Methods and apparatus for processing fractional flow reserve based on angiography images

Abstract

This invention relates to a method and apparatus for processing fractional flow reserve (FLR) based on angiographic images. The method includes: acquiring an angiographic image as a first image; performing target detection and semantic segmentation on the first image for the coronary artery inlet vessel and the stenotic segment vessel; performing coronary artery inlet diameter analysis to generate the coronary artery inlet diameter; performing stenotic segment vessel diameter and stenosis rate analysis to generate a first stenotic segment diameter sequence and a first stenosis rate; and using user-marked FLR analysis points as target points P. i ; Confirm the generation of the corresponding vascular branch C i The invention extracts vascular branch features to generate branch feature data sequences and sets; inputs the branch feature data sets into an artificial neural network model for computation, outputting a set of branch computation results; and outputs the fractional flow reserve analysis results of each branch in the set as the target point. This invention reduces measurement difficulty and improves measurement safety and efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a processing method and apparatus for analyzing fractional flow reserve based on angiography images. Background Technology

[0002] Fractional flow reserve (FFR) refers to the ratio of the mean pressure in the coronary artery distal to the stenosis to the mean pressure in the aorta at the coronary ostium under maximal coronary congestion. Normally, FFR values ​​are obtained through percutaneous coronary intervention (PCI). However, this method is not only complex but also invasive, posing certain risks and potential harm to the patient. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing a method, apparatus, electronic device, and computer-readable storage medium for analyzing fractional flow reserve (FFR) from angiographic images. Based on an image target detection and semantic segmentation model, the invention identifies coronary artery ostial segments and stenotic segments in coronary angiographic images. It then performs quantitative coronary artery angiography (QCA) analysis on the diameter changes and stenosis rates of the identified coronary ostial segments and stenotic segments. When the user marks FFR measurement points on the image, an artificial neural network (ANN) model is used to analyze the FFR based on the aforementioned QCA analysis results to obtain the corresponding FFR values. This invention not only avoids personal injury caused by invasive measurements but also significantly reduces measurement difficulty, improves measurement safety, and increases measurement efficiency.

[0004] To achieve the above objectives, a first aspect of the present invention provides a method for processing fractional flow reserve based on angiographic image analysis, the method comprising:

[0005] Acquire angiographic images as the first image;

[0006] Based on a preset image target detection and semantic segmentation model, the first image is processed for target detection and semantic segmentation of coronary artery inlet vessels and stenotic vessels, thereby obtaining a target detection box A and one or more target detection boxes B on the first image; each of the target detection boxes A and B includes a segment of vessel mask image, the target type corresponding to the target detection box A is the coronary artery inlet vessel type, and the target type corresponding to the target detection box B is the stenotic vessel type;

[0007] The coronary artery inlet diameter is analyzed for the vascular mask image of the target detection box A to generate the corresponding coronary artery inlet diameter; and the stenosis segment diameter and stenosis rate are analyzed for the vascular mask images of each of the target detection boxes B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate.

[0008] The fractional blood flow reserve analysis points marked by the user on the first image are used as the corresponding target points P. i and will the user's target P i The blood vessel diameter measurement result at the location is used as the corresponding target point diameter; where 1≤i≤n, and n is the total number of measurement points;

[0009] For each of the aforementioned target points P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n );

[0010] The branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i The results of fractional blood flow reserve analysis are output.

[0011] Preferably, the image target detection and semantic segmentation model includes the Mask R-CNN model;

[0012] When the image object detection and semantic segmentation model is specifically the Mask R-CNN model, it includes a feature extraction network layer, a region candidate network layer, a region alignment network layer, and a region head network layer.

[0013] The feature extraction network layer is connected to the region candidate network layer, the region candidate network layer is connected to the region alignment network layer, and the region alignment network layer is connected to the region head network layer.

[0014] The feature extraction network layer is specifically composed of a five-level residual network and a corresponding five-level feature pyramid network; the region candidate network layer includes a five-level region candidate network, which corresponds to the five-level feature pyramid network; when implementing the five-level residual network, the ResNet-50 residual network structure is used and it serves as the backbone network of the feature extraction network layer.

[0015] The region head network layer includes two sub-networks: a target detection branch network and a target segmentation branch network. The target detection branch network is used to output the target detection box A, which is the type of coronary artery inlet vessel, and the target detection box B, which is the type of stenotic vessel. The target segmentation branch network is used to output the vascular mask image of the corresponding coronary artery inlet vessel in the target detection box A, and the vascular mask image of the corresponding stenotic vessel in the target detection box B.

[0016] Preferably, the step of performing coronary artery inlet diameter analysis on the blood vessel mask image of the target detection box A to generate the corresponding coronary artery inlet diameter specifically includes:

[0017] Use the blood vessel mask image of the target detection box A as the current blood vessel mask image;

[0018] The current blood vessel mask image is used to identify the blood vessel edges and the blood vessel center line to generate the corresponding current blood vessel edges and the current center line;

[0019] The multiple pixels included in the current center line are sequentially labeled as the corresponding first pixel X. 1,j Where 1≤j≤m1, m1 is the total number of pixels on the current center line; the first pixel X 1,j＝1 The last first pixel is the blood flow inflow point of the blood vessel corresponding to the current blood vessel mask image. The blood outflow point of the blood vessel corresponding to the current blood vessel mask image;

[0020] After each of the first pixel points X1,j Multiple first intersecting line segments are obtained by constructing intersecting line segments with the current blood vessel edge; and the shortest first intersecting line segment is selected as the length of the first intersecting line segment with the current first pixel X. 1,j The corresponding first pixel point blood vessel diameter L 1,j ;

[0021] For all the first pixel points, the blood vessel diameter L 1,j The coronary artery inlet diameter is generated by averaging.

