Cerebrovascular hemodynamic measurement method and device based on transcranial color Doppler

Abstract

The invention discloses a cerebral vascular hemodynamic measurement method and device based on transcranial color Doppler. The method comprises the following steps: training a plurality of sample images to establish a multi-target cerebrovascular segmentation model; inputting a to-be-segmented image into the multi-target cerebrovascular segmentation model to obtain a segmentation region of each cerebrovascular in the to-be-segmented image; performing central axis transformation on the segmented region of each cerebral vessel, and extracting a skeleton structure of each cerebral vessel; and based on the skeleton structure of each cerebral vessel, calculating to obtain an optimal measurement position and an optimal measurement angle for measuring cerebral vessel hemodynamic parameters. According to the method, more accurate analysis can be provided for diagnosis and evaluation of cerebral vessels under transcranial color Doppler, errors caused by manual operation are reduced, and improvement of accuracy and reliability of medical image analysis is facilitated.

Description

Technical Field

[0001] This invention relates to medical imaging, and more particularly to a method and apparatus for measuring cerebral hemodynamics based on transcranial color Doppler. Background Technology

[0002] Cerebrovascular diseases are the second leading cause of death worldwide and have a high rate of disability, seriously threatening human health. The Circle of Willis (CoW), as a core structure for compensatory regulation of cerebral blood flow, directly affects the stability of blood supply to brain tissue. Studies show that only about 20% of the population has an intact Circle of Willis, and structural variations may lead to a decline in cerebral blood flow reserve capacity. Transcranial color Doppler (TCCD) is a very important clinical imaging technique that can help doctors diagnose cerebrovascular diseases, such as stenosis, hemorrhage, and aneurysms, more quickly and efficiently. However, due to the limitations of the skull's acoustic window and the complexity of vascular structures, the TCD ultrasound examination process relies heavily on the operator's experience, and the detection and analysis process is very time-consuming; the operator's interpretation of the images has a certain degree of subjectivity, affecting the consistency of parameter interpretation.

[0003] Therefore, there is an urgent need for a technical solution that can automatically identify cerebral vascular structures and intelligently recommend measurement locations and angles to improve the efficiency and reliability of TCCD in cerebral vascular assessment. Summary of the Invention

[0004] The main objective of this invention is to provide a method and device for measuring cerebral hemodynamics based on transcranial color Doppler ultrasound.

[0005] To achieve the above objectives, the present invention provides a method for cerebrovascular analysis based on transcranial color Doppler ultrasound, comprising:

[0006] Training multiple sample images to establish a multi-target cerebral blood vessel segmentation model;

[0007] The image to be segmented is input into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented;

[0008] A midline transformation is performed on the segmented regions of each cerebral blood vessel to extract the skeletal structure of each vessel; and

[0009] Based on the skeletal structure of each cerebral blood vessel, the optimal measurement location and angle for measuring cerebral blood vessel hemodynamic parameters are calculated.

[0010] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by this invention, the step of training multiple sample images to establish a multi-target cerebral blood vessel segmentation model includes:

[0011] Multiple cerebral blood vessels in each sample image are segmented to obtain labeled sample data containing the segmented regions of each cerebral blood vessel. The sample images include multiple sets of transcranial color Doppler cerebral vascular image data.

[0012] A training dataset is constructed based on the labeled sample data and the corresponding sample images;

[0013] Based on the training dataset, the multi-target cerebral blood vessel segmentation model is constructed.

[0014] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by this invention, the step of region segmentation of multiple cerebral blood vessels in each sample image includes:

[0015] The multiple sample images are divided into a template image group and a training image group, and the training image group is calibrated using the template image group;

[0016] Extract segmented regions of multiple cerebral blood vessels from each sample image in the calibrated training image set.

[0017] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by this invention, in the step of extracting segmented regions of multiple cerebral blood vessels in each sample image of the calibrated training image group, the segmented regions of multiple cerebral blood vessels are extracted by using traditional image processing methods by utilizing the difference in color or density between the blood vessel region and the surrounding tissue.

