Risk behavior prediction method and system based on capillary image change

Abstract

The invention discloses a risk behavior prediction method and system based on microvascular image change, and relates to the technical field of image processing, and the method comprises the steps: collecting a plurality of frames of preprocessed microvascular images, and marking a microvascular region in the microvascular image to generate a basic template; comparing the basic template at the T-k moment with the basic template at the T moment to generate basic template change data; performing matching score calculation according to the basic template change data to obtain a template change score; comparing the template change score with a preset change score value to generate an image change tag, the image change tag including an abnormal image change tag and a normal image change tag; and inputting the template change score of the abnormal image change label into the established risk prediction model to obtain the risk behavior, the risk prediction model based on the regression network is constructed in combination with the abnormal image change label and the historical blood vessel image data, and the risk behavior can be predicted in advance.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for predicting risk behavior based on microvascular image changes. Background Art

[0002] As an important part of the human microcirculatory system, changes in the state of microvessels can reflect the health of the body. Changes in blood flow velocity, morphology and structure of microvessels usually indicate potential health problems, especially in the early stages of certain major diseases, such as cardiovascular disease and diabetes. Microvascular abnormalities are common early signals. Therefore, how to accurately and real-time monitor microvascular images to identify abnormal vascular behavior in advance has become a focus of current research.

[0003] At present, the microvascular image analysis technology in the existing technology still has some problems in capturing subtle changes. First, traditional image processing methods mostly rely on feature extraction of static images, lack of effective monitoring of dynamic changes, and it is difficult to provide reliable early warning of microvascular morphological changes or blood flow abnormalities; secondly, the existing methods are highly dependent on image quality, especially when imaging jitter, out-of-focus blur or noise interference occurs, and accuracy is often unable to be guaranteed, resulting in inaccurate detection of vascular abnormalities. Furthermore, the current analysis technology lacks effective historical data association and risk prediction capabilities, and fails to make full use of previous image information to judge future vascular health status, limiting the feasibility of early disease prevention. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for predicting risk behavior based on microvascular image changes.

[0005] A risk behavior prediction method based on microvascular image changes, comprising: Collecting multiple frames of preprocessed microvascular images, marking microvascular regions in the microvascular images to generate a basic template; Compare the basic template at time Tk with the basic template at time T to generate basic template change data; Calculate the matching score based on the basic template change data to obtain the template change score; Comparing the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; The template change scores of abnormal image change labels are input into the established risk prediction model to obtain risk behaviors.

[0006] Furthermore, the logic of annotating the microvascular region in the microvascular image to generate the basic template is as follows: The area containing microvessels in the microvascular image is divided into n segments, and the contours of the microvessels are marked by drawing polygons. The vertices of the drawn polygons constitute an edge point set, and the marked area containing microvessels is saved as a marking file as a marking result. A mapping between the marking file and the corresponding original image is established, and a basic template is generated.

[0007] Furthermore, the logic for comparing the basic template at time Tk with the basic template at time T is: The edge point sets of the microvascular basic template at time Tk and time T are extracted respectively. For the edge points in the edge point set, the distribution of their adjacent points in different directions and distances is calculated to generate a shape context histogram. The shape context histograms at time Tk and time T are compared, and the basic template change data between the two shape context histograms is calculated.

[0008] Furthermore, the generation logic of the shape context histogram at time Tk and time T is: Extract the edge point set of the microvascular basic template at Tk time = ; where a is the number of edge points of the microvascular basic template at time Tk; ( , ), calculate the point and all other edge points The polar diameter and angle , where 1≤i, j≤a; the polar diameter The calculation formula is: , For edge points and edge points The Euclidean distance between them; the angle β is calculated as: ,in, For edge points Relative to the edge point Angle; for each edge point , count the distribution of this point relative to all other edge points in different distance intervals and angle intervals as the shape context histogram at time Tk , for each edge point ,set up distance intervals, and set angle intervals, the corresponding shape context can be expressed as: ,in, is the zth distance interval and the The number of edge points within the angle interval, Represents the counting operation; similarly, the shape context histogram at time T is calculated .