[0022] Preferably, the step of analyzing the stenosis segment diameter and stenosis rate of the vascular mask image of each of the target detection boxes B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate specifically includes:

[0023] Use the blood vessel mask image of the current target detection box B as the current blood vessel mask image;

[0024] The current blood vessel mask image is used to identify the blood vessel edges and the blood vessel center line to generate the corresponding current blood vessel edges and the current center line;

[0025] The multiple pixels included in the current center line are sequentially labeled as the corresponding second pixel X. 2,h Where 1≤h≤m2, m2 is the total number of pixels on the current center line; the first second pixel X 2,h＝1 The last second pixel is the blood flow inflow point of the blood vessel corresponding to the current blood vessel mask image. The blood outflow point of the blood vessel corresponding to the current blood vessel mask image;

[0026] After each of the second pixel points X 2,h Multiple second intersection segments are obtained by constructing line segments that intersect with the current blood vessel edge; and the shortest second intersection segment is selected as the length of the line segment intersecting with the current second pixel X. 2,h The corresponding second pixel point blood vessel diameter L 2,h ;

[0027] Additionally, based on the blood vessel diameter L of the second pixel... 2,h＝1 Second pixel blood vessel diameter The construction can reflect the first and second pixel points X 2,h＝1 To the last second pixel The linear change function f(h) represents the linear relationship between blood vessel changes.

[0028] f(h) = L 2,h＝1 +k*(h-1),

[0029]

[0030] And according to the linear transformation function f(h), for each of the second pixel points X 2,h The corresponding linearly varying diameter length is used to calculate and generate the corresponding linear diameter L′ of the second pixel. 2,h ,

[0031]

[0032] Then, based on the blood vessel diameter L of the second pixel... 2,h and the linear diameter L′ of the second pixel 2,h Calculate the X of each of the second pixels. 2,h The corresponding second pixel narrowing rate R 2,h ,

[0033]

[0034] The blood vessel diameter L of the second pixel 2,h As the corresponding first stenotic segment diameter, and sorting all the first stenotic segment diameters in ascending order of the second pixel index h, a sequence of first stenotic segment diameters corresponding to the current blood vessel mask image is generated; and the stenosis rate R of all the obtained second pixels is used as the first stenotic segment diameter. 2,h The maximum value is selected as the first stenosis rate corresponding to the current vascular mask image.

[0035] Preferably, the step of targeting each of the target points P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n Specifically, this includes:

[0036] In the first image, the starting position of the blood vessel in the blood vessel mask image of the target detection box A is taken as the coronary artery inlet position; and according to the blood flow direction, the distance from the coronary artery inlet position to each of the target points P is calculated. i The blood flow path is used as the connection between each of the target points P. i The corresponding vascular branch C i ;

[0037] Each of the aforementioned vascular branches C i All the first stenotic segment diameter sequences of the target detection boxes B that pass through are included in the corresponding first sequence set; and in each first sequence set, the first stenotic segment diameters of all the first stenotic segment diameter sequences are sorted according to the blood flow direction to obtain the corresponding first branch stenotic segment diameter sequence.

[0038] From each of the aforementioned vascular branches C i Among all the target detection boxes B that have passed through, the maximum value is selected as the corresponding first branch maximum narrowing rate;

[0039] This will be related to each of the aforementioned vascular branches C i The corresponding target point P i The diameter of the target point is taken as the diameter of the corresponding first branch target point;

[0040] For each of the aforementioned vascular branches C i The coronary artery inlet diameter, the corresponding first branch stenosis segment diameter sequence, the first branch target point diameter, and the first branch maximum stenosis rate are combined to form the corresponding branch feature data sequence S. i ;

[0041] All branch feature data sequences S are processed in ascending order of target index i. i The data is sorted to obtain the branch feature data set D. in (S1…S i …S n ).

[0042] Preferably, the branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n Specifically, this includes:

[0043] The branch feature data set D in (S1…S i …S n The data is divided into n one-dimensional segments; each segment corresponds to a branch feature data sequence S. i ;

[0044] Each of the first data segments is input into the artificial neural network (ANN) model for computation to obtain the corresponding branch operation result U. i ;

[0045] The results of all branch operations, U, are processed in ascending order of target index i. i Sort the results to obtain the set D of the branch operation results. out (U1…U i …U n ).

[0046] Furthermore, the artificial neural network (ANN) model includes an input layer, one or more hidden layers, and an output layer; the input layer includes multiple input layer nodes; each hidden layer includes multiple hidden layer nodes; and the output layer includes one output layer node.

[0047] The input layer is used to input each segment of the first data segment into the corresponding input layer node as the corresponding node output value;

[0048] Each hidden node of the first hidden layer is fully connected to all the input layer nodes to form a corresponding first fully connected network. The first hidden layer is used to input the node output values ​​of all the input layer nodes in the corresponding first fully connected network into a preset fully connected linear operation function to generate a corresponding first result, and input the first result into a preset activation function to generate a corresponding second result, and use the second result as the node output value of the current hidden layer node.

[0049] Each hidden node in the next hidden layer is fully connected to all hidden nodes in the previous hidden layer to form a corresponding second fully connected network. The next hidden layer is used to input the node output values ​​of all hidden nodes in the previous hidden layer in the corresponding second fully connected network into a preset fully connected linear operation function to generate a corresponding third result, and input the third result into a preset activation function to generate a corresponding fourth result, and use the fourth result as the node output value of the current hidden layer node.