[0018] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by the present invention, in the step of extracting the segmented regions of multiple cerebral blood vessels in each sample image of the calibrated training image group, the segmented regions of multiple cerebral blood vessels in each sample image are obtained by hand drawing.

[0019] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by this invention, the step of performing a midline transformation on the segmented region of each cerebral blood vessel and extracting the skeletal structure of each cerebral blood vessel includes:

[0020] The segmented regions of each cerebral blood vessel are binarized to obtain a foreground mask image;

[0021] A skeleton extraction algorithm is used to perform a mid-axis transformation on the foreground mask image to obtain a skeleton image including the skeletal structure of cerebral blood vessels.

[0022] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by the present invention, in the step of performing a mid-axis transformation on the foreground mask image using a skeleton extraction algorithm to obtain a skeleton image including the skeleton structure of cerebral blood vessels, the skeleton image including the skeleton structure of cerebral blood vessels is obtained by progressively eroding the boundary pixels of the foreground mask image and retaining the endpoints.

[0023] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by the present invention, in the step of performing a mid-axis transformation on the foreground mask image using a skeleton extraction algorithm to obtain a skeleton image including the skeleton structure of cerebral blood vessels, the distance map from each point in the foreground mask image to the boundary is calculated and the curvature abrupt change points or slope discontinuities in the distance map are extracted as skeletons to obtain a skeleton image including the skeleton structure of cerebral blood vessels.

[0024] In the transcranial color Doppler-based cerebral hemodynamic measurement method provided by this invention, the steps of calculating the optimal measurement position and optimal measurement angle for measuring cerebral hemodynamic parameters based on the skeletal structure of each cerebral blood vessel include:

[0025] Multiple skeleton points are sampled in the skeleton structure;

[0026] A distance transformation is performed on the skeleton points, and the distance from each point to the edge of the blood vessel is calculated as an estimate of the blood vessel thickness.

[0027] Each skeleton point is scored, and the point where the estimated blood vessel thickness exceeds the first preset threshold and the local curvature is lower than the second preset threshold is taken as the best measurement location;

[0028] Based on the local orientation of the neighborhood structure or skeleton at the optimal measurement location, the extension direction of the blood vessel at the optimal measurement location is estimated as the optimal measurement angle.

[0029] According to another aspect of the present invention, a transcranial color Doppler-based cerebral vascular hemodynamic measurement device is also provided for implementing the transcranial color Doppler-based cerebral vascular hemodynamic measurement method as described above, comprising:

[0030] The segmentation model building module is used to train multiple sample images to build a multi-target cerebral blood vessel segmentation model;

[0031] The image segmentation module is used to input the image to be segmented into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented;

[0032] A skeleton structure extraction module is used to perform a midline transformation on the segmented regions of each cerebral blood vessel to extract the skeleton structure of each cerebral blood vessel; and

[0033] The parameter calculation module is used to calculate the optimal measurement position and optimal measurement angle for measuring cerebral blood vessel hemodynamic parameters based on the skeletal structure of each cerebral blood vessel.

[0034] A labeling module is used to label images containing skeletal structures of the brain based on the optimal measurement location and optimal measurement angle;

[0035] The display module is used to display in real time the segmented region of each cerebral blood vessel in the image to be segmented, the skeletal structure of each cerebral blood vessel, and the optimal measurement position and angle of the marker.

[0036] This invention combines image segmentation, skeleton extraction, and measurement point optimization to achieve automatic identification of cerebral vascular structures and intelligent measurement of hemodynamic parameters. It significantly improves the accuracy and efficiency of transcranial color Doppler in cerebral vascular diagnosis, reduces errors caused by human operation, and helps improve the accuracy and reliability of medical image analysis. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort:

[0038] Figure 1 The diagram shows a flowchart of a method for measuring cerebral hemodynamics based on transcranial color Doppler ultrasound according to an embodiment of the present invention.