[0009] Furthermore, the shape context histograms at time Tk and T are compared, and the basic template change data between the two shape context histograms is calculated. The logic is: By comparing the shape context histograms at time Tk and T, calculate the Tk and T moments The chi-square distance S of the shape context histogram is calculated as follows: S= , the chi-square distance of a edge points is output as the basic template change data.

[0010] Furthermore, the cost matrix C is constructed, and the chi-square distances in the basic change data are arranged in a Q×P matrix format, where the product of Q and P is equal to a. For each row of the cost matrix C, the mean is calculated to obtain the chi-square mean, and the chi-square means of each row are accumulated and calculated, and the final sum is output as the template change score.

[0011] Furthermore, the generation logic of the image change label is: Compare the template change score with the change score label value. If the template change score is greater than or equal to the change score label value, generate an abnormal image change label. The template change score is compared with the change score label value. If the template change score is less than the change score label value, a normal image change label is generated.

[0012] Furthermore, the steps for constructing the established risk prediction model are: Acquire historical vascular image detection training data, and divide the historical vascular image detection training data into an image detection training set and an image detection test set; the historical vascular image detection training data includes a template change score of an abnormal image change label and its corresponding predicted risk behavior; Construct a regression network, use the template change scores of abnormal image change labels in the image detection training set as the regression network input data, and use the estimated risk behavior in the image detection training set as the regression network output data, train the regression network, and obtain an initial machine learning network; The initial machine learning network is validated using the image detection test set, and the output of the initial machine learning network that is less than or equal to the preset test error threshold is used as the pre-built risk prediction model.

[0013] Furthermore, the risky behaviors include machine-side risks and vascular physiological risks.

[0014] A risk behavior prediction system based on microvascular image changes, the system comprising: Template generation module: used to collect multiple frames of preprocessed microvascular images, mark the microvascular areas in the microvascular images, and generate basic templates; Template comparison module: used to compare the basic template at time Tk with the basic template at time T to generate basic template change data; Change scoring module: used to calculate the matching score based on the basic template change data to obtain the template change score; Label generation module: used to compare the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; Risk prediction module: used to input the template change score of the abnormal image change label into the established risk prediction model to obtain risk behavior.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes accurate comparison of microvascular images and extraction of template change data through calculation of shape context histogram and chi-square distance, thereby effectively identifying subtle changes in microvascular images, and automatically generates abnormal or normal image change labels by calculating template change scores through cost matrix and comparing them with preset values, and preliminarily determines whether there are abnormalities in the microvascular image during the change process, thereby improving the depth of microvascular image analysis; in addition, the present invention combines abnormal image change labels with historical vascular image data to construct a risk prediction model based on regression network, which can predict machine-side risks and vascular physiological risks in advance, thereby providing early warning signals, and improving the clinical ability to monitor abnormalities of microvascular images and prevent machine-side failure risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 A flowchart of a risk behavior prediction method based on microvascular image changes provided in Example 1 of the present invention; Figure 2 A schematic diagram of a basic template of a risk behavior prediction method based on microvascular image changes provided in Example 1 of the present invention; Figure 3 This is a module diagram of a risk behavior prediction method based on microvascular image changes provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Example 1 See also Figure 1 As shown, this embodiment discloses a risk behavior prediction method based on microvascular image changes, the method comprising: S110: collecting multiple frames of microvascular images after preprocessing, and marking microvascular regions in the microvascular images to generate a basic template; It should be noted that the preprocessing includes but is not limited to image denoising, contrast and brightness adjustment, grayscale, edge detail enhancement, artifact removal, and image size standardization: Specifically, the logic of annotating the microvascular area in the microvascular image to generate the basic template is: like Figure 2 As shown, the area containing microvessels in the microvessel image is divided into n segments, and the contours of the microvessels are marked by drawing polygons, the vertices of the drawn polygons constitute an edge point set, and the marked area containing microvessels is saved as a marking file as a marking result, a mapping between the marking file and the corresponding original image is established, and a basic template is generated; It should be noted that: the image annotation software includes but is not limited to LabelImg, ImageJ and CVAT, and the template generation software includes but is not limited to matlab, ImageJ and ITK-SNAP.