[0050] The output layer nodes of the output layer are fully connected to all the hidden layer nodes of the last hidden layer to form a corresponding third fully connected network. The output layer is used to input the node output values ​​of all the hidden layer nodes of the last hidden layer in the corresponding third fully connected network into a preset fully connected linear operation function to generate a corresponding fifth result. The fifth result is then input into a preset activation function to generate a corresponding sixth result, and the sixth result is used as the branch operation result U. i ;

[0051] The activation function used by the hidden layer and the output layer is the ReLU function by default.

[0052] A second aspect of the present invention provides an apparatus for implementing the method described in the first aspect above, comprising: an acquisition module, an image target detection and semantic segmentation module, a quantitative coronary angiography analysis and processing module, a target processing module, a vascular branch processing module, and a fractional flow reserve processing module;

[0053] The acquisition module is used to acquire angiography images as the first image;

[0054] The image target detection and semantic segmentation module is used to perform target detection and semantic segmentation processing on the first image based on a preset image target detection and semantic segmentation model, thereby obtaining a target detection box A and one or more target detection boxes B on the first image; each of the target detection boxes A and B includes a segment of vascular mask image, the target type corresponding to the target detection box A is the coronary inlet vessel type, and the target type corresponding to the target detection box B is the stenotic vessel type;

[0055] The quantitative coronary angiography analysis and processing module is used to perform coronary artery inlet diameter analysis on the vascular mask image of the target detection box A to generate the corresponding coronary artery inlet diameter; and to perform stenosis segment diameter and stenosis rate analysis on the vascular mask image of each target detection box B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate.

[0056] The target processing module is used to use the fractional blood flow reserve analysis points marked by the user on the first image as the corresponding target points P. i and will the user's target P i The blood vessel diameter measurement result at the location is used as the corresponding target point diameter; where 1≤i≤n, and n is the total number of measurement points;

[0057] The vascular branch processing module is used to process each of the target points P. i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n );

[0058] The fractional flow reserve processing module is used to process the branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i The results of fractional blood flow reserve analysis are output.

[0059] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0060] The processor is used to couple with the memory, read and execute instructions in the memory to implement the steps of the method described in the first aspect above;

[0061] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0062] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a computer, cause the computer to perform the instructions described in the first aspect.

[0063] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for analyzing fractional flow reserve (FFR) from angiographic images. Based on an image target detection and semantic segmentation model, it identifies coronary artery ostial segments and stenotic segments in coronary angiographic images. QCA analysis is then performed on the diameter changes and stenosis rates of the identified coronary ostial segments and stenotic segments. When a user marks FFR measurement points on the image, an ANN model is used to analyze the FFR based on the aforementioned QCA analysis results to obtain the corresponding FFR values. This invention not only avoids personal injury caused by invasive measurements but also significantly reduces the difficulty of measurement, improving measurement safety and efficiency. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of a method for processing fractional flow reserve based on angiography image analysis according to Embodiment 1 of the present invention;

[0065] Figure 2 This is a module structure diagram of a processing device based on angiography image analysis fraction of blood flow reserve provided in Embodiment 2 of the present invention;

[0066] Figure 3 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0068] Embodiment 1 of the present invention provides a method for processing fractional flow reserve based on angiography image analysis, such as... Figure 1 The schematic diagram shows a method for processing fractional flow reserve based on angiography image analysis provided in Embodiment 1 of the present invention. This method mainly includes the following steps:

[0069] Step 1: Obtain an angiography image as the first image.

[0070] Here, coronary angiography involves injecting contrast agent into the blood vessel being measured and capturing images of the contrast agent passing through the coronary artery under X-ray. The image data obtained by coronary angiography is the coronary angiography image. The angiography image in the current step is by default the coronary angiography image with a relatively obvious contrast agent filling effect during a single coronary angiography process.

[0071] Step 2: Based on the preset image target detection and semantic segmentation model, target detection and semantic segmentation processing of coronary artery inlet vessels and stenotic vessels are performed on the first image to obtain a target detection box A and one or more target detection boxes B on the first image;

[0072] Among them, both target detection boxes A and B include a segment of vascular mask image. The target type corresponding to target detection box A is the coronary artery inlet vessel type, and the target type corresponding to target detection box B is the stenotic segment vessel type.

[0073] There are various ways to implement image object detection and semantic segmentation models. One of these methods is based on the neural network architecture of the Mask R-CNN model. When the image object detection and semantic segmentation model is specifically the Mask R-CNN model, its neural network structure can be found in the paper "Mask R-CNN" by Kaiming He, Georgia Gkioxari, Piotr Doll'ar, and Ross Girshick. It includes: a feature extraction network layer, a Region Proposal Network (RPN) layer, a Region Of Interest Align (ROI Align) network layer, and a Region Of Interest Head (ROI HEAD) network layer. The feature extraction network layer is connected to the region proposal network layer, the region proposal network layer is connected to the region alignment network layer, and the region alignment network layer is connected to the region head network layer.