[0039] Figure 2 The figure shown is a schematic diagram of data annotation for model training in one embodiment of the present invention, which shows the outline information of multiple cerebral vascular structures obtained by manual drawing, including but not limited to the ipsilateral and contralateral middle cerebral artery M1 segment, anterior cerebral artery A1 and A2 segments, posterior cerebral artery P1 and P2 segments, anterior communicating artery, posterior communicating artery, etc., and different colors are used to distinguish each vascular region.

[0040] Figure 3 The diagram shown is a schematic of the model inference results in one embodiment of the present invention. It shows the cerebral blood vessel segmentation results obtained by the segmentation model after processing transcranial color Doppler ultrasound images. Different color regions in the figure correspond to different cerebral blood vessel structures. The segmentation results are stored in the form of Numpy Array and can be used for subsequent skeleton extraction and measurement location calculation.

[0041] Figure 4 The diagram shown is a visual representation of the optimal measurement position and angle in one embodiment of the present invention. Detailed Implementation

[0042] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0044] Figure 1 The diagram shows a flowchart of a cerebrovascular analysis method based on ultrasound data and a large language model, provided by an embodiment of the present invention. Figure 1 As shown, the method for multi-target cerebral vessel segmentation and optimal hemodynamic measurement based on transcranial color Doppler includes the following steps:

[0045] Step S1: Train multiple sample images to establish a multi-target cerebral blood vessel segmentation model;

[0046] Specifically, in one embodiment of the present invention, the sample images include multiple sets of transcranial color Doppler cerebral vascular image data. First, a large amount of transcranial color Doppler ultrasound image data is acquired, which may be derived from clinical collections or public databases. To ensure data quality and consistency with model training, the image data is preprocessed, including but not limited to denoising, contrast enhancement, and image standardization. Subsequently, the preprocessed cerebral vascular ultrasound images are segmented to obtain the contour information of cerebral vessels in each sample image. Then, the above image data and their corresponding annotation information are used to construct a training dataset, which is input into a multi-target cerebral vascular segmentation model for training. Therefore, step S1 includes:

[0047] Step S11: Perform region segmentation on multiple cerebral blood vessels in each sample image to obtain labeled sample data containing the segmented region of each cerebral blood vessel;

[0048] Specifically, in this step, the sample images are obtained by manual drawing of different cerebral blood vessel segmentation regions in the images by relevant professionals. Some sample images are used as template images, and the remaining sample images are registered to extract segmentation regions. Therefore, step S11 includes dividing the multiple sample images into a template image group and a training image group, using the template image group to calibrate the training image group; and extracting multiple cerebral blood vessel segmentation regions from each sample image in the calibrated training image group.

[0049] Furthermore, by utilizing the differences in color or density between the vascular region and surrounding tissues, segmented regions are extracted using traditional image processing methods; alternatively, segmented regions are obtained by manual drawing by professionals. The depicted cerebral blood vessels include, but are not limited to, major cerebral vascular structures such as the ipsilateral and contralateral middle cerebral artery M1 segment, anterior cerebral artery A1 and A2 segments, posterior cerebral artery P1 and P2 segments, anterior communicating artery, and ipsilateral and contralateral posterior communicating artery. Figure 2 The figure shown is a schematic diagram of data annotation used for model training in an embodiment of the present invention. The figure shows the artificial depiction results of the above-mentioned cerebral vascular structure, and different colors are used to distinguish each vascular region.

[0050] Step S12: Construct a training dataset based on the labeled sample data and the corresponding sample images;

[0051] Step S13: Construct the multi-target cerebral blood vessel segmentation model based on the training dataset.

