of the vascular bounding box; based on the coordinates and confidence score of the vascular bounding box, it is determined whether the current image frame is a valid image frame;

[0036] Preferably, the blood vessel detection model is the YOLOv5s detection model;

[0037] The step of determining whether the current image is a valid image frame based on the coordinates and confidence of the blood vessel bounding box is as follows: determining whether blood vessels are detected in the current image frame based on the coordinates and confidence of the blood vessel bounding box; if blood vessels are detected at least twice in the three most recent image frames containing the current image frame, then the current image frame is determined to be valid.

[0038] Preferred: When the output confidence level is >0.7, it is determined that a blood vessel has been detected;

[0039] Preferred: If the output confidence level is >0.7, the area of ​​the detected blood vessel region is <0.5% of the total image area, the blood vessel is continuously visible, and the aspect ratio is within the standard range of 1:2 to 1:4, then it is determined that a blood vessel has been detected.

[0040] Step S2 further includes: the operation suggestion module is used to determine whether the detected valid image frames are standard based on the clarity and morphology of the blood vessel imaging, and to provide operation suggestions to the operator to assist in obtaining high-quality images; specifically: the operation suggestion module is used to determine whether the detected valid image frames are standard based on the clarity and morphology of the blood vessel imaging, access the image frame management module to insert the current image frame into the image structure; and provide operation suggestions based on the feedback from the image frame management module.

[0041] The determination of image sharpness and morphology standards based on vascular imaging specifically involves: calculating the sharpness of the vessel wall edge using the gradient magnitude of the Sobel operator; detecting the signal-to-noise ratio (SNR) of the blood flow region using the ratio of signal intensity to background noise within the region of interest (ROI); determining an image sharpness score based on the relationship between edge sharpness and a sharpness threshold, and the relationship between SNR and a SNR threshold; determining a vascular morphology score based on the relationship between the aspect ratio of the measured vessel and the ideal range, and the relationship between the smoothness of the vessel contour and a rate of change of curvature threshold; determining an image frame score based on the image sharpness score and the vascular morphology score; and determining a standard image frame if the image frame score is greater than the scoring standard, otherwise, a non-standard image frame.

[0042] Preferably, the scoring criteria are preset values ​​obtained from the ultrasound operation request, which are related to the operational requirements of the current ultrasound operation request; the sharpness threshold, signal-to-noise ratio threshold, ideal range, and rate of change of curvature threshold are preset values; for example, the sharpness threshold is 0.8, the signal-to-noise ratio threshold is 15dB; the ideal range is 1.5~3.0; and the rate of change of curvature threshold is... Of course, the settings of these thresholds are related to the blood vessel area being detected and the ultrasound equipment used; generally, these thresholds are fixed values ​​in the operation suggestion module.

[0043] The image frame management module uses a graph structure to store valid image frames and their acquisition methods. Specifically, upon receiving a request to insert a valid image frame into the graph structure, if the current image frame is a standard image frame, it is made a child node of all leaf nodes of the node corresponding to the first standard image frame in its preorder nodes. If the current image frame is a non-standard image frame, the prediction adjustment number from the current image frame to the standard image frame is determined, and the current image frame is made a child node of the current node. The current node is then updated to the node containing the current image frame. In the graph structure, nodes represent image frames, and edges represent relationships between image frames. When using the graph structure to store image frames, the standard score, acquisition method, blood vessel location, and acquisition timestamp of the image frame are stored as node attributes.

[0044] Preferably, each node has a unique node identifier;

[0045] Preferably, the root node of the graph structure is the first acquired standard image frame / image frame, recording the complete acquisition method; the graph structure is a directed graph.

[0046] The determination of the prediction adjustment number from the current image frame to the standard image frame specifically involves: analyzing the historical image frame sequence, taking historical image frames similar to the current image frame as target image frames, determining the standard image frame closest to the target image frame in the historical image frame sequence, and taking the number of historical image frames between them as the adjustment number; repeating the above steps until the historical image frame sequence is processed to obtain one or more adjustment numbers; using the feature value of the adjustment number as the prediction adjustment number; since the number of historical image frames similar to the current image frame may be one or more, the obtained adjustment number may also be one or more.