[0074] The feature extraction network layer in this embodiment of the invention specifically consists of a five-level residual network (ResNet) and a corresponding five-level feature pyramid network (FPN); the region candidate network layer includes a five-level region candidate network, which corresponds to the five-level feature pyramid network; when implementing the five-level residual network, this embodiment of the invention uses the ResNet-50 residual network structure and uses it as the backbone network of the feature extraction network layer;

[0075] The region head network layer of this invention includes two sub-networks: a target detection branch network and a target segmentation branch network. The target detection branch network is used to output a target detection box A for detecting a coronary artery inlet vessel and a target detection box B for detecting a stenotic vessel. The target segmentation branch network is used to output the vascular mask image of the corresponding coronary artery inlet vessel in target detection box A and the vascular mask image of the corresponding stenotic vessel in target detection box B.

[0076] Step 3: Analyze the coronary artery inlet diameter of the vascular mask image of target detection box A to generate the corresponding coronary artery inlet diameter; and analyze the stenosis segment diameter and stenosis rate of the vascular mask images of each target detection box B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate.

[0077] Specifically, this includes: Step 31, performing coronary artery inlet diameter analysis on the blood vessel mask image of the target detection box A to generate the corresponding coronary artery inlet diameter;

[0078] Specifically, this includes: step 311, using the blood vessel mask image of the target detection box A as the current blood vessel mask image;

[0079] Step 312: Recognize the vessel edges and center lines of the current vessel mask image to generate the corresponding current vessel edges and center lines;

[0080] Here, when identifying the blood vessel edges and center lines of the current blood vessel mask image, all image information within the target detection box A of the current blood vessel mask image is first used as the first detection box image; then, the first detection box image is binarized to generate a first binary image; wherein, the pixel values ​​of all pixels in the first binary image are only of two types: the foreground pixel value of the foreground pixel and the background pixel value of the background pixel, and the foreground pixel value and the background pixel value can be defined by the user in the specific implementation;

[0081] Then, all pixels covered by the current blood vessel mask image are traversed on the first binary image. If there is at least one background pixel within the four neighboring pixels of the currently traversed pixel, the currently traversed pixel is taken as the blood vessel edge pixel. After the traversal is completed, all blood vessel edge pixels are connected in a clockwise or counterclockwise direction to obtain the closed blood vessel edge curve of the current blood vessel mask image, which is the current blood vessel edge.

[0082] Furthermore, without altering the topological properties of the blood vessel image (the connectivity of the blood vessels), a centerline is extracted from the current blood vessel mask image on the first binary image using a topology thinning method to generate the current centerline. Specifically, this involves: pre-creating connectivity judgment rules for the blood vessels, including multiple sub-rules, where each sub-rule corresponds to one or more 3×3 pixel value template matrices, and each pixel value template matrix corresponds to a pixel arrangement structure that maintains connectivity; then, traversing all pixels covered by the current blood vessel mask image on the first binary image; during traversal, the pixels of the eight neighborhood pixels of the currently traversed pixel are... The extracted pixel values ​​are combined with the pixel values ​​of the currently traversed pixel to form a 3×3 pixel value matrix, denoted as the first matrix. Then, one or more 3×3 pixel value template matrices from each sub-rule of the connectivity judgment rule are compared sequentially with the first matrix. During the comparison, if the first matrix completely matches any of the pixel value template matrices, it means the currently traversed pixel is a key pixel for maintaining the connectivity of its neighboring pixels and cannot be deleted; the process jumps to the next pixel. During the comparison, if the first matrix does not match the pixel value template matrices of any of the sub-rules of the connectivity judgment rule, it means the currently traversed pixel is not... Key pixels that maintain connectivity with neighboring pixels can be deleted, and are marked as deletable pixels. After traversing all pixels covered by the current blood vessel mask image, all pixels marked as deletable are converted to background pixels. This achieves a topological refinement of the current blood vessel mask image without destroying its topological properties (connectivity). After converting all deletable pixels to background pixels, the above steps are repeated for all pixels covered by the topologically refined current blood vessel mask image. The process iterates through the image, and during each iteration, it uses the pixel value template matrix based on the connectivity rules to determine whether each pixel is a key pixel, whether it needs to be retained, or whether it should be marked as a deletable pixel. After each iteration, all pixels marked as deletable pixels are converted to background pixels. This process is repeated iteratively until the number of pixels marked as deletable is 0 after each iteration, indicating that the current blood vessel mask image can no longer be further refined. At this point, the remaining pixels in the current blood vessel mask image are connected sequentially according to the blood flow direction, and the resulting curve is the current centerline.

[0083] Step 313: Sequentially label the multiple pixels included by the current center line as the corresponding first pixel X. 1,j ;

[0084] Where 1≤j≤m1, m1 is the total number of pixels on the current center line; the first pixel X 1,j＝1 The last first pixel of the blood flow inflow point corresponding to the blood vessel in the current blood vessel mask image. This represents the blood outflow point of the blood vessel corresponding to the current vascular mask image;

[0085] Step 314, passing through each first pixel X 1,j Multiple first intersection segments are obtained by constructing line segments that intersect with the current blood vessel edge; the shortest first intersection segment is selected as the length of the line segment intersecting with the current first pixel X. 1,j The corresponding first pixel point blood vessel diameter L 1,j ;

[0086] Here, according to each first pixel point X 1,j Its arrangement relationship with the pixels in its eight neighboring regions (top left, top, top right, right, bottom right, bottom, bottom left, left) is determined by passing through each first pixel X. 1,j Draw four straight lines: The first straight line (top left - X) 1,j - Bottom left), second straight line (above -X) 1,j -below), the third straight line (upper right -X) 1,j - Bottom left) and the fourth line (right-X) 1,j -Left side); Extract the intersection segments of the first, second, third, and fourth straight lines with the current blood vessel edge as the corresponding first intersection segments 1, 2, 3, and 4; From the first intersection segments 1, 2, 3, and 4, extract the shortest one as the blood vessel diameter L of the first pixel. 1,j ;

[0087] Step 315, calculate the blood vessel diameter L of all the first pixels. 1,j The coronary artery inlet diameter is generated by calculating the mean value.