[0052] Specifically, in this step, a model capable of multi-target cerebral blood vessel segmentation is constructed based on the image features of labeled sample data containing segmented regions. The model includes, but is not limited to, models based on dynamic shape models, machine learning models, deep learning models, radiomics models, traditional computer vision algorithm models, or models combining biological marker information.

[0053] Furthermore, in one embodiment, the segmentation model may employ a deep learning model based on the YOLO-11 architecture. To improve the segmentation accuracy of the model under complex vascular structures, an attention mechanism can be introduced into the backbone network of the model, for example, embedding a Linear Attention structure in a deep feature extraction module to enhance the model's ability to perceive spatial information and vascular boundaries.

[0054] Through the above training process, a multi-target cerebral vascular segmentation model capable of automatically identifying and segmenting various cerebral vascular structures is obtained. This model can be used in subsequent steps for rapid and accurate extraction of cerebral vascular regions from the images to be processed.

[0055] Step S2: Input the image to be segmented into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented.

[0056] Specifically, in one embodiment of the present invention, firstly, video acquisition cards with interfaces such as HDMI, DP, and DVI acquire image data in real time from a transcranial color Doppler ultrasound device. The image data is a two-dimensional dynamic image sequence reflecting the direction, velocity, and distribution of blood flow within cerebral blood vessels. During the acquisition process, the video acquisition card converts the HDMI, DP, and DVI signals output by the ultrasound device into a processable digital image stream and transmits it to a computing device for subsequent processing.

[0057] After initial caching and frame synchronization, the acquired image data is fed into the multi-target cerebral blood vessel segmentation model trained in step S1. This model can be deployed on a local server or edge computing device, supporting GPU-accelerated inference to meet real-time processing requirements.

[0058] During model inference, each image frame serves as input, undergoes feature extraction, object detection, and segmentation modules, and outputs the corresponding segmentation result. The segmentation result represents the cerebral vascular region at the pixel level and is stored in NumPy Array format. Specifically, for each cerebral vascular structure (such as the ipsilateral middle cerebral artery M1 segment and the contralateral anterior cerebral artery A1 segment), the model outputs a two-dimensional array, where the values ​​in the array are 0 or 1, representing the pixel distribution of that vessel in the original image.

[0059] In one embodiment, the segmentation results can be categorized and managed according to blood vessel type, resulting in the following structured output:

[0060] ·mask_ACA_A2(Numpy Array)

[0061] ·mask_Ipsilateral_MCA_M1(Numpy Array)

[0062] ·mask_Contralateral_PCA_P1(Numpy Array)

[0063] ·……

[0064] Each array maintains the same size as the original image, facilitating subsequent skeleton extraction, midline transformation, and measurement position calculation.

[0065] In addition, to improve segmentation accuracy and stability, a confidence threshold mechanism can be introduced during model inference to retain only segmented regions with a confidence level higher than a set threshold, in order to avoid false detections or artifact interference.

[0066] Figure 3The diagram shown illustrates the model inference results in one embodiment of the present invention, demonstrating the cerebral blood vessel segmentation results obtained after processing transcranial color Doppler ultrasound images using a segmentation model. Different colored regions in the diagram correspond to different cerebral blood vessel structures. The segmentation results are stored in Numpy Array format and can be used for subsequent skeleton extraction and measurement location calculation.

[0067] The above segmentation process enables real-time and automated segmentation of multiple cerebral vascular structures in transcranial color Doppler images, providing high-quality structural input for subsequent steps.

[0068] Step S3: Perform a midline transformation on the segmented regions of each cerebral blood vessel to extract the skeletal structure of each cerebral blood vessel;

[0069] Specifically, in one embodiment of the present invention, the midline transformation includes binarizing the segmented region to obtain the foreground region, and using a skeleton extraction algorithm to generate a skeleton image representing the main structure of the cerebral blood vessels. Therefore, step S3 includes:

[0070] Step S31: Binarize the segmented region of each cerebral blood vessel to obtain a foreground mask image;

[0071] Specifically, in one embodiment of the present invention, the segmented region is first binarized to obtain a foreground mask image.