[0047] Preferred method: Based on the image frame acquisition method, the similarity of image frames is determined by a combination of one or more of the image features, image sharpness score and / or vascular morphology score of the image frame;

[0048] Preferably, the characteristic value of the number of adjustments is the average or minimum value of the one or more adjustment numbers;

[0049] Alternative: Cluster the number of adjustments to obtain one or more adjustment number classification sets, and use the mean of the adjustment number of the smallest adjustment number classification set as the feature value of that adjustment number;

[0050] The historical image frame sequence is a sequence of valid image frames obtained when detecting the same vascular region and / or using the same type of ultrasound equipment. This historical image frame sequence can be obtained when the ultrasound equipment is being tested, when a skilled operator is conducting the test, or when an operator with similar testing experience to the current operator is conducting the test.

[0051] Step S3: Determine whether an operation needs to be withdrawn. If so, determine the withdrawal location and withdrawal operation method. Specifically: the image frame management module determines whether a withdrawal operation needs to be withdrawn; if so, it determines the withdrawal location; the operation suggestion module determines the withdrawal operation method based on the difference between the operation method at the withdrawal location and the operation method of the current image frame; and feeds back the withdrawal operation method to the operator; return to step S1 until the planned operation time length.

[0052] Preferred: After completing the request to insert the diagram structure, determine whether an operation reversal and / or forward operation is required;

[0053] The retraction operation mode is the difference between the current operation mode and the operation mode at the retraction position; the operation mode includes probe pose and equipment parameters, etc.; for example: probe displacement vector, pressure adjustment amount and changes in equipment parameters; for example: in the retraction operation mode, the probe moves 3mm to the right, reduces the pressure by 0.2N, and increases the gain by +2dB;

[0054] The image frame management module determines whether a retraction operation is needed by: upon receiving an insertion graph structure request, determining the path length between the node containing the nearest standard image frame and the current node; if the path length is greater than a preset path length, determining that a retraction operation is needed; thus avoiding repeated invalid operations and reinforcing erroneous operation methods.

[0055] The determination of the withdrawal position is specifically as follows: when the withdrawal method is maximum withdrawal, the node containing the standard image frame closest to the current node is taken as the withdrawal position; otherwise, the node with the fewest prediction adjustments between the node containing the standard image frame closest to the current node and the current node is taken as the withdrawal position; and the node at the withdrawal position is taken as the current node.

[0056] Alternative: Among nodes whose path length to the current node is less than the acceptable withdrawal length, the node with the fewest predicted adjustments is selected as the withdrawal position; Further: The acceptable withdrawal length is less than the preset length; It can find a feasible operation path based on the operator's operating habits in the case of incorrect operation mode;

[0057] Step S3 further includes: the image frame management module determines whether a forward operation is required; if so, the image frame sequence between the standard image frame closest to the current node and the current image frame is input into the intelligent suggestion model to obtain a suggestion acquisition method; the suggestion acquisition method is fed back to the operation suggestion module, and the operation suggestion module determines the difference between the suggestion acquisition method and the current acquisition method to constitute a forward operation;

[0058] The intelligent suggestion model is an artificial intelligence model pre-trained using a standard historical image frame sequence; the standard historical image frame sequence is an image frame sequence obtained by a skilled detection operator when performing detection on the same vascular region and / or using the same type of ultrasound equipment;

[0059] The image frame management module determines whether a forward operation is needed, specifically: if no standard image frame has been obtained within the forward feedback time, and N consecutive image frames are non-standard image frames; for example: in carotid artery ultrasound examination, if two consecutive frames show blurred blood vessel walls, and the most recent standard image frame is the image frame acquired 10 seconds ago, a forward operation is suggested through the intelligent suggestion model, reducing the probe tilt angle by 2° and increasing the gain by 2dB; the forward feedback time and the N value are both derived from the ultrasound operation request;

[0060] Alternatives: Forward feedback duration and N value are inherent settings in the image frame management module;

[0061] Table 1: Example of node management in the image frame management module;

[0062]