[0088] Here, using the mean calculation method can reduce calculation errors in the image processing process;

[0089] Step 32: Analyze the stenosis segment diameter and stenosis rate of the vascular mask image of each target detection box B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate;

[0090] Specifically, this includes: step 321, using the blood vessel mask image of the current target detection box B as the current blood vessel mask image;

[0091] Step 322: Recognize the blood vessel edges and center lines of the current blood vessel mask image to generate the corresponding current blood vessel edges and center lines;

[0092] Step 323: Sequentially label the multiple pixels included in the current center line as the corresponding second pixel X. 2,h ;

[0093] Where 1≤h≤m2, m2 is the total number of pixels on the current center line; the first second pixel X 2,h＝1This refers to the blood flow inflow point of the corresponding blood vessel in the current blood vessel mask image, specifically the last second pixel. This represents the blood outflow point of the blood vessel corresponding to the current vascular mask image;

[0094] Step 324, passing through each second pixel point X 2,h Draw a line segment that intersects the current blood vessel edge to obtain multiple second intersecting line segments; and select the shortest second intersecting line segment as the length of the line segment intersecting the current second pixel X. 2,h The corresponding second pixel point blood vessel diameter L 2,h ;

[0095] This is similar to step 314, and will not be elaborated further here;

[0096] Step 325, based on the blood vessel diameter L of the second pixel. 2,h＝1 Second pixel blood vessel diameter The construction can reflect the first and second pixel points X 2,h＝1 To the last second pixel The linear change function f(h) represents the linear relationship between blood vessel changes.

[0097] f(h) = L 2,h＝1 +k*(h-1),

[0098]

[0099] Here, in the embodiments of the present invention, it is assumed that the diameter of the coronary artery should decrease linearly from the coronary artery inlet to the end point of each branch under normal circumstances without stenosis of the intermediate segment of the vessel due to lesions. The above-mentioned linear change function f(h) is the function used to simulate this linear decrease process.

[0100] Step 326: Based on the linear transformation function f(h), for each second pixel point X 2,h The corresponding linearly varying diameter length is used to calculate and generate the corresponding linear diameter L′ of the second pixel. 2,h ,

[0101]

[0102] Here, the linear diameter L′ of the second pixel is... 2,h That is, the simulation of each second pixel point X under normal conditions based on the above linear change function f(h) 2,h The corresponding normal blood vessel diameter;

[0103] Step 327, based on the blood vessel diameter L of the second pixel. 2,h The linear diameter L′ of the second pixel 2,h Calculate the X value of each second pixel. 2,hThe corresponding second pixel narrowing rate R 2,h ,

[0104]

[0105] Here, the blood vessel diameter L at the second pixel is... 2,h The actual diameter of the blood vessel, and the linear diameter L′ of the second pixel. 2,h To simulate the theoretical blood vessel diameter under normal conditions, The ratio of the actual diameter to the theoretical diameter can reflect the value of the second pixel X. 2,h The degree of blood flow, naturally This reflects the second pixel X. 2,h The degree of blood flow obstruction is also the second pixel X. 2,h The corresponding second pixel narrowing rate R 2,h ;

[0106] Step 328, adjust the diameter L of the blood vessel at the second pixel. 2,h As the corresponding first stenotic segment diameter, all first stenotic segment diameters are sorted in ascending order of second pixel index h to generate a first stenotic segment diameter sequence corresponding to the current blood vessel mask image; and the stenosis rate R of all obtained second pixels is used as the basis for the sequence. 2,h The maximum value is selected as the first stenosis rate corresponding to the current vascular mask image.

[0107] Here, the first stenotic segment diameter sequence reflects the diameter of the vessel at each point on the corresponding stenotic segment, and the first stenosis rate is the maximum stenosis rate on the corresponding stenotic segment.

[0108] It should be noted that, when calculating the first stenosis rate corresponding to the current blood vessel mask image, this embodiment of the invention supports multiple processing methods; one of which, as can be seen from steps 327-328 above, first calculates the complete stenosis rate sequence, and then selects the maximum value from it as the first stenosis rate; another method is to first calculate the blood vessel diameter L from all second pixel points. 2,h The minimum value is selected as the minimum blood vessel diameter L of the second pixel. min and L min The corresponding pixel index y is substituted into the linear transformation function f(h=y) to calculate the linear reference diameter L′ corresponding to that pixel index y. 2,h＝y Then L min and L′ 2,h＝y Substitution The calculation results in R, which is then output as the first narrowing rate. The former method obtains the narrowing rate at each position along the center line of the narrow segment while obtaining the first narrowing rate, while the latter method is simpler and faster because it does not require outputting the narrowing rate at each position.

[0109] Step 3 above is the QCA analysis process performed using image processing methods. This QCA analysis process yields the analysis results of the coronary artery inlet diameter, the diameter sequence of all stenotic segments, and the stenosis rate on the first image. Based on this, in this embodiment of the invention, subsequent steps 4-6 can further measure the FFR values ​​of the marked measurement points on any stenotic segment of the blood vessel in the first image.