[0072] Let the original segmented image be M(x,y), then its binarization result is:

[0073]

[0074] Here, B(x,y) indicates whether the pixel (x,y) belongs to the blood vessel region.

[0075] Step S32: Use a skeleton extraction algorithm to perform a mid-axis transformation on the foreground mask image to obtain a skeleton image including the skeleton structure of cerebral blood vessels.

[0076] Specifically, in one embodiment of the present invention, a morphological thinning method is used to perform a mid-axis transformation by progressively eroding boundary pixels while preserving endpoints until thinning can no longer be performed. That is, a skeleton image including the skeletal structure of the brain's blood vessels is obtained by progressively eroding the boundary pixels of the foreground mask image while preserving endpoints.

[0077] In this embodiment, a skeletonization method based on morphological thinning is adopted. Its core idea is to gradually remove edge pixels while keeping the image topology unchanged, until the image shrinks to a center line that is one pixel wide.

[0078] The processing steps of this algorithm include:

[0079] 1. Edge pixel recognition: In each iteration, edge pixels in the current image are identified, that is, foreground pixels that meet certain neighborhood conditions.

[0080] 2. Topology Preservation Judgment: For each candidate pixel, determine whether its removal would change the connectivity or topology of the image. Commonly used judgment criteria include:

[0081] • Does the neighborhood of a pixel contain multiple connected components?

[0082] • Does removal cause certain areas to break or disappear?

[0083] 3. Pixel Removal: If the topology preservation condition is met, mark the pixel as removable and delete it in the current iteration.

[0084] 4. Iterative shrinkage: Repeat the above steps until the image can no longer be refined, that is, all remaining foreground pixels are skeleton points.

[0085] This process can be formally expressed as:

[0086]

[0087] Among them, T n S(x,y) represents the image state after the nth iteration, and S(x,y) is the final skeleton image.

[0088] Through the above processing steps, a skeleton image S(x,y) is finally obtained, where a pixel value of 1 represents a skeleton point and 0 represents the background. This skeleton structure can be used to calculate the optimal measurement position and angle in the subsequent step S4.

[0089] In another embodiment, or based on a distance transformation method, a distance map from each point in the foreground region to the boundary is calculated, and points of curvature abrupt change or slope discontinuity are extracted as a skeleton. That is, by calculating the distance map from each point in the foreground mask image to the boundary and extracting points of curvature abrupt change or slope discontinuity from the distance map as a skeleton, a skeleton image including cerebral blood vessels is obtained.

[0090] Step S4: Based on the skeletal structure of each cerebral blood vessel, calculate the optimal measurement location and optimal measurement angle for measuring cerebral blood vessel hemodynamic parameters;

[0091] Specifically, in one embodiment of the present invention, based on the skeleton structure extracted in step S3, the optimal measurement location and optimal measurement angle for measuring hemodynamic parameters are calculated. This process includes three sub-steps: vessel thickness estimation, skeleton point scoring, and direction estimation. Specifically, it includes: sampling multiple skeleton points in the skeleton structure; performing distance transformation on the skeleton points and calculating the distance from each point to the vessel edge as an estimated vessel thickness; scoring each skeleton point and identifying points whose estimated vessel thickness exceeds a first preset threshold and whose local curvature is lower than a second preset threshold as optimal measurement locations; and estimating the extension direction of the vessel at the optimal measurement location as the optimal measurement angle based on the neighborhood structure or the local direction of the skeleton orientation.

[0092] First, a distance transformation is performed on the vascular region corresponding to the skeleton structure to estimate the vascular thickness at each skeleton point. Let the binarized vascular mask image be B(x,y), then its distance transformation result is:

[0093]

[0094] in, Let D(x,y) represent the set of pixels at the edge of the blood vessel, and let D(x,y) represent the Euclidean distance from pixel (x,y) to the edge. For skeleton points, this distance value can be regarded as the local radius of the blood vessel, reflecting the thickness of the blood vessel.