[0063] Preferably, the method further includes step S4: the image frame management module determines a personalized path based on the graph structure and provides it to the operator; specifically, the image frame management module traverses the graph structure starting from the root node, determines the personalized graph path from the root node to the leaf node, organizes the acquisition methods of all nodes involved in the personalized graph path in chronological order to obtain the personalized path; and provides the personalized path to the operator; since the operator may obtain the standard image frame through multiple paths starting from an image frame and its acquisition method when making multiple attempts, this is related to the detection methods and detection preferences of different operators; through the image frame management of the graph structure, detection experience for the current image frame can be obtained from the big data of image frames containing existing monitoring experience, guiding the current operator to conduct ultrasonic section standardized operation simulation training;

[0064] Preferably, the personalized path includes the shortest operation path, the highest quality path, and / or the fastest achievement path, and the corresponding personalized graph path is the path with the shortest node span, the path with the highest average standard score (based on standard score), and / or the shortest time path (based on timestamp).

[0065] Preferably, the leaf node is any leaf node, some leaf nodes, or one of all leaf nodes;

[0066] Preferred: After the planned operation time is completed, proceed to step S4;

[0067] As attached Figure 2As shown, during an operation, the standard score was set to 80 points; nodes F0 to F9 were generated, and a graph structure was constructed based on their scores; this graph structure is a directed graph; thus, the operator can form their own operating habits based on big data guidance while completing the ultrasound operation request; new operating methods and habits are formed by merging and expanding branches through rollback.

[0068] As can be seen, the shortest operation path is F0→F2→F4→F8, with a cumulative displacement of (+1.0-0.5+0.5,-0.3+0.4-0.1)=(+1.0x,0.0y); equipment parameter adjustments: gain from 65dB→71dB, frequency stabilized at 7.5MHz; the score change path is 85→82→88→91 points, with an average score increment of 2 points; by explicitly inheriting the standard image frame relationship, the digital transfer of operational experience is realized, and the causal relationship of operations is explicitly modeled. Compared with the linear sequence management method, it supports different ranges of local attempts by the operator, thus improving the efficiency of ultrasound lifting.

[0069] Preferred method: After completing the ultrasound operation request, generate an operation request completion report by traversing the graph structure; the report includes: shortest operation path, longest operation path, average score increment, etc.

[0070] Based on the same inventive concept, the present invention also provides an intelligent ultrasound vascular analysis system, which is used to implement the above-mentioned intelligent ultrasound vascular analysis method; the system includes an API gateway, a vascular quality assessment module, an operation suggestion module, and an image frame management module; after receiving an ultrasound operation request, it generates an ultrasound task corresponding to the operator, and creates, loads, and initializes the API gateway, vascular quality assessment module, operation suggestion module, and image frame management module for the ultrasound task;

[0071] Preferred configuration: After loading the ultrasound task, context loading is performed, and the module parameters are initialized based on the patient's historical DICOM data (such as previous vessel diameters) (e.g., various thresholds and ranges during standard scoring); further, the device preset parameters and operator habit configuration (left-handed / right-handed mode) are configured based on the ultrasound operation request.

[0072] Preferred: The ultrasound operation suggestions and the current image frame are displayed synchronously;

[0073] As attached Figure 3As shown, during system initialization, the ultrasound device sends an ultrasound task start request to the API gateway via the HTTP protocol. The API gateway simultaneously triggers the parallel initialization of the vascular quality assessment module, operation suggestion module, and image frame management module. The quality assessment module loads the pre-trained vascular detection model; the operation suggestion module loads the historical image frame sequence and intelligent suggestion model; and the image frame management module constructs a dynamic graph structure.

[0074] The quality assessment module, operation suggestion module, and image frame management module independently execute initialization tasks, synchronizing their progress through shared memory. Once all modules have completed initialization, the API gateway returns a 201 Created response, which includes the task ID and initial parameter configuration.

[0075] A computer program (also referred to as a program, software, software application, script, or code) can be written in any form of programming language, including assembly or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but does not necessarily, correspond to a file in a file system. A program can be stored as part of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to said program, or in multiple co-located files (e.g., a file storing one or more modules, subroutines, or code portions). A computer program can be deployed to execute on a single computer or on multiple computers located at a single site or distributed across multiple sites and interconnected by a communications network.

