Blood vessel stenosis rate prediction method, prediction model and prediction system

Abstract

The invention provides a blood vessel stenosis rate prediction method, model and system, and the method comprises the steps: obtaining an angiography image, carrying out the blood vessel segmentation processing of the angiography image, and obtaining a blood vessel segmentation map; then the blood vessel segmentation map is used as input of a pre-trained deep learning-based blood vessel stenosis rate prediction model, a blood vessel stenosis rate distribution map is output after prediction is carried out through the prediction model, and the blood vessel stenosis rate distribution map comprises a plurality of first pixels. According to the prediction model, method and system, the deep learning technology is combined with clinical medicine, the blood vessel stenosis condition is accurately calculated through the deep learning technology, the method and system can adapt to medical images of different qualities, the prediction speed is high, the prediction result is accurate and consistent, the diagnosis specification and standard of doctors are kept, and the diagnosis efficiency is improved. And the defect of poor interpretability of deep learning is overcome, and the method is a high-accuracy blood vessel stenosis rate prediction method.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a vascular stenosis rate prediction method and a vascular stenosis rate prediction system, and a vascular stenosis rate prediction model based on deep learning that the method and the system rely on. Background Art

[0002] Magnetic resonance angiography (MRA) and computed tomography angiography (CTA) have been widely used in vascular stenosis identification technology, and the existing technology has provided a variety of related technologies. Automatic identification of vascular stenosis in MRA / CTA images and evaluation of the degree of stenosis provide clinicians with computer-aided diagnosis methods, which can improve the diagnostic accuracy of intracranial and coronary artery stenosis and reduce the workload of doctors.

[0003] The mainstream technical methods for identifying vascular stenosis in the prior art include manual identification, centerline-based method and vascular tracking method. The manual identification method is a method in which a clinician determines vascular stenosis by measuring and identifying MRA and / or CTA images through tools. The method for identifying vascular stenosis based on the centerline method, such as the method proposed in the document "Automatic Identification Method of Vascular Stenosis Based on Vascular Segmentation of Coronary Angiography Images", includes two parts: vascular segmentation and vascular stenosis identification. In the vascular segmentation part, the image is first enhanced by an improved model based on Frangi Hessian, and then the vascular region is segmented using a statistical region fusion method. In the vascular stenosis identification part, the level set algorithm is first used to refine the segmentation result to obtain the vascular skeleton, and then the vascular edge is extracted to measure the vascular diameter. Finally, the local minimum point method is used to calculate the percentage of vascular segment stenosis in the entire image, and the stenosis segment is located and graded. Another example based on the centerline method is a vascular centerline extraction framework proposed in the document "Deep Centerline: A Multi-task Fully Convolutional Network for Centerline Extraction", which combines a deep learning fully convolutional network FCN with a minimum path extractor. FCN can simultaneously predict the minimum distance map of the blood vessels and detect all the blood vessel endpoints. The minimum distance map is the minimum distance mapping from the point on the blood vessel to the blood vessel boundary, and the point on the line should be the local maximum point. Then, the minimum path algorithm is used to extract the blood vessel centerline points and serialize them based on the data of the given starting root point, the list of blood vessel branch endpoints and the predicted blood vessel minimum distance map. Finally, the diameter and stenosis rate are calculated in sequence along the serialized centerline to determine the stenosis location and stenosis level. An example of the blood vessel tracking method is the method proposed in the literature "Coronary Artery Centerline Extraction in Cardiac CTAngiography Using a CNN-Based Orientation Classifier", which includes specifying a starting position on the original image or blood vessel segmentation map, intercepting a small area at this position, and predicting the blood vessel radius and blood vessel direction of the current area through a deep learning model. Therefore, it can move a certain distance along the direction (that is, reach the next position of the blood vessel), intercept and predict, and execute in a loop to achieve centerline tracking and blood vessel radius prediction and calculate stenosis rate until the blood vessel reaches the end.

[0004] A large number of practices have proved that the judgment of vascular stenosis by manual identification methods is affected by the knowledge, experience and subjective factors of the doctor in charge, as well as the measurement angle and tool errors, and different clinical physicians frequently make inconsistent judgments. Identification of vascular stenosis based on the centerline method may cause adhesions in vascular segmentation and errors in centerline extraction, which may result in the inability to accurately identify the location of vascular stenosis; also because the centerline extraction speed is slow, the centerline may not be centered due to the bulge of the vascular segmentation map, causing a large error in the calculation of the vascular stenosis rate, resulting in low recognition accuracy; also because the extracted centerline is disordered, the stenosis rate cannot be directly calculated, and thus vascular stenosis cannot be accurately identified. If the recognition rate is to be improved, the centerline serialization link needs to be added, which increases the calculation cost. The vascular tracking method requires the specification of the starting point and cannot be calculated in parallel, which results in the tracking on the original image being affected by the image quality, and the tracking on the vascular segmentation map being affected by vascular adhesion and abnormal bulges, which makes it impossible to accurately identify and judge vascular stenosis.

[0005] Based on the above defects of the prior art, a technical solution is proposed that can quickly and accurately determine and identify the stenosis rate of blood vessels. Summary of the invention

[0006] In order to solve the above-mentioned problems in the prior art, the embodiments of the present invention aim at the defects in the prior art and, based on obtaining Magnetic Resonance Angiography (MRA) and Computed Tomography Angiography (CTA), provide a vascular stenosis rate prediction method, a vascular stenosis rate prediction model based on the vascular stenosis rate, and a vascular stenosis rate prediction system.

[0007] The first aspect of the present invention provides a method for predicting vascular stenosis rate, comprising:

[0008] Acquiring angiographic images; wherein the angiographic images include magnetic resonance angiography and / or computed tomography angiography;

[0009] Performing blood vessel segmentation processing on the angiography image to obtain a blood vessel segmentation map;

[0010] The blood vessel segmentation map is used as the input of a pre-trained blood vessel stenosis rate prediction model, and a blood vessel stenosis rate distribution map is output; wherein the blood vessel stenosis rate distribution map includes a plurality of first pixels.

[0011] Preferably, the method for predicting vascular stenosis rate further includes:

[0012] A calculation process is performed according to the pixel value V of each first pixel in the vascular stenosis rate distribution map to generate a first vascular stenosis rate at a position corresponding to the first pixel; wherein V>0.