[0095] Next, sample several skeleton points from the skeleton image, denoted as set P = {p1, p2, ..., p...}. n}, where each point p i =(x i ,y i To improve computational efficiency, sampling can be performed at regular pixel intervals, such as once every 10 pixels.

[0096] For each sampling point p i Calculate its score S i The scoring function can be defined as:

[0097] S i =D(x) i ,y i )

[0098] Iterate through all sampling points and select the point with the highest score as the optimal measurement location:

[0099]

[0100] Finally, based on the overall structure of the skeleton image, the extension direction of the blood vessels is estimated as the optimal measurement angle. In this embodiment, principal component analysis (PCA) is used to linearly fit the skeleton points and extract their principal axis directions.

[0101] Specifically, the coordinates of all skeleton points are extracted to form a data matrix X = {(x1,y1),(x2,y2),…,(x...} n ,y n The data is processed using PCA to obtain the principal direction vector.

[0102]

[0103] Direction vector This indicates the direction of blood vessel extension in the image. The center point is c = (x...). c ,y c The two endpoints of the direction line can be defined:

[0104]

[0105] Where L is the length parameter of the direction line, used for visualization.

[0106] Through the above processing steps, the optimal position p for hemodynamic parameter measurement is finally obtained. * And the optimal angle and direction lines (pt1, pt2), which can be marked and visualized in the image.

[0107] Figure 4 The diagram shown is a visualization of the optimal measurement location and direction in one embodiment of the present invention. Different colored areas represent segmented cerebral vascular structures, gray dots mark the optimal measurement locations automatically generated by skeletal structure analysis and thickness scoring algorithms, and dashed lines represent the recommended pulse wave (PW) measurement directions to guide the acquisition of hemodynamic parameters.

[0108] This invention provides a method for cerebral vascular segmentation and hemodynamic measurement based on transcranial color Doppler ultrasound. Through model training, image reasoning, skeleton extraction, and measurement optimization, it achieves automatic identification of cerebral vascular structures and intelligent recommendation of optimal measurement positions and angles. This method can effectively improve the accuracy and consistency of blood flow parameter acquisition, reduce human error, and help enhance the clinical application value of transcranial Doppler ultrasound in cerebral vascular assessment.

[0109] Accordingly, the present invention also provides a transcranial color Doppler (TCD) cerebral hemodynamic measurement device for implementing the TCD cerebral hemodynamic measurement method described above, comprising: a segmentation model establishment module for training multiple sample images to establish a multi-target cerebral blood vessel segmentation model; an image segmentation module for inputting the image to be segmented into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented; a skeleton structure extraction module for performing a midline transformation on the segmentation region of each cerebral blood vessel to extract the skeleton structure of each cerebral blood vessel; a parameter calculation module for calculating the optimal measurement position and optimal measurement angle for measuring cerebral hemodynamic parameters based on the skeleton structure of each cerebral blood vessel; a marking module for marking the image containing the skeleton structure of cerebral blood vessels based on the optimal measurement position and optimal measurement angle; and a display module for displaying in real time the segmentation region of each cerebral blood vessel in the image to be segmented, the skeleton structure of each cerebral blood vessel, and the optimal measurement position and optimal measurement angle of the marking.

[0110] Specifically, the segmentation model building module includes: a sample image segmentation unit, used to segment multiple cerebral blood vessels in each sample image to obtain labeled sample data containing the segmented region of each cerebral blood vessel, wherein the sample images include multiple sets of transcranial color Doppler cerebral blood vessel image data; a data construction unit, used to construct a training dataset based on the labeled sample data and the corresponding sample images; and a model construction unit, used to construct the multi-target cerebral blood vessel segmentation model based on the training dataset.