[0076] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0077] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0078] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0079] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. An intelligent ultrasound vascular analysis method, characterized in that, The method includes: Step S1: Real-time image frame stream input and preprocessing based on ultrasound operation request; Step S2: The vascular quality assessment module determines whether the current image frame is a valid image frame based on the vascular bounding box coordinates and confidence level; The operation suggestion module determines whether a valid image frame is standard based on the clarity and morphology of the blood vessel imaging, inserts the current image frame into the graph structure, and provides operation suggestions. The image frame management module uses a graph structure to store valid image frames. For a valid current image frame, if it is a standard image frame, it is made a child node of all leaf nodes of the node corresponding to the first standard image frame in its preorder nodes; otherwise, the prediction adjustment number from the current image frame to the standard image frame is determined, and the current image frame is made a child node of the current node. Update the current node to the node containing the current image frame; Step S3: The image frame management module determines whether to perform a retraction operation; if so, it determines the retraction position; and determines the retraction operation method based on the difference between the operation method of the retraction position and the operation method of the current image frame. Return to step S1 until the planned operation time length; where: Upon receiving the request to insert a graph structure, the path length between the node containing the nearest standard image frame and the current node is determined. If the path length is greater than a preset path length, a withdrawal operation is performed. When the withdrawal method is maximum withdrawal, the node containing the nearest standard image frame to the current node is taken as the withdrawal position; otherwise, the node with the fewest prediction adjustments between the node containing the nearest standard image frame to the current node and the current node is taken as the withdrawal position; the node at the withdrawal position is taken as the current node. Step S4: The image frame management module determines the personalized path based on the graph structure and provides it to the operator. The personalized path is determined in the following way: Image frame management based on directed graph structure obtains the detection experience for the current image frame from the big data of image frames containing existing detection experience, guides the current operator to perform ultrasonic section standardized operation simulation training and forms a directed graph representing the personalized operation path, and provides the personalized path based on the directed graph.

2. The ultrasonic vascular intelligent analysis method according to claim 1, characterized in that, Step S1 specifically involves: the ultrasound device pushing image frame streams in real time through a medical-specific interface; the API gateway performing identity verification and format validation on the image frame streams, filtering out invalid pushed image frame streams, and pushing the filtered current image frame to the vascular quality assessment module in real time.

3. The ultrasonic intelligent vascular analysis method according to claim 2, characterized in that, When using a graph structure to save image frames, one or more of the image frame’s standard score, acquisition method, blood vessel location, and / or acquisition timestamp are saved as node attributes.

4. The ultrasonic vascular intelligent analysis method according to claim 3, characterized in that, The method for determining the prediction adjustment number from the current image frame to the standard image frame is as follows: analyze the historical image frame sequence, take the historical image frames that are similar to the current image frame as the target image frame, determine the standard image frame that is closest to the target image frame in the historical image frame sequence, and take the number of historical image frames between the two as the adjustment number. Repeat the above steps until the historical image frame sequence has been processed to obtain one or more adjustment counts; use the feature value of the adjustment count as the predicted adjustment count.

5. The ultrasonic intelligent vascular analysis method according to claim 4, characterized in that, Medical-specific interfaces use DICOM Streaming or RTSP protocols and are encoded in H.

264.

6. The ultrasonic vascular intelligent analysis method according to claim 5, characterized in that, Step S4 specifically involves: the image frame management module traversing the graph structure starting from the root node, determining the personalized graph path from the root node to the leaf node, organizing the acquisition methods of all nodes involved in the personalized graph path in chronological order to obtain the personalized path; providing the personalized path to the operator; the personalized path includes the shortest operation path, the highest quality path, and / or the fastest achievement path, and the corresponding personalized graph path is the path with the shortest node span, the path with the highest average standard score, and / or the shortest time path.

7. The ultrasonic intelligent vascular analysis method according to claim 6, characterized in that, After completing the request to insert the graph structure, determine whether an operation reversal and / or forward operation is required.

8. The ultrasonic vascular intelligent analysis method according to claim 7, characterized in that, The vascular quality assessment module is used to detect target blood vessels; the vascular quality assessment module is equipped with a pre-trained vascular detection model; the input image frames are sent to the vascular detection model; the vascular detection model performs inference on each image frame and outputs the coordinates and confidence score of the vascular bounding box; based on the coordinates and confidence score of the vascular bounding box, it is determined whether the current image frame is a valid image frame.

Images