[0013] Further preferably, the calculation process according to the pixel value V of each first pixel in the vascular stenosis rate distribution map is specifically as follows:

[0014] According to a preset calculation rule, the pixel value V of each first pixel is calculated and processed to obtain the vascular stenosis rate at the position corresponding to the first pixel.

[0015] Preferably, the prediction calculation rule is to perform a square root operation on the pixel value V to obtain the vascular stenosis rate at the position corresponding to the first pixel.

[0016] Preferably, the method for predicting vascular stenosis rate further includes:

[0017] Detecting stenotic blood vessels and occluded blood vessels in a blood vessel stenosis rate distribution map according to pixel values ​​V of a plurality of first pixels;

[0018] According to the correspondence between the vascular stenosis rate distribution map and the vascular segmentation map or the angiography image, the stenotic blood vessels and the occluded blood vessels are matched to the vascular segmentation map and / or the angiography image to generate a visualized detection vascular segmentation map and / or a visualized detection angiography image;

[0019] The visualized detected blood vessel segmentation map and / or the visualized detected blood vessel angiography image are output to a display device.

[0020] Further preferably, the vascular stenosis rate prediction method further includes: classifying the stenotic blood vessels according to the degree of stenosis based on the pixel value V of the first pixel corresponding to the stenotic blood vessel and a preset vascular stenosis rate grading threshold; wherein the preset vascular stenosis rate grading threshold is multiple.

[0021] Further preferably, the number of the preset vascular stenosis rate classification thresholds is 3, and the vascular stenosis rate prediction method further includes:

[0022] Classifying the stenotic blood vessel into a mild stenotic blood vessel, a moderate stenotic blood vessel, and a severe stenotic blood vessel according to the pixel value V of the first pixel corresponding to the stenotic blood vessel and three preset vascular stenosis rate classification thresholds;

[0023] Different marking methods are selected to mark the mildly stenotic blood vessels, the moderately stenotic blood vessels and the severely stenotic blood vessels.

[0024] A second aspect of the present invention provides a deep learning-based vascular stenosis rate prediction model, including a vascular segmentation map and a first vascular stenosis rate distribution map matched therewith, and using a 3D U-net convolutional neural network model as a backbone network, and using a training set to train the 3D U-net convolutional neural network model to obtain a vascular stenosis rate prediction model. The specific method includes:

[0025] Data preprocessing process: using a data cutting method to perform sliding cutting processing on the vascular segmentation map and the first vascular stenosis rate distribution map matched therewith, to obtain a vascular segmentation sub-map and a vascular stenosis rate distribution sub-map with a model input size of N×N×N; wherein N>0, and N is an integer;

[0026] Encoding process: The vascular segmentation sub-map and the vascular stenosis rate distribution sub-map are used as inputs of the 3D U-net convolutional neural network model;

[0027] An M×M×M 3D convolution kernel is used for instance normalization, and a leaky rectified linear unit is used for correction. The step size is set to 2 to achieve downsampling, and the number of feature maps doubles with the number of downsampling times, and the network depth is set to 5. M>0, and M is an integer.

[0028] Decoding process: Upsampling is performed using the nearest neighbor upsampling method, which is cascaded at the same level as the encoding process, and the convolution results of the encoding process and the decoding process are spliced;

[0029] Output process: The model input size is restored to N×N×N, and a linear activation function is used to generate a stenosis rate prediction map that matches the vascular segmentation map.

[0030] Preferably, N=64, and M=3.

[0031] Preferably, the method for generating the first vascular stenosis rate distribution map includes:

[0032] Performing vascular centerline extraction processing on the vascular segmentation image to generate a vascular centerline image;

[0033] Performing optimized centerline processing on the vascular centerline image to generate an optimized centerline vascular image;

[0034] Performing midline serialization processing on the optimized midline vascular image to generate a midline serialized vascular image;

[0035] Performing simulated diameter-changing processing on the midline serialized blood vessel image to generate a simulated diameter-changing blood vessel image; wherein the simulated diameter-changing blood vessel image includes a plurality of first pixels;

[0036] Each pixel value of the first pixel is set to V to obtain a first blood vessel stenosis rate distribution map.

[0037] A third aspect of the present invention provides a vascular stenosis rate prediction system, which includes a data acquisition module for acquiring angiographic images;

[0038] A data processing module is used to perform blood vessel segmentation processing on the angiography image to obtain a blood vessel segmentation map;

[0039] A prediction module, comprising a pre-trained vascular stenosis rate prediction model, is used to use the vascular segmentation map as an input of the pre-trained vascular stenosis rate prediction model, and output a vascular stenosis rate distribution map; wherein the vascular stenosis rate distribution map comprises a plurality of first pixels;

[0040] The display module is used to receive the image data outputted thereto in any one of the vascular stenosis rate prediction methods of claims 1-7, and to display and output the image data.

[0041] A fourth aspect of the present invention provides a computer-readable storage medium having computer program instructions stored thereon, which implement the above method when the computer program instructions are executed by a processor.

[0042] The embodiments of the present invention provide a method and system for predicting vascular stenosis rate, and a vascular stenosis rate prediction model based on deep learning, which they rely on. It combines deep learning technology with clinical medicine, and efficiently and accurately calculates the stenosis of blood vessels through deep learning technology, and can be adapted to medical images of different qualities. The vascular stenosis rate prediction is performed using a vascular stenosis prediction model based on deep learning, which has a fast prediction speed, accurate and consistent prediction results, maintains the norms and standards of doctors' diagnosis, and solves the shortcomings of poor interpretability of deep learning. The vascular stenosis rate prediction method provided by the present invention performs predictions, which can provide a powerful reference for the diagnosis and treatment of vascular diseases and ischemic acute stroke through standardized and normalized vascular stenosis detection and positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 An exemplary CTA image of human brain tissue provided by the present invention;

[0044] Figure 2 An exemplary blood vessel segmentation map provided by the present invention;

[0045] Figure 3 A flow chart of a processing method for generating a first blood vessel stenosis rate distribution map according to a blood vessel segmentation map provided by the present invention;

[0046] Figure 4 A schematic diagram of a portion of an exemplary blood vessel centerline diagram provided by the present invention;

[0047] Figure 5 A schematic diagram of a portion of another exemplary blood vessel centerline diagram provided by the present invention;

[0048] Figure 6 This is an example diagram of the blood vessel centerline image processing provided by the present invention;

[0049] Figure 7 A schematic diagram of a portion of a first vascular stenosis rate distribution map provided by the present invention;