[0110] Step 4: Use the fractional flow reserve analysis points marked by the user on the first image as the corresponding target points P. i and will the user target point P i The blood vessel diameter measurement result at that location is used as the corresponding target point diameter;

[0111] Where i is the target index, 1≤i≤n, and n is the total number of measurement points.

[0112] Here, target P i This refers to the fractional flow reserve measurement point marked by the user on the first image, which is a location point on a certain segment of the blood vessel in the first image. Normally, the user will mark the target point P on a blood vessel at a certain distance from the last stenotic segment of the blood flow branch with the stenotic segment. i The user completed the target point P. i After marking, the target point P can be obtained using the image scale on the first image. i The diameter of the blood vessel cross-section at the location is used as the target point P. i The blood vessel diameter measurement result at that location is also the corresponding target point diameter.

[0113] Step 5, for each target point P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each blood vessel branch C i The corresponding first stenotic segment diameter sequences and first stenosis rates, and the corresponding target point P i The target diameter and coronary artery inlet diameter are used to extract vascular branch features and generate corresponding branch feature data sequences S. i And for all the branch feature data sequences S obtained i The branch feature data set D is obtained by sorting. in (S1…S i …S n );

[0114] Specifically, this includes: Step 51, in the first image, taking the starting position of the blood vessel in the blood vessel mask image of the target detection box A as the coronary artery inlet position; and according to the blood flow direction, from the coronary artery inlet position to each target point P. i The blood flow path is used as a reference for each target point P.i Corresponding vascular branch C i ;

[0115] Here, vascular branch C i This refers to the distance from the coronary artery inlet to each target point P. i A single blood flow path at a given location;

[0116] Step 52, divide each blood vessel branch C i The first stenotic segment diameter sequences of all target detection boxes B along the path are included in the corresponding first sequence set; and in each first sequence set, the first stenotic segment diameters of all first stenotic segment diameter sequences are sorted according to the blood flow direction to obtain the corresponding first branch stenotic segment diameter sequence.

[0117] Here, if we consider the distance from the coronary artery inlet to the current target point P... i If a single blood flow path at a location contains only one narrow segment of a blood vessel, then branch C of that vessel... i If only one target detection box B is passed, the corresponding first sequence set will only include the first stenotic segment diameter sequence of this unique target detection box B, and the first branch stenotic segment diameter sequence is also the first stenotic segment diameter sequence of this unique target detection box B; if from the coronary artery inlet position to the current target point P i If a single blood flow path at a location passes through multiple narrowed vascular segments, then the vascular branch C... i Having passed through multiple target detection boxes B, the corresponding first sequence set includes the first stenotic segment diameter sequences of multiple target detection boxes B. Sort these multiple first stenotic segment diameter sequences according to the blood flow direction to achieve the full sorting process in this step. The resulting first branch stenotic segment diameter sequence then includes the current vessel branch C. i From the coronary artery inlet to the corresponding target point P i The parameters showing the change in vessel diameter of all narrowed segments of the blood flow path at the location;

[0118] It should be noted that after obtaining the diameter sequence of the first branch narrow segment, in order to facilitate subsequent model calculations, this embodiment of the invention will also perform data shaping to ensure that its data length is consistent with the length required by the model.

[0119] Step 53, from each vascular branch C i Among all the target detection boxes B that pass through, the maximum value is selected as the corresponding first branch maximum narrowing rate;

[0120] Here, if blood vessel branch C i If the number of target detection boxes B along the path is unique, then the maximum stenosis rate of the first branch is the maximum stenosis rate of that unique target detection box B, which is also the first stenosis rate; if the blood vessel branch C iIf the number of target detection boxes B along the path is not unique, then the maximum stenosis rate of the first branch should be the maximum value among the first stenosis rates of multiple target detection boxes B, which is the value of the vascular branch C. i The maximum stenosis rate on the corresponding single blood flow path;

[0121] Step 54, connect with each vascular branch C i Corresponding target point P i The target diameter is taken as the corresponding first branch target diameter;

[0122] Here, each vascular branch C i The diameter of the target point in the first branch is actually the target point P at the end of that branch. i The diameter of the blood vessel's cross-section at its location;

[0123] Step 55, for each vascular branch C i The coronary artery inlet diameter, the corresponding first branch stenosis segment diameter sequence, the first branch target point diameter, and the first branch maximum stenosis rate are combined to form the corresponding branch feature data sequence S. i ;

[0124] Here, the branch feature data sequence S is composed of the coronary artery inlet diameter, the first branch stenosis segment diameter sequence, the first branch target point diameter, and the first branch maximum stenosis rate. i There are several ways to combine data, one of which is: branch feature data sequence S i = Coronary artery inlet diameter + First branch stenotic segment diameter sequence + First branch target point diameter + First branch maximum stenosis rate;

[0125] Step 56: Process all branch feature data sequences S in ascending order of target index i. i Sort the data to obtain the branch feature data set D. in (S1…S i …S n ).

[0126] Step 6, set the branch feature data D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i Output the results of fractional blood flow reserve analysis;

[0127] Specifically, this includes: Step 61, which involves processing the branch feature data set D... in (S1…Si …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n );

[0128] Specifically, this includes: Step 611, which involves processing the branch feature data set D... in (S1…S i …S n The data is divided into n one-dimensional segments; each segment corresponds to a branch feature data sequence S. i ;

[0129] Step 612: Input each first data segment into the artificial neural network (ANN) model for computation to obtain the corresponding branch operation result U. i ;

[0130] The Artificial Neural Network (ANN) model includes an input layer, one or more hidden layers, and an output layer; the input layer includes multiple input layer nodes; each hidden layer includes multiple hidden layer nodes; and the output layer includes one output layer node.

[0131] When the Artificial Neural Network (ANN) model processes the first input data segment:

[0132] The input layer is used to input the data of each segment of the first data segment into the corresponding input layer node as the corresponding node output value;

[0133] Each hidden node of the first hidden layer is fully connected to all input layer nodes to form a corresponding first fully connected network. The first hidden layer is used to input the node output values ​​of all input layer nodes in the corresponding first fully connected network into a preset fully connected linear operation function to generate a corresponding first result, and input the first result into a preset activation function to generate a corresponding second result, and use the second result as the node output value of the current hidden layer node.

[0134] Each hidden node in the next hidden layer is fully connected to all hidden nodes in the previous hidden layer to form a corresponding second fully connected network. The next hidden layer is used to input the node output values ​​of all hidden nodes in the previous hidden layer in the corresponding second fully connected network into a preset fully connected linear operation function to generate a corresponding third result. The third result is then input into a preset activation function to generate a corresponding fourth result, and the fourth result is used as the node output value of the current hidden layer node.

[0135] The output layer nodes are fully connected to all hidden layer nodes of the last hidden layer to form the corresponding third fully connected network. The output layer is used to input the node output values ​​of all hidden layer nodes of the last hidden layer in the corresponding third fully connected network into a preset fully connected linear operation function to generate the corresponding fifth result. The fifth result is then input into a preset activation function to generate the corresponding sixth result, and the sixth result is used as the branch operation result U. i ;

[0136] In the above operations, the activation function used by the hidden layer and the output layer is the Rectified Linear Units (ReLU) function by default;

[0137] Step 613: Calculate the result U of all branches in ascending order of target index i. i Sort the results to obtain the set D of branch operation results. out (U1…U i …U n );

[0138] Step 62, calculate the results of each branch operation in the set U. i As the corresponding target point P i The results of fractional blood flow reserve analysis are output.

[0139] Figure 2 This is a module structure diagram of a processing device for analyzing fractional flow reserve based on angiography images, provided in Embodiment 2 of the present invention. This device can be a terminal device or server implementing the method of the embodiments of the present invention, or it can be a device connected to the aforementioned terminal device or server to implement the method of the embodiments of the present invention. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 2 As shown, the device includes: an acquisition module 201, an image target detection and semantic segmentation module 202, a quantitative coronary angiography analysis and processing module 203, a target processing module 204, a vascular branch processing module 205, and a fractional flow reserve processing module 206.

[0140] The acquisition module 201 is used to acquire angiography images as the first image.

[0141] The image target detection and semantic segmentation module 202 is used to perform target detection and semantic segmentation processing on the first image based on a preset image target detection and semantic segmentation model, thereby obtaining a target detection box A and one or more target detection boxes B on the first image; both target detection boxes A and B include a segment of vascular mask image, the target type corresponding to target detection box A is the coronary inlet vessel type, and the target type corresponding to target detection box B is the stenotic vessel type.

[0142] The quantitative coronary angiography analysis and processing module 203 is used to perform coronary artery inlet diameter analysis on the vascular mask image of target detection box A to generate the corresponding coronary artery inlet diameter; and to perform stenosis segment diameter and stenosis rate analysis on the vascular mask images of each target detection box B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate.

[0143] The target processing module 204 is used to take the fractional blood flow reserve analysis points marked by the user on the first image as the corresponding target points P. i and will the user target point P i The blood vessel diameter measurement result at the location is used as the corresponding target point diameter; where i is the target point index, 1≤i≤n, and n is the total number of measurement points.

[0144] The vascular branch processing module 205 is used for processing various target points P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each blood vessel branch C i The corresponding first stenotic segment diameter sequences and first stenosis rates, and the corresponding target point P i The target diameter and coronary artery inlet diameter are used to extract vascular branch features and generate corresponding branch feature data sequences S. i And for all the branch feature data sequences S obtained i The branch feature data set D is obtained by sorting. in (S1…S i …S n ).

[0145] Fractional flow reserve processing module 206 is used to process branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i The results of fractional blood flow reserve analysis are output.

[0146] The present invention provides a processing device for analyzing fractional flow reserve based on angiography images, which can execute the method steps in the above method embodiments. Its implementation principle and technical effect are similar, and will not be repeated here.

[0147] It should be noted that the division of the various modules in the above device is merely a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, these modules can be implemented entirely in software via processing element calls; they can be fully implemented in hardware; or some modules can be implemented by processing element calls to software, while others are implemented in hardware. For example, the acquisition module can be a separate processing element, or it can be integrated into a chip in the above device. Alternatively, it can be stored as program code in the memory of the above device, and called and executed by a processing element of the device. The implementation of other modules is similar. Moreover, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed through integrated logic circuits in the hardware of the processor element or through software instructions.

[0148] For example, these modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more Digital Signal Processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). As another example, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a System-on-a-Chip (SOC).

[0149] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer described above can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions described above can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, Bluetooth, microwave, etc.) means. The computer-readable storage medium described above can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The aforementioned available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0150] Figure 3 This is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention. This electronic device can be the aforementioned terminal device or server, or it can be a terminal device or server connected to the aforementioned terminal device or server that implements the method of the embodiments of the present invention. Figure 3 As shown, the electronic device may include: a processor 301 (e.g., CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transmission and reception operations of the transceiver 303. The memory 302 may store various instructions for performing various processing functions and implementing the methods and processes provided in the above embodiments of the present invention. Preferably, the electronic device involved in the embodiments of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The communication port 306 is used for communication between the electronic device and other peripherals.