[0020] S120: Compare the basic template at time Tk with the basic template at time T to generate basic template change data; Specifically, the logic for comparing the basic template at time Tk with the basic template at time T is: Extract the edge point sets of the microvascular basic template at time Tk and time T respectively, calculate the distribution of its adjacent points in different directions and distances for the edge points in the edge point set, generate a shape context histogram, compare the shape context histograms at time Tk and time T, and calculate the basic template change data between the two shape context histograms; It should be noted that: the context of each edge point can be represented by a 2D polar coordinate histogram, which describes the spatial distribution of the point relative to other points; Specifically, the generation logic of the shape context histogram at time Tk and time T is: Extract the edge point set of the microvascular basic template at Tk time = ; where a is the number of edge points of the microvascular basic template at time Tk; ( , ), calculate the point and all other edge points The polar diameter and angle , where 1≤i, j≤a; the polar diameter The calculation formula is: , For edge points and edge points The Euclidean distance between them; the angle β is calculated as: ,in, For edge points Relative to the edge point Angle; for each edge point , count the distribution of this point relative to all other edge points in different distance intervals and angle intervals as the shape context histogram at time Tk , for each edge point ,set up distance intervals, and set angle intervals, the corresponding shape context can be expressed as: ,in, is the zth distance interval and the The number of edge points within the angle interval, Represents the counting operation; similarly, the shape context histogram at time T is calculated ; Specifically, the shape context histograms at time Tk and T are compared, and the basic template change data between the two shape context histograms is calculated. The logic is: By comparing the shape context histograms at time Tk and T, calculate the Tk and T moments The chi-square distance S of the shape context histogram is calculated as follows: S= , the chi-square distance of a edge points is output as the basic template change data; It should be noted that: the smaller the chi-square distance is, the more similar the two shape context histograms are, and the lower the probability of microvascular image offset or jitter; S130: Calculate a matching score according to the basic template change data to obtain a template change score; Specifically, a cost matrix C is constructed, and the chi-square distances in the basic change data are arranged in a Q×P matrix format, where the product of Q and P is equal to a. For each row of the cost matrix C, the mean is calculated to obtain the chi-square mean, and the chi-square means of each row are accumulated and calculated, and the final sum is output as the template change score.

[0021] S140: Compare the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; Specifically, the generation logic of image change labels is: Compare the template change score with the change score label value. If the template change score is greater than or equal to the change score label value, generate an abnormal image change label. Compare the template change score with the change score label value. If the template change score is less than the change score label value, generate a normal image change label. S150: inputting the template change score of the abnormal image change label into a predetermined risk prediction model to obtain risk behavior; Specifically, the steps for constructing the established risk prediction model are: Acquire historical vascular image detection training data, and divide the historical vascular image detection training data into an image detection training set and an image detection test set; the historical vascular image detection training data includes a template change score of an abnormal image change label and its corresponding predicted risk behavior; Specifically, the risky behaviors include machine-side risks and vascular physiological risks; It is important to understand that the machine-side risks include but are not limited to imaging jitter, out-of-focus blur and noise interference, and the vascular physiological risks include but are not limited to vascular dilation or contraction, microvascular blockage or rupture and morphological changes or bifurcation abnormalities; It should be noted that: the historical vascular image detection training data is obtained by manually recording the template change score when the abnormal image change label appears, and the machine-side risk and / or vascular physiological risk that appears within a period of time after the abnormal image change label is recorded; Construct a regression network, use the template change scores of abnormal image change labels in the image detection training set as the regression network input data, and use the estimated risk behavior in the image detection training set as the regression network output data, train the regression network, and obtain an initial machine learning network; The initial machine learning network is validated using the image detection test set, and the initial machine learning network with a test error less than or equal to a preset threshold is output as a pre-built risk prediction model; It should be noted that the regression network is one of support vector machine regression, decision tree regression or random forest regression.

[0022] Example 2 See also Figure 3 As shown, based on the unified inventive concept, this embodiment discloses a risk behavior prediction system based on microvascular image changes, which includes a template generation module, a template comparison module, a change scoring module, a label generation module and a risk prediction module.