[0050] Figure 8 A schematic diagram of a simulated blood vessel provided by the present invention;

[0051] Fig. 9 A schematic diagram of a training process of a vascular stenosis rate prediction model provided by an embodiment of the present invention;

[0052] Fig.10 A flow chart of a method for predicting vascular stenosis rate provided by an embodiment of the present invention;

[0053] Fig.11 A flowchart of another method for predicting vascular stenosis rate provided by an embodiment of the present invention;

[0054] Fig.12 A flowchart of another method for predicting vascular stenosis rate provided by an embodiment of the present invention;

[0055] Fig.13 A flowchart of another method for predicting vascular stenosis rate provided by an embodiment of the present invention;

[0056] Fig.14 An overall flow chart of a training model for predicting vascular stenosis rate provided by an embodiment of the present invention;

[0057] Fig.15 An overall flow chart of a method for predicting a vascular stenosis rate using a vascular stenosis rate prediction model provided by the present invention is provided in an embodiment of the present invention;

[0058] Fig.16 A system block diagram of a vascular stenosis rate prediction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0060] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0061] In order to more clearly introduce a vascular stenosis rate prediction method and device provided by the present invention, here, a vascular stenosis rate prediction model based on deep learning provided by the present invention on which it relies is first introduced.

[0062] The vascular stenosis rate prediction model provided by the present invention adopts 3D U-NET as the backbone network and is pre-trained by the training set provided by the present invention. The training set of the vascular stenosis rate prediction model provided by the present invention includes: a vascular segmentation map and a first vascular stenosis rate distribution map matched therewith.

[0063] In one example of an embodiment of the present invention, Figure 1 An exemplary CTA image of human brain tissue provided by the present invention is shown in FIG. Figure 2 is an exemplary blood vessel segmentation map provided by the present invention. Figure 1 As shown in the figure, according to the image characteristics of MRA / CTA, the bright blocks in the brain tissue are the cross-sectional effects of the blood vessels at this level. According to the traditional image processing method or deep learning model, the blood vessels are segmented or extracted, and all non-vascular tissues and noise are removed to achieve the blood vessel segmentation map after 3D reconstruction of the blood vessels, as shown in the figure. Figure 2 As shown, the blood vessel indicated by 2200 is an intracranial blood vessel, and the blood vessels indicated by 2100 and 2300 are extracranial blood vessels. In the present invention, we generally refer to the process of obtaining a blood vessel segmentation map after processing MRA / CTA as blood vessel segmentation processing.

[0064] In clinical practice, physicians can Figure 2 The vascular segmentation diagram shown in the figure is used to diagnose vascular stenosis and occlusion. However, it is necessary to continuously rotate the image angle and observe the direction of the blood vessels in turn to identify whether there are stenotic or occluded blood vessels. The presence or absence of stenotic blood vessels and the degree of stenosis may be determined by the physician directly based on experience, or by using tools such as the ruler of the film reading software to measure and compare. Therefore, the above methods will introduce large errors due to subjective reasons or the measurement process, thereby affecting the calculation of vascular stenosis rate and the division of stenosis degree.

[0065] In an optional embodiment provided by the present invention, a large number of Figure 2 The blood vessel segmentation map shown in is used as a training set for training the blood vessel stenosis rate prediction model provided by the present invention. In order to train the blood vessel stenosis rate prediction model provided by the present invention, the blood vessel segmentation map needs to be processed to obtain a first blood vessel stenosis rate distribution map as a training set.

[0066] Figure 3 The present invention provides a flow chart of a processing method for generating a first vascular stenosis rate distribution map based on a vascular segmentation map. Figure 3 As shown, the step of processing the blood vessel segmentation map to obtain a first blood vessel stenosis rate distribution map includes:

[0067] Step 110, performing vascular centerline extraction processing on the vascular segmentation image to generate a vascular centerline image.

[0068] Specifically, a preset blood vessel centerline extraction method is used to extract the centerline of the blood vessel segmentation map to obtain a blood vessel centerline map. The preset blood vessel centerline extraction method can be selected from the blood vessel centerline extraction methods in the prior art. For example, in the embodiment of the present invention, a skeleton algorithm is selected as the preset blood vessel centerline extraction method.

[0069] In a specific example of an embodiment of the present invention, for example, the present invention provides Figure 2 By performing midline extraction processing on the vascular segmentation image shown in , a vascular midline image can be obtained. Figure 4 This is a schematic diagram of a portion of an exemplary blood vessel centerline map provided by the present invention. As shown in the figure, the centerline extracted from the blood vessel segmentation map obtained from the MRA / CTA image is essentially composed of points.

[0070] Step 120, performing centerline optimization processing on the vascular centerline image to generate an optimized centerline vascular image.

[0071] Specifically, in an ideal case, a portion of the blood vessel centerline image obtained by step 110 should be as follows: Figure 4 As shown in , they are neatly arranged. However, in actual operation, when performing the vascular centerline extraction process, high requirements are placed on the accuracy and stability of the vascular segmentation stage. Adhesions, bulges, etc. will seriously affect the effect of centerline extraction. In some complex locations, the centerline may be seriously misaligned or the centerline point may oscillate frequently. Therefore, only performing the vascular centerline extraction process in step 110 often cannot directly obtain the following. Figure 4 This effect is due to the limitations of the algorithm, which may cause the midline to oscillate, deviate (not centered), etc. Figure 5 Shown and Figure 6 Therefore, if we further use the vascular centerline image obtained by centerline extraction (such as Figure 5 As shown in , the vessel radius and stenosis rate calculated along the current centerline have deviations and cannot meet the requirements for training data.

[0072] Therefore, in order to improve the problem of midline oscillation in the vascular centerline map, the present invention uses a preset midline optimization processing method to optimize the midline processing of the vascular centerline map to obtain an optimized midline vascular map. In the embodiment of the present invention, the preset midline optimization processing method can be selected from the midline optimization processing method in the prior art. For example: Figure 6 As shown, for the vascular midline diagram, that is Figure 6 The original midline in is optimized using the preset midline optimization method to obtain the optimized midline vascular map, that is, Figure 6 The optimized midline in .

[0073] Step 130 , performing centerline serialization processing on the optimized centerline vascular image to generate a centerline serialized vascular image.

[0074] Specifically, a preset midline serialization method is used to perform midline serialization processing on the optimized midline vascular image to obtain a midline serialized vascular image. The preset midline serialization method is selected from the midline serialization methods in the prior art.