[0151] exist Figure 3The system bus mentioned can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used to enable communication between the database access device and other devices (e.g., clients, read-write libraries, and read-only libraries). Memory may include Random Access Memory (RAM) and may also include non-volatile memory, such as at least one disk drive.

[0152] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0153] It should be noted that the embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to execute the methods and processes provided in the above embodiments.

[0154] This invention also provides a chip for executing instructions, which is used to execute the methods and processes provided in the above embodiments.

[0155] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for analyzing fractional flow reserve (FFR) from angiographic images. Based on an image target detection and semantic segmentation model, it identifies coronary artery ostial segments and stenotic segments in coronary angiographic images. QCA analysis is then performed on the diameter changes and stenosis rates of the identified coronary ostial segments and stenotic segments. When a user marks FFR measurement points on the image, an ANN model is used to analyze the FFR based on the aforementioned QCA analysis results to obtain the corresponding FFR values. This invention not only avoids personal injury caused by invasive measurements but also significantly reduces the difficulty of measurement, improving measurement safety and efficiency.

[0156] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0157] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0158] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for processing fractional flow reserve based on angiographic image analysis, characterized in that, The method includes: Acquire angiographic images as the first image; Based on a preset image target detection and semantic segmentation model, the first image is processed for target detection and semantic segmentation of coronary artery inlet vessels and stenotic vessels, thereby obtaining a target detection box A and one or more target detection boxes B on the first image; each of the target detection boxes A and B includes a segment of vessel mask image, the target type corresponding to the target detection box A is the coronary artery inlet vessel type, and the target type corresponding to the target detection box B is the stenotic vessel type; The coronary artery inlet diameter is analyzed for the vascular mask image of the target detection box A to generate the corresponding coronary artery inlet diameter; and the stenosis segment diameter and stenosis rate are analyzed for the vascular mask images of each of the target detection boxes B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate. The fractional blood flow reserve analysis points marked by the user on the first image are used as the corresponding target points P. i and will the user's target P i The blood vessel diameter measurement result at the location is used as the corresponding target point diameter; where 1≤i≤n, and n is the total number of measurement points; For each of the aforementioned target points P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n ); The branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i Output the results of fractional blood flow reserve analysis; Specifically, the step of analyzing the stenosis segment diameter and stenosis rate of the vascular mask image of each of the target detection boxes B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate includes: Use the blood vessel mask image of the current target detection box B as the current blood vessel mask image; The current blood vessel mask image is used to identify the blood vessel edges and the blood vessel center line to generate the corresponding current blood vessel edges and the current center line; The multiple pixels included in the current center line are sequentially labeled as the corresponding second pixel X. 2,h Where 1≤h≤m2, m2 is the total number of pixels on the current center line; the first second pixel X 2,h＝1 The last second pixel is the blood flow inflow point of the blood vessel corresponding to the current blood vessel mask image. The blood outflow point of the blood vessel corresponding to the current blood vessel mask image; After each of the second pixel points X 2,h Multiple second intersection segments are obtained by constructing line segments that intersect with the current blood vessel edge; and the shortest second intersection segment is selected as the length of the line segment intersecting with the current second pixel X. 2,h The corresponding second pixel point blood vessel diameter L 2,h ; Additionally, based on the blood vessel diameter L of the second pixel... 2,h＝1 Second pixel blood vessel diameter The construction can reflect the first and second pixel points X 2,h＝1 To the last second pixel The linear change function f(h) represents the linear relationship between blood vessel changes. f(h)=L 2,h＝1 +k*(h-1), And according to the linear transformation function f(h), for each of the second pixel points X 2,h The corresponding linearly varying diameter length is used to calculate and generate the corresponding linear diameter L′ of the second pixel. 2,h , Then, based on the blood vessel diameter L of the second pixel... 2,h and the linear diameter L′ of the second pixel 2,h Calculate the X of each of the second pixels. 2,h The corresponding second pixel narrowing rate R 2,h , The blood vessel diameter L of the second pixel 2,h As the corresponding first stenotic segment diameter, and sorting all the first stenotic segment diameters in ascending order of the second pixel index h, a sequence of first stenotic segment diameters corresponding to the current blood vessel mask image is generated; and the stenosis rate R of all the obtained second pixels is used as the first stenotic segment diameter. 2,h The maximum value is selected as the first stenosis rate corresponding to the current vascular mask image.

2. The method for processing fractional flow reserve based on angiography image analysis according to claim 1, characterized in that, The image object detection and semantic segmentation model includes the Mask R-CNN model; When the image object detection and semantic segmentation model is specifically the Mask R-CNN model, it includes a feature extraction network layer, a region candidate network layer, a region alignment network layer, and a region head network layer. The feature extraction network layer is connected to the region candidate network layer, the region candidate network layer is connected to the region alignment network layer, and the region alignment network layer is connected to the region head network layer. The feature extraction network layer is specifically composed of a five-level residual network and a corresponding five-level feature pyramid network; the region candidate network layer includes a five-level region candidate network, which corresponds to the five-level feature pyramid network; when implementing the five-level residual network, the ResNet-50 residual network structure is used and it serves as the backbone network of the feature extraction network layer. The region head network layer includes two sub-networks: a target detection branch network and a target segmentation branch network. The target detection branch network is used to output the target detection box A, which is the type of coronary artery inlet vessel, and the target detection box B, which is the type of stenotic vessel