[0111] Specifically, the image segmentation module includes: a segmentation model storage unit for storing multi-target cerebral blood vessel segmentation models, which can be built based on a deep learning architecture and support the identification of multiple cerebral blood vessel structures, including the middle cerebral artery, anterior cerebral artery, posterior cerebral artery, and communicating artery; and a segmentation processing unit for receiving transcranial color Doppler ultrasound image data acquired in real time through a video capture card (such as HDM, DP, DVI, etc.), inputting it into the segmentation model for inference, and outputting the segmented regions of each cerebral blood vessel. The segmentation results are stored in NumPy array format and contour markings are applied to the original image for subsequent processing.

[0112] Specifically, the skeleton structure extraction module is used to perform a midline transformation on the cerebral blood vessel region output by the segmentation processing unit, and extract the vascular skeleton structure using a morphological thinning algorithm or a distance transformation method to generate a one-pixel wide centerline image.

[0113] Specifically, the parameter calculation module performs distance transformation in the skeleton image, estimates the vessel thickness at each skeleton point, and selects the optimal measurement location using a scoring function. Subsequently, principal component analysis (PCA) is used to fit the skeleton points and extract the principal axis direction of the vessels as the optimal measurement angle. The final results are marked in the image as gray dots and dashed lines to guide the acquisition of hemodynamic parameters.

[0114] The device provided in this invention can be integrated into a transcranial color Doppler ultrasound system or deployed as an independent image processing module on an edge computing device or server. It has efficient and automated cerebrovascular analysis capabilities, significantly improving the efficiency and consistency of clinical diagnosis.

[0115] This invention also provides a cerebrovascular analysis device based on ultrasound data and a large language model, which may include:

[0116] Memory, used to store computer programs;

[0117] When a processor executes a computer program stored in the aforementioned memory, it can perform the following steps:

[0118] Multiple sample images are trained to establish a multi-target cerebral blood vessel segmentation model; the image to be segmented is input into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented; the segmentation region of each cerebral blood vessel is transformed along the midline to extract the skeleton structure of each cerebral blood vessel; and based on the skeleton structure of each cerebral blood vessel, the optimal measurement position and the optimal measurement angle for measuring cerebral blood vessel hemodynamic parameters are calculated.

[0119] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps;

[0120] Multiple sample images are trained to establish a multi-target cerebral blood vessel segmentation model; the image to be segmented is input into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented; the segmentation region of each cerebral blood vessel is transformed along the midline to extract the skeleton structure of each cerebral blood vessel; and based on the skeleton structure of each cerebral blood vessel, the optimal measurement position and the optimal measurement angle for measuring cerebral blood vessel hemodynamic parameters are calculated.

[0121] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM) > random access memory (RAM), magnetic disks, or optical disks.

[0122] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0123] Similarly, it should be understood that, in order to simplify this disclosure and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this method of disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0124] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0125] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0126] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0127] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A method for measuring cerebral hemodynamics based on transcranial color Doppler ultrasound, characterized in that, include: Training multiple sample images to establish a multi-target cerebral blood vessel segmentation model; The image to be segmented is input into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented; A midline transformation is performed on the segmented region of each cerebral blood vessel to extract the skeletal structure of each cerebral blood vessel; as well as Based on the skeletal structure of each cerebral blood vessel, the optimal measurement location and angle for measuring cerebral blood vessel hemodynamic parameters are calculated.

2. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler as described in claim 1, characterized in that, The steps for training multiple sample images to build a multi-target cerebral blood vessel segmentation model include: Multiple cerebral blood vessels in each sample image are segmented to obtain labeled sample data containing the segmented regions of each cerebral blood vessel. The sample images include multiple sets of transcranial color Doppler cerebral vascular image data. A training dataset is constructed based on the labeled sample data and the corresponding sample images; Based on the training dataset, the multi-target cerebral blood vessel segmentation model is constructed.

3. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler ultrasound according to claim 2, characterized in that, The steps for region segmentation of multiple cerebral blood vessels in each sample image include: The multiple sample images are divided into a template image group and a training image group, and the training image group is calibrated using the template image group; Extract segmented regions of multiple cerebral blood vessels from each sample image in the calibrated training image set.

4. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler ultrasound according to claim 3, characterized in that, In the step of extracting segmented regions of multiple cerebral blood vessels in each sample image of the calibrated training image set, the segmented regions of multiple cerebral blood vessels are extracted by conventional image processing methods by utilizing the differences in color or density between the blood vessel region and the surrounding tissue.

5. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler ultrasound according to claim 3, characterized in that, In the step of extracting segmented regions of multiple cerebral blood vessels in each sample image of the calibrated training image set, the segmented regions of multiple cerebral blood vessels in each sample image are obtained by hand drawing.

6. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler as described in claim 1, characterized in that, The steps of performing a midline transformation on the segmented regions of each cerebral blood vessel and extracting the skeletal structure of each cerebral blood vessel include: The segmented regions of each cerebral blood vessel are binarized to obtain a foreground mask image; A skeleton extraction algorithm is used to perform a mid-axis transformation on the foreground mask image to obtain a skeleton image including the skeletal structure of cerebral blood vessels.

7. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler as described in claim 6, characterized in that, In the step of performing a mid-axis transformation on the foreground mask image using a skeleton extraction algorithm to obtain a skeleton image of the skeleton structure including cerebral blood vessels, the skeleton image of the skeleton structure including cerebral blood vessels is obtained by progressively eroding the boundary pixels of the foreground mask image and preserving the endpoints.

8. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler as described in claim 6, characterized in that, In the step of performing a mid-axis transformation on the foreground mask image using a skeleton extraction algorithm to obtain a skeleton image of a skeleton structure including cerebral blood vessels, the skeleton image is obtained by calculating the distance map from each point in the foreground mask image to the boundary and extracting the curvature abrupt change points or slope discontinuities in the distance map as the skeleton.

9. The method for measuring cerebral vascular hemodynamics based on transcranial color Doppler as described in claim 1, characterized in that, Based on the skeletal structure of each cerebral blood vessel, the steps for calculating the optimal measurement location and angle for measuring cerebral hemodynamic parameters include: Multiple skeleton points are sampled in the skeleton structure; A distance transformation is performed on the skeleton points, and the distance from each point to the edge of the blood vessel is calculated as an estimate of the blood vessel thickness. Each skeleton point is scored, and the point where the estimated blood vessel thickness exceeds the first preset threshold and the local curvature is lower than the second preset threshold is taken as the best measurement location; Based on the local orientation of the neighborhood structure or skeleton at the optimal measurement location, the extension direction of the blood vessel at the optimal measurement location is estimated as the optimal measurement angle.

10. A transcranial color Doppler-based cerebral hemodynamic measurement device, used to implement the transcranial color Doppler-based cerebral hemodynamic measurement method according to any one of claims 1-9, characterized in that, include: The segmentation model building module is used to train multiple sample images to build a multi-target cerebral blood vessel segmentation model; The image segmentation module is used to input the image to be segmented into the multi-target cerebral blood vessel segmentation model to obtain the segmentation region of each cerebral blood vessel in the image to be segmented; The skeleton structure extraction module is used to perform a midline transformation on the segmented region of each cerebral blood vessel to extract the skeleton structure of each cerebral blood vessel. as well as The parameter calculation module is used to calculate the optimal measurement position and optimal measurement angle for measuring cerebral blood vessel hemodynamic parameters based on the skeletal structure of each cerebral blood vessel. A labeling module is used to label images containing skeletal structures of the brain based on the optimal measurement location and optimal measurement angle; The display module is used to display in real time the segmented region of each cerebral blood vessel in the image to be segmented, the skeletal structure of each cerebral blood vessel, and the optimal measurement position and angle of the marker.

Images