[0075] The purpose of centerline serialization is to provide a basis for calculating the vascular stenosis rate. This is because, in order to calculate the vascular stenosis rate, it is necessary to calculate the vascular radius at each point, and calculate the ratio of the vascular radius corresponding to all the centerline points in a certain distance before and after it, and compare it with its adjacent points. Therefore, centerline serialization is required to determine which centerline points are the points before and after the current point.

[0076] Step 140 , performing simulated diameter-changing processing on the centerline serialized vascular image to generate a simulated diameter-changing vascular image.

[0077] Specifically, the centerline serialized vascular image obtained after processing in step 120 and step 130 solves the problem of centerline oscillation and makes the centerline smooth, but it does not improve the degree of centerline deviation, and even aggravates the degree of centerline deviation in some local positions.

[0078] Therefore, the present invention uses a preset simulated diameter change method to simulate the diameter change of the centerline serialized vascular image, so as to obtain a simulated diameter change vascular image with improved deflection. The preset simulated diameter change method can be selected from the simulated diameter change methods in the prior art. For example, in the present invention, strict centering is used as the criterion to simulate the diameter change of the centerline serialized vascular image, so as to obtain a simulated diameter change vascular image. Figure 6 As shown in the rightmost schematic diagram, the simulated variable-diameter blood vessel image includes multiple first pixels.

[0079] Step 150, setting the pixel value of each first pixel in the simulated variable-diameter blood vessel map to V, and obtaining a first blood vessel stenosis rate distribution map.

[0080] Specifically, in the present invention, the center line of the simulated variable diameter blood vessel image processed by steps 110 to 140 is serialized, and the blood vessel stenosis rate at each center line point can be accurately calculated. The reference radius at the location is used for the blood vessel cross-sectional radius of each center line point in the simulated variable diameter blood vessel image, and the pixel value is 1, and then the pixel value of the pixel point on the cross section is set to V. Among them, V can be the stenosis rate, the square of the stenosis rate, or the cube of the stenosis rate.

[0081] In an optional solution of the present invention, the value of V is equal to the square of the stenosis rate, which can achieve the purpose of expanding the difference. Figure 7As shown, the pixel values ​​of the points on the cross section marked by the dotted line in the figure are all 0.9.

[0082] In the embodiment provided by the present invention, after the blood vessel segmentation map is processed in steps 110 to 150, a first blood vessel stenosis rate distribution map can be obtained.

[0083] Figure 8 A schematic diagram of a simulated blood vessel provided by the present invention. Figure 8 As shown, what is simulated is a first vascular stenosis rate distribution map. In the present invention, the vascular segmentation map can be obtained by performing image segmentation and 3D reconstruction on the MRA / CTA image. The vascular segmentation map is then processed through steps 110 to 150 to obtain a matching first vascular stenosis rate distribution map.

[0084] After the blood vessel segmentation map is processed by steps 110 to 150 provided by the embodiment of the present invention, the obtained data includes the blood vessel segmentation map and the first blood vessel stenosis rate distribution map matched therewith. A plurality of blood vessel segmentation maps and the first blood vessel stenosis rate distribution maps matched therewith can constitute the training set of the present invention for training the blood vessel stenosis rate prediction model provided by the present invention, and can be applied to the training of a blood vessel stenosis rate prediction model provided by the present invention.

[0085] The vascular stenosis rate model provided in the embodiment of the present invention is a special vascular stenosis rate prediction model obtained by training based on a deep learning model. The following specifically introduces a method for training the vascular stenosis rate prediction model using the training set provided by the present invention.

[0086] A vascular stenosis rate prediction model provided by an embodiment of the present invention adopts a 3D U-net convolutional neural network model as a backbone network, and the training set adopted is the vascular segmentation map introduced above and the first vascular stenosis rate distribution map matched therewith. Fig. 9 A schematic diagram of the training process of a vascular stenosis rate prediction model provided by an embodiment of the present invention is shown in FIG. Fig. 9 As shown, the specific process of training the deep learning-based vascular stenosis rate prediction model provided by the present invention includes:

[0087] Data preprocessing process S1: Use a data cutting method to perform sliding cutting processing on the vascular segmentation map and the first vascular stenosis rate distribution map that matches it, and obtain a vascular segmentation sub-map and a vascular stenosis rate distribution sub-map with a model input size of N×N×N; where N>0 and N is an integer.

[0088] Specifically, in the data preprocessing process of the embodiment provided by the present invention, since the first vascular stenosis rate distribution map in the training set is subjected to the simulated diameter change processing in step 140, the size of the first vascular stenosis rate distribution map is relatively large, so the vascular segmentation map and the first vascular stenosis rate distribution map matched therewith are slidably cut into vascular segmentation sub-maps and vascular stenosis rate distribution sub-maps with a model input size of N×N×N by a data cutting method for model training. Wherein, N>0, and N is an integer.

[0089] In an optional solution provided by an embodiment of the present invention, N=64 is taken, and the vascular segmentation map and the first vascular stenosis rate distribution map matching it are slidingly cut into vascular segmentation sub-maps and vascular stenosis rate distribution sub-maps with a model input size of 64×64×64 for model training.

[0090] Coding process S2: The vascular segmentation sub-image and the vascular stenosis rate distribution sub-image are used as the input of the 3D U-net convolutional neural network model. An M×M×M 3D convolution kernel is used for instance normalization, and a leaky rectified linear unit is used for correction.

[0091] Specifically, in the encoding process, the embodiment provided by the present invention selects a 3D U-net convolutional neural network model as the backbone network, and uses the vascular segmentation sub-graph and vascular stenosis rate distribution sub-graph of the model input size N×N×N obtained in the data preprocessing process as the input of the 3D U-net convolutional neural network model, and uses a 3D convolution kernel of M×M×M for strength normalization, and uses a leaky corrected linear unit for correction. Among them, the step size is set to 2 to achieve downsampling, and the number of feature maps doubles with the number of downsampling times, and the network depth is set to 5; M>0, and M is an integer. In the scheme provided by the embodiment of the present invention, the leaky corrected linear unit can be selected from the leaky corrected linear method in the prior art, such as the LeakyReLU correction function.