[0023] The template generation module S210 is used to collect multiple frames of microvascular images after preprocessing, and mark the microvascular regions in the microvascular images to generate a basic template; Specifically, the logic of annotating the microvascular area in the microvascular image to generate the basic template is: The area containing microvessels in the microvascular image is divided into n segments, and the outline of the microvessels is marked by drawing polygons. The marked area containing microvessels is saved as a marking file as the marking result. The marking file needs to be associated with the corresponding original image. The marked area containing microvessels is imported into the analysis software to generate a basic template.

[0024] Template comparison module S220: used to compare the basic template at time Tk with the basic template at time T to generate basic template change data; Specifically, the logic for comparing the basic template at time Tk with the basic template at time T is: Extract the edge point sets of the microvascular basic template at time Tk and time T respectively, calculate the distribution of its adjacent points in different directions and distances for the edge points in the edge point set, generate a shape context histogram, compare the shape context histograms at time Tk and time T, and calculate the basic template change data between the two shape context histograms; Specifically, the generation logic of the shape context histogram at time Tk and time T is: Extract the edge point set of the microvascular basic template at Tk time = ; where a is the number of edge points of the microvascular basic template at time Tk; ( , ), calculate the point and all other edge points The polar diameter and angle , where 1≤i, j≤a; the polar diameter The calculation formula is: , For edge points and edge points The Euclidean distance between them; the angle β is calculated as: ,in, For edge points Relative to the edge point Angle; for each edge point , count the distribution of this point relative to all other edge points in different distance intervals and angle intervals as the shape context histogram at time Tk , for each edge point ,set up distance intervals, and set angle intervals, the corresponding shape context can be expressed as: ,in, is the zth distance interval and the The number of edge points within the angle interval, Represents the counting operation; similarly, the shape context histogram at time T is calculated ; Specifically, the shape context histograms at time Tk and T are compared, and the basic template change data between the two shape context histograms is calculated. The logic is: By comparing the shape context histograms at time Tk and T, calculate the Tk and T moments The chi-square distance S of the shape context histogram is calculated as follows: S= , the chi-square distance of a edge points is output as the basic template change data; Change scoring module S230: used to calculate the matching score according to the basic template change data to obtain the template change score; Specifically, a cost matrix C is constructed, and the chi-square distances in the basic change data are arranged in a Q×P matrix format, where the product of Q and P is equal to a. For each row of the cost matrix C, the mean is calculated to obtain the chi-square mean, and the chi-square means of each row are accumulated and calculated, and the final sum is output as the template change score.

[0025] The label generation module S240 is used to compare the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; Specifically, the generation logic of image change labels is: Compare the template change score with the change score label value. If the template change score is greater than or equal to the change score label value, generate an abnormal image change label. Compare the template change score with the change score label value. If the template change score is less than the change score label value, generate a normal image change label. Risk prediction module S250: used to input the template change score of the abnormal image change label into the established risk prediction model to obtain risk behavior; Specifically, the steps for constructing the established risk prediction model are: Acquire historical vascular image detection training data, and divide the historical vascular image detection training data into an image detection training set and an image detection test set; the historical vascular image detection training data includes a template change score of an abnormal image change label and its corresponding predicted risk behavior; Specifically, the risky behaviors include machine-side risks and vascular physiological risks; Construct a regression network, use the template change scores of abnormal image change labels in the image detection training set as the regression network input data, and use the estimated risk behavior in the image detection training set as the regression network output data, train the regression network, and obtain an initial machine learning network; The initial machine learning network is validated using the image detection test set, and the initial machine learning network with a test error less than or equal to a preset threshold is output as a pre-built risk prediction model; The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0026] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0027] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0028] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a risk behavior prediction method based on microvascular image changes. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0029] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0030] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0031] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0032] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A risk behavior prediction method based on microvascular image changes, characterized in that: The method comprises: Collecting multiple frames of preprocessed microvascular images, marking microvascular regions in the microvascular images to generate a basic template; Compare the basic template at time Tk with the basic template at time T to generate basic template change data; Calculate the matching score based on the basic template change data to obtain the template change score; Comparing the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; The template change scores of abnormal image change labels are input into the established risk prediction model to obtain risk behaviors.

2. The risk behavior prediction method based on microvascular image changes according to claim 1, characterized in that: The logic of annotating the microvascular area in the microvascular image to generate the basic template is: The area containing microvessels in the microvascular image is divided into n segments, and the contours of the microvessels are marked by drawing polygons. The vertices of the drawn polygons constitute an edge point set, and the marked area containing microvessels is saved as a marking file as a marking result. A mapping between the marking file and the corresponding original image is established, and a basic template is generated.