[0092] In a preferred solution of an embodiment of the present invention, M=3, that is, a 3D convolution kernel of 3×3×3 is used in the encoding stage, instance normalization is performed and a leaky rectified linear unit LeakyReLU is used for correction, the step size is set to 2, downsampling is achieved, and the number of feature maps doubles with the number of downsampling. The network depth is set to 5. In an embodiment of the present invention, since excessive depth will lead to an exponential increase in parameters, the requirement for computing power is too high. Therefore, in an embodiment of the present invention, for the selection of each parameter, the calculation speed and the amount of calculation will be fully considered, and the corresponding training set will be balanced and determined.

[0093] In the embodiments provided by the present invention, the core purpose of the instance normalization process is to adjust the pixel values ​​of the image data to a fixed range, usually from 0 to 1, to improve the stability and accuracy of the image processing algorithm. Through the instance normalization process, the difference in image intensity values ​​under different devices or different acquisition conditions can be eliminated, the model training speed can be accelerated, and the training stability of the model can be improved.

[0094] In an embodiment of the present invention, a method for performing instance normalization is selected from the prior art. In an optional example provided in an embodiment of the present invention, the core network used is a 3D U-net convolutional neural network model. The specific steps and implementation method for performing instance normalization include:

[0095] Generate Random 3D Image Data: Generate a random 3D image data using the NumPy library.

[0096] Define a normalization function: convert the 3D image data into float type and then normalize it to the range of [0,1].

[0097] Applying a normalization function: applying a normalization function to the generated 3D image data to obtain normalized image data.

[0098] Decoding process S3: The nearest neighbor upsampling method is used to perform the same-level cascade with the encoding process, and the convolution results of the encoding process and the decoding process are spliced.

[0099] Specifically, the upsampling in the decoding process provided by the embodiment of the present invention adopts the nearest neighbor upsampling method, and is cascaded at the same level with the encoding stage, and the convolution results of the two are spliced ​​together to perform regional property analysis at multiple resolutions.

[0100] The upsampling in the decoding process of the embodiment of the present invention adopts the nearest neighbor upsampling method and does not require training, which effectively reduces the training parameters of the network and has better robustness and speed.

[0101] Output process S4: The model input size is restored to N×N×N, and a linear activation function is used to generate a vascular stenosis rate prediction map that matches the vascular segmentation map.

[0102] Specifically, in the output process provided by the embodiment of the present invention, the output layer is restored to the model input size N×N×N, and the Linear activation function is used to generate the final vascular stenosis rate prediction map. Among them, N is selected to be the same size as in the data preprocessing step. In an optional solution of the present invention, in the data preprocessing step, N=64 is taken. Therefore, in the output process, the output layer is restored to the model input size 64×64×64.

[0103] The above describes a deep learning-based vascular stenosis rate prediction model provided by an embodiment of the invention. Next, based on the vascular stenosis rate prediction model provided by the invention, we will introduce in detail the vascular stenosis rate prediction method provided by the invention.

[0104] Fig.10 A flow chart of a method for predicting vascular stenosis rate provided by an embodiment of the present invention is shown in FIG. Fig.10 The method specifically comprises the following steps:

[0105] Step 210, obtaining angiographic images.

[0106] Specifically, the angiographic images acquired by the present invention include magnetic resonance angiography (MRA) and computed tomography angiography (CTA). The type of angiographic images acquired depends on the source of the image data.

[0107] Step 220, performing blood vessel segmentation processing on the angiography image to obtain a blood vessel segmentation map.

[0108] Specifically, a preset blood vessel segmentation processing technology is used to perform blood vessel segmentation processing on the MRA image or CTA image obtained in step 210 to obtain a blood vessel segmentation map. The preset blood vessel segmentation processing method can be a traditional image processing method or a deep learning model, which removes all non-vascular tissues and noises when implementing blood vessel segmentation or extraction, and implements a blood vessel segmentation map after 3D reconstruction of the blood vessels.

[0109] Step 230 , using the blood vessel segmentation map as an input to a pre-trained blood vessel stenosis rate prediction model, and outputting a blood vessel stenosis rate distribution map.

[0110] Specifically, the blood vessel segmentation map is input into a deep learning-based blood vessel stenosis rate prediction model provided by the present invention, and the output after being processed by the blood vessel stenosis rate prediction model is the blood vessel stenosis rate distribution map of the present invention, wherein the blood vessel stenosis rate distribution map includes a plurality of first pixels.

[0111] In a preferred embodiment of the present solution, the vascular stenosis rate distribution map is also output to a display device connected to the system for display, and / or the vascular stenosis rate distribution map is output to a printing device connected to the system for printing. The purpose of display output and / or printing is to facilitate doctors or other relevant personnel to view and analyze.

[0112] In an optional example of the embodiment of the present invention, Figure 7As shown, in a schematic diagram of a portion of a vascular stenosis rate distribution map provided by an embodiment of the present invention, each of a plurality of first pixels of the vascular stenosis rate distribution map has its own pixel value V, and the pixel value V can reflect the situation of the vascular stenosis rate. The value of V is greater than 0. When the first pixel value V is equal to 0, it represents that the position of the first pixel is the background, rather than being inside the blood vessel.

[0113] Fig.11 A flowchart of another method for predicting vascular stenosis rate provided by an embodiment of the present invention is shown in FIG. Fig.11 As shown, in a preferred solution of the embodiment of the present invention, the method for predicting vascular stenosis rate provided by the embodiment of the present invention further includes:

[0114] Step 240 , performing calculation processing according to the pixel value V of each first pixel in the vascular stenosis rate distribution map, to generate a first vascular stenosis rate at a position corresponding to the first pixel.

[0115] Specifically, when the vascular stenosis rate prediction model provided by the embodiment of the present invention is trained, when the first vascular stenosis rate distribution map in the training set is generated, in step 150, each pixel value of the first pixel is set to V, and V can be selected as the stenosis rate, the square of the stenosis rate, or the cube. Therefore, the preset calculation rule here should be the inverse operation of the assignment in step 150. For example, if the V assigned to the first pixel in step 150 is the square of the vascular stenosis rate, then the preset calculation rule here is square root, and a positive value is taken; if the V assigned to the first pixel in step 150 is the cube of the vascular stenosis rate, then the preset calculation rule here is cube root, and a positive value is taken.

[0116] In an optional solution provided by an embodiment of the present invention, the pixel value V of each first pixel is calculated and processed according to a preset calculation rule to obtain the vascular stenosis rate at the position corresponding to the first pixel.

[0117] In an optional solution of the embodiment of the present invention, the prediction calculation rule is to perform a square root operation on the pixel value V to obtain the vascular stenosis rate at the position corresponding to the first pixel.

[0118] In a preferred solution of the embodiment of the present invention, the vascular stenosis rate is output and displayed according to a preset output rule. For example, it is output and displayed in the form of a curve chart, a list, etc. according to a certain arrangement rule. And in a further solution, the values ​​in the table can correspond to each first pixel in the vascular stenosis rate distribution map, so that when the value in the selected list is reached, the corresponding first pixel is highlighted in the vascular stenosis rate distribution map.

[0119] Fig.12 A flowchart of another method for predicting vascular stenosis rate provided by an embodiment of the present invention is shown in FIG. Fig.12As shown, in a preferred embodiment provided by the present invention, the method for predicting vascular stenosis rate further includes the following steps:

[0120] Step 250: Detect stenotic blood vessels and occluded blood vessels in the blood vessel stenosis rate distribution map according to the pixel values ​​V of the plurality of first pixels.

[0121] Specifically, the specific method of the embodiment of the present invention is to determine whether the range of the V value is within a first preset range. For example, when the V value is greater than or equal to 0.5 and less than or equal to 0.9, the blood vessel corresponding to the first pixel point is determined to be a stenotic blood vessel. When the V value is greater than 0.9 and less than 1, the blood vessel corresponding to the first pixel point is determined to be an occluded blood vessel.

[0122] Step 260, according to the correspondence between the vascular stenosis rate distribution map and the vascular segmentation map or the angiography image, the stenotic vessels and occluded vessels are mapped to the vascular segmentation map and / or the angiography image to generate a visualized detection vascular segmentation map and / or a visualized detection angiography image.

[0123] Specifically, the vascular stenosis rate prediction model provided by the present invention has a vascular segmentation map as input, and an output is a vascular stenosis rate prediction model corresponding to the vascular segmentation map. The vascular segmentation map is obtained by processing the angiographic image through vascular segmentation, and the vascular segmentation map and the angiographic image also have a corresponding correspondence. Therefore, the stenotic vessels and occluded vessels in the vascular stenosis rate distribution map output by the present invention can be matched to the vascular segmentation map and / or the angiographic image according to the corresponding correspondence, thereby obtaining a visualized detection vascular segmentation map and / or a visualized detection angiographic image. In addition, the display of the stenotic vessels and occluded vessels is marked by using different display marking methods so that they can be observed intuitively.

[0124] Step 270: output the visualized detected blood vessel segmentation map and / or the visualized detected blood vessel angiography image to a display device.

[0125] Specifically, the visually detected blood vessel segmentation map and / or the visually detected blood vessel angiography image is displayed via a display device that is communicatively connected to the processor.

[0126] In an optional solution of the embodiment of the present invention, the visualized detection blood vessel segmentation map and / or the visualized detection blood vessel angiography image may also be printed by a printing device that is communicatively connected to the processor.

[0127] In a specific example of the embodiment of the present invention, stenosis and occlusion can be directly detected through a deep learning detection model, and a series of bounding boxes are predicted on the original image or the vascular segmentation map, and each bounding box represents a detected stenotic blood vessel or occluded blood vessel. This adds the original image with bounding boxes as the visualization detection angiography image that is the final output of the present invention, and the vascular segmentation map with added bounding boxes is the visualization detection vascular segmentation map that is the final output of the present invention.

[0128] The embodiment of the present invention also provides another method for predicting the vascular stenosis rate. This method can classify the degree of vascular stenosis according to the value of the cardiovascular stenosis rate. Fig.13 It is a flowchart of another method for predicting the vascular stenosis rate provided by the embodiment of the present invention. As Fig.13 shown, after step 250 is executed, the method further includes:

[0129] Step 2501, classifying the stenotic blood vessels according to the pixel value V of the first pixel corresponding to the stenotic blood vessel and a preset vascular stenosis rate classification threshold according to the degree of stenosis.

[0130] Specifically, in the solution provided by the embodiment of the present invention, the stenotic blood vessels are classified according to the pixel value V of the first pixel corresponding to the stenotic blood vessel and a preset vascular stenosis rate classification threshold. That is to say, it is judged according to the magnitude of the pixel value V of each first pixel in the vascular stenosis rate distribution map to determine the degree of vascular stenosis. For example, the stenotic blood vessels are divided into mildly stenotic blood vessels, moderately stenotic blood vessels, and severely stenotic blood vessels.

[0131] In an alternative solution provided by the embodiment of the present invention, the number of preset vascular stenosis rate classification thresholds is 3. For example: the first preset vascular stenosis rate classification threshold V1, the second preset vascular stenosis rate classification threshold V2, and the third preset vascular stenosis rate classification threshold V3. Among them, V1>0, V2>V1, and V3>V2.

[0132] Classifying the stenotic blood vessels according to the pixel value V of the first pixel corresponding to the stenotic blood vessel and a preset vascular stenosis rate classification threshold specifically includes judging the magnitude of the pixel value V of each first pixel to determine the degree of vascular stenosis. Such as:

[0133] When V<V1, it is determined that the stenotic blood vessel at the position corresponding to the first pixel is a mildly stenotic blood vessel.

[0134] When V≥V1 and V<V2, it is determined that the stenotic blood vessel at the position corresponding to the first pixel is a moderately stenotic blood vessel.

[0135] When V≥V2 and <V3, it is determined that the stenotic blood vessel at the position corresponding to the first pixel is a severely stenotic blood vessel.

[0136] In the embodiment provided by the present invention, V1, V2, and V3 are preset values ​​in the system. For example, the values ​​of V1, V2, and V3 are 30%, 50%, and 70%, respectively. In this way, the stenotic blood vessels detected in step 250 can be divided into mild stenotic blood vessels, moderate stenotic blood vessels, and severe stenotic blood vessels according to the size of the V value.

[0137] Furthermore, in the vascular stenosis rate prediction method provided by the embodiment of the present invention, the number of preset vascular stenosis rate classification thresholds can be increased, and the vascular stenosis rate can be divided into more levels according to the size of the pixel value V of each first pixel in the vascular stenosis rate distribution map. For example, four preset vascular stenosis rate classification thresholds are set: the first preset vascular stenosis rate classification threshold V1, the second preset vascular stenosis rate classification threshold V2, the third preset vascular stenosis rate classification threshold V3 and the fourth preset vascular stenosis rate classification threshold V4, and their values ​​are respectively set to 50%, 75%, 90%, and 95%. Therefore, the stenotic blood vessels can be divided into 4 different levels according to the degree of stenosis, and the detected stenotic blood vessels can be divided into mild stenotic blood vessels, moderate stenotic blood vessels, severe stenotic blood vessels and serious stenotic blood vessels. Through the above invention, the vascular stenosis level is divided according to the size of the pixel value V of each first pixel in the vascular stenosis rate distribution map.