3. The risk behavior prediction method based on microvascular image changes according to claim 2 is characterized in that: The logic for comparing the basic template at time Tk with the basic template at time T is: The edge point sets of the microvascular basic template at time Tk and time T are extracted respectively. For the edge points in the edge point set, the distribution of their adjacent points in different directions and distances is calculated to generate a shape context histogram. The shape context histograms at time Tk and time T are compared, and the basic template change data between the two shape context histograms is calculated.

4. The risk behavior prediction method based on microvascular image changes according to claim 3 is characterized in that: The generation logic of the shape context histogram at time Tk and time T is: Extract the edge point set of the microvascular basic template at Tk time = ; where a is the number of edge points of the microvascular basic template at time Tk; ( , ), calculate the point and all other edge points The polar diameter and angle , where 1≤i, j≤a; the polar diameter The calculation formula is: , For edge points and edge points The Euclidean distance between them; the angle β is calculated as: ,in, For edge points Relative to the edge point Angle; for each edge point , count the distribution of this point relative to all other edge points in different distance intervals and angle intervals as the shape context histogram at time Tk , for each edge point ,set up distance intervals, and set angle intervals, the corresponding shape context can be expressed as: ,in, is the zth distance interval and the The number of edge points within the angle interval, Represents the counting operation; similarly, the shape context histogram at time T is calculated .

5. The risk behavior prediction method based on microvascular image changes according to claim 4, characterized in that: Compare the shape context histograms at time Tk and T, and calculate the basic template change data between the two shape context histograms. The logic is: By comparing the shape context histograms at time Tk and T, calculate the Tk and T moments The chi-square distance S of the shape context histogram is calculated as follows: S= , the chi-square distance of a edge points is output as the basic template change data.

6. The risk behavior prediction method based on microvascular image changes according to claim 5, characterized in that: Construct the cost matrix C, arrange the chi-square distances in the basic change data in a Q×P matrix format, where the product of Q and P is equal to a, and for each row of the cost matrix C, calculate the mean to obtain the chi-square mean, add up the chi-square means of each row, and output the final sum as the template change score.

7. The method for predicting risk behavior based on microvascular image changes according to claim 6, characterized in that: The generation logic of image change labels is: Compare the template change score with the change score label value. If the template change score is greater than or equal to the change score label value, generate an abnormal image change label. The template change score is compared with the change score label value. If the template change score is less than the change score label value, a normal image change label is generated.

8. The method for predicting risk behavior based on microvascular image changes according to claim 7, characterized in that: The steps for constructing the established risk prediction model are: Acquire historical vascular image detection training data, and divide the historical vascular image detection training data into an image detection training set and an image detection test set; the historical vascular image detection training data includes a template change score of an abnormal image change label and its corresponding predicted risk behavior; Construct a regression network, use the template change scores of abnormal image change labels in the image detection training set as the regression network input data, and use the estimated risk behavior in the image detection training set as the regression network output data, train the regression network, and obtain an initial machine learning network; The initial machine learning network is validated using the image detection test set, and the output of the initial machine learning network that is less than or equal to the preset test error threshold is used as the pre-built risk prediction model.

9. The method for predicting risk behavior based on microvascular image changes according to claim 8, characterized in that: The risky behaviors include machine-side risks and vascular physiological risks.

10. A risk behavior prediction system based on microvascular image changes, used to execute any one of the risk prediction methods based on microvascular image changes of claims 1-9, characterized in that: The system comprises: Template generation module: used to collect multiple frames of preprocessed microvascular images, mark the microvascular areas in the microvascular images, and generate basic templates; Template comparison module: used to compare the basic template at time Tk with the basic template at time T to generate basic template change data; Change scoring module: used to calculate the matching score based on the basic template change data to obtain the template change score; Label generation module: used to compare the template change score with a preset change score value to generate an image change label, wherein the image change label includes an abnormal image change label and a normal image change label; Risk prediction module: used to input the template change score of the abnormal image change label into the established risk prediction model to obtain risk behavior.

Images