[0138] In an embodiment of the present invention, after the detected stenotic blood vessels are graded according to the degree of stenosis, different marking methods are used to mark the blood vessels of different grades. Optional marking methods include marking the classified stenotic blood vessels and occluded blood vessels with different colors, which can more intuitively display the condition of the blood vessels for diagnostic reference. For example, use green lines to mark mildly stenotic blood vessels, yellow lines to mark moderately stenotic blood vessels, orange lines to mark severely stenotic blood vessels, purple lines to mark severely stenotic blood vessels, and dark red lines to mark occluded blood vessels. This example in the present invention is only to illustrate the use of different marks to represent blood vessels of different stenosis grades, and does not limit the marking method to only color lines. Other methods can also be used to mark blood vessels of different stenosis grades.

[0139] In a further example of the method for predicting the vascular stenosis rate provided by the embodiment of the present invention, the method further includes: according to the correspondence between the vascular stenosis rate distribution map and the vascular segmentation map or the angiography image, the mild stenosis blood vessels, moderate stenosis blood vessels, severe stenosis blood vessels and occluded blood vessels are corresponded to the vascular segmentation map or the angiography image, and a visualized stenosis graded blood vessel segmentation map and / or a visualized stenosis graded angiography image is generated, and then the visualized stenosis graded blood vessel segmentation map and / or the visualized stenosis graded angiography image is output to a display device. The principle of corresponding the mild stenosis blood vessels, moderate stenosis blood vessels, severe stenosis blood vessels and occluded blood vessels to the vascular segmentation map or the angiography image and the method of displaying and outputting are the same as those in step 260 and step 270, and will not be repeated here.

[0140] The vascular stenosis rate prediction model and vascular stenosis rate prediction method provided by the present invention have been introduced. Fig.14 The overall flow chart of a training model for predicting vascular stenosis rate provided by the embodiment of the present invention is shown in FIG. Fig.14 , which can help to more intuitively understand the training method of the vascular stenosis rate prediction model of the present invention. Fig.15 The overall flow chart of a method for predicting vascular stenosis rate using the vascular stenosis rate prediction model provided by the present invention is provided in an embodiment of the present invention, with reference to Fig.15 , it can be more intuitively understood that the vascular stenosis rate prediction model provided by the present invention is applied to implement a vascular stenosis rate prediction method provided by an embodiment of the present invention.

[0141] Next, a vascular stenosis rate prediction system provided by an embodiment of the present invention is described in detail. Fig.16 This is a system block diagram of a vascular stenosis rate prediction system provided by an embodiment of the present invention. As shown in the figure, the vascular stenosis rate device 1000 includes: a data acquisition module 1501, a data processing module 1502, a prediction module 1503, and a display module 1504.

[0142] In an embodiment provided by the present invention, the data acquisition module 1501, the data processing module 1502, the prediction module 1503 and the display module 1504 are connected to each other via wired or wireless communication methods and can transmit data to each other.

[0143] The data acquisition module 1501 is used to acquire angiographic images, wherein the angiographic images include magnetic resonance angiography and / or computed tomography angiography. The specific type of the acquired angiographic images depends on the data source.

[0144] The data processing module 1502 is used to segment or extract blood vessels based on the image characteristics of MRA / CTA and adopt traditional image processing methods or deep learning models, remove all non-vascular tissues and noise, and achieve vascular segmentation map after 3D reconstruction of blood vessels.

[0145] The prediction module 1503 is pre-installed with a pre-trained vascular stenosis rate prediction model, which is a vascular stenosis rate prediction model based on deep learning provided by an embodiment of the present invention. The prediction module is used to use the vascular segmentation map obtained after processing by the data processing module 1502 as the input of the pre-trained vascular stenosis rate prediction model, and directly output a vascular stenosis rate distribution map after the vascular stenosis rate prediction model is calculated. The vascular stenosis rate distribution map includes a plurality of first pixels.

[0146] The display module 1504 is used to receive all the image data, vascular stenosis rate data, and vascular marker display outputted therein in the vascular stenosis prediction method provided by the present invention, and to display and output the received data. The data received by the display module may come from one or more of the data acquisition module 1501, the data processing module 1502, and the prediction module 1503.

[0147] The embodiments of the present invention provide a method and system for predicting a vascular stenosis rate, and a vascular stenosis rate prediction model based on deep learning, which they rely on. They combine deep learning technology with clinical medicine, and efficiently and accurately calculate the stenosis of blood vessels through deep learning technology, and can be adapted to medical images of different qualities. The vascular stenosis rate prediction is performed using a vascular stenosis prediction model based on deep learning, which has a fast prediction speed, accurate and consistent prediction results, maintains the norms and standards of doctors' diagnosis, and solves the shortcomings of poor interpretability of deep learning. The vascular stenosis rate prediction method provided by the present invention performs predictions, which can provide a powerful reference for the diagnosis and treatment of vascular diseases and ischemic acute stroke through standardized and normalized vascular stenosis detection and positioning.

[0148] In some possible implementations, a vascular stenosis rate prediction method and a vascular stenosis rate prediction system provided by the present invention and a vascular stenosis rate prediction model based on deep learning on which it depends can be implemented by a processing component calling a computer-readable instruction stored in a memory. In one example, the processing component includes but is not limited to a separate processor, or a discrete component, or a combination of a processor and a discrete component. The processor may include a controller having an execution instruction function in an electronic device, and the processor may be implemented in any appropriate manner, for example, by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components. Inside the processor, executable instructions can be executed by hardware circuits such as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers.

[0149] The first and second in the following text are only for distinction and have no other meanings.

[0150] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0151] The disclosed embodiment also proposes a computer-readable storage medium, on which computer program instructions are stored, and the above method is implemented when the computer program instructions are executed by the processor. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. The computer-readable storage medium may be, for example, (but not limited to) an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a protruding structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be construed as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., light pulses through a fiber optic cable), or an electrical signal transmitted through wires.

[0152] Computer program instructions for performing operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages—such as Smalltalk, C++, etc., and conventional procedural programming languages—such as the "C" language or similar programming languages.

[0153] The computer-readable program instructions described herein can be downloaded to each computing / processing device from a computer-readable storage medium, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. The network adapter card or the network interface in each computing / processing device receives the computer-readable program instructions from the network, and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device. The computer-readable program instructions can be executed completely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer, partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer by any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect by the Internet). In some embodiments, by utilizing state information of computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0154] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0155] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0156] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0157] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0158] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK) and the like.

[0159] The various embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Without departing from the scope and spirit of the various embodiments described, many modifications and changes are obvious to those of ordinary skill in the art, and any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention. The selection of terms used in this article is intended to best explain the principles of the various embodiments, practical applications, or improvements to the technology in the market, or to enable other ordinary technicians in the technical field to understand the various embodiments disclosed herein.

Claims

1. A method for predicting vascular stenosis rate, characterized in that: Methods for predicting vascular stenosis rate include: Acquiring angiographic images; wherein the angiographic images include magnetic resonance angiography and / or computed tomography angiography; Performing blood vessel segmentation processing on the angiography image to obtain a blood vessel segmentation map; The blood vessel segmentation map is used as an input of a pre-trained blood vessel stenosis rate prediction model to output a blood vessel stenosis rate distribution map; wherein the blood vessel stenosis rate distribution map includes a plurality of first pixels.

2. The method for predicting vascular stenosis rate according to claim 1, characterized in that: The method for predicting vascular stenosis rate also includes: A calculation process is performed according to the pixel value V of each first pixel in the vascular stenosis rate distribution map to generate a first vascular stenosis rate at a position corresponding to the first pixel; wherein V>0.

3. The method for predicting vascular stenosis rate according to claim 2, characterized in that: The calculation process according to the pixel value V of each first pixel in the vascular stenosis rate distribution map is specifically as follows: According to a preset calculation rule, the pixel value V of each first pixel is calculated and processed to obtain the vascular stenosis rate at the position corresponding to the first pixel.

4. The method for predicting vascular stenosis rate according to claim 1, characterized in that: The prediction calculation rule is to perform a square root operation on the pixel value V to obtain the vascular stenosis rate at the position corresponding to the first pixel.

5. The method for predicting vascular stenosis rate according to claim 1, characterized in that: The method for predicting vascular stenosis rate also includes: Detecting stenotic blood vessels and occluded blood vessels in a blood vessel stenosis rate distribution map according to pixel values ​​V of a plurality of first pixels; According to the correspondence between the vascular stenosis rate distribution map and the vascular segmentation map or the angiography image, the stenotic blood vessels and the occluded blood vessels are matched to the vascular segmentation map and / or the angiography image to generate a visualized detection vascular segmentation map and / or a visualized detection angiography image; The visualized detected blood vessel segmentation map and / or the visualized detected blood vessel angiography image are output to a display device.

6. The method for predicting vascular stenosis rate according to claim 5, characterized in that: The vascular stenosis rate prediction method further includes: classifying the stenotic blood vessels according to the degree of stenosis based on the pixel value V of the first pixel corresponding to the stenotic blood vessel and a preset vascular stenosis rate classification threshold; wherein the preset vascular stenosis rate classification threshold is multiple.

7. The method for predicting vascular stenosis rate according to claim 6, characterized in that: The number of the preset vascular stenosis rate classification thresholds is 3, and the vascular stenosis rate prediction method further includes: Classifying the stenotic blood vessel into a mild stenotic blood vessel, a moderate stenotic blood vessel, and a severe stenotic blood vessel according to the pixel value V of the first pixel corresponding to the stenotic blood vessel and three preset vascular stenosis rate classification thresholds; Different marking methods are selected to mark the mildly stenotic blood vessels, the moderately stenotic blood vessels and the severely stenotic blood vessels.

8. A vascular stenosis rate prediction model, characterized in that: The training set of the vascular stenosis rate prediction model includes a vascular segmentation map and a first vascular stenosis rate distribution map matched therewith, and a 3D U-net convolutional neural network model is used as a backbone network. The specific method of using the training set to train the 3D U-net convolutional neural network model to obtain the vascular stenosis rate prediction model includes: Data preprocessing process: using a data cutting method to perform sliding cutting processing on the vascular segmentation map and the first vascular stenosis rate distribution map matched therewith, to obtain a vascular segmentation sub-map and a vascular stenosis rate distribution sub-map with a model input size of N×N×N; wherein N>0, and N is an integer; Encoding process: The vascular segmentation sub-map and the vascular stenosis rate distribution sub-map are used as inputs of the 3D U-net convolutional neural network model; An M×M×M 3D convolution kernel is used for instance normalization, and a leaky rectified linear unit is used for correction. The step size is set to 2 to achieve downsampling, and the number of feature maps doubles with the number of downsampling times, and the network depth is set to 5. M>0, and M is an integer. Decoding process: Upsampling is performed using the nearest neighbor upsampling method, which is cascaded at the same level as the encoding process, and the convolution results of the encoding process and the decoding process are spliced; Output process: The model input size is restored to N×N×N, and a linear activation function is used to generate a stenosis rate prediction map that matches the vascular segmentation map.

9. The vascular stenosis rate prediction model according to claim 8, characterized in that: The method for generating the first vascular stenosis rate distribution map includes: Performing vascular centerline extraction processing on the vascular segmentation image to generate a vascular centerline image; Performing optimized centerline processing on the vascular centerline image to generate an optimized centerline vascular image; Performing midline serialization processing on the optimized midline vascular image to generate a midline serialized vascular image; Performing simulated diameter-changing processing on the midline serialized blood vessel image to generate a simulated diameter-changing blood vessel image; wherein the simulated diameter-changing blood vessel image includes a plurality of first pixels; Each pixel value of the first pixel is set to V to obtain a first blood vessel stenosis rate distribution map.

10. A vascular stenosis rate prediction system, characterized in that: The device includes: A data acquisition module, used for acquiring angiographic images; A data processing module is used to perform blood vessel segmentation processing on the angiography image to obtain a blood vessel segmentation map; A prediction module, comprising a pre-trained vascular stenosis rate prediction model, is used to use the vascular segmentation map as an input of the pre-trained vascular stenosis rate prediction model, and output a vascular stenosis rate distribution map; wherein the vascular stenosis rate distribution map comprises a plurality of first pixels; The display module is used to receive the image data outputted thereto in any one of the vascular stenosis rate prediction methods of claims 1-7, and to display and output the image data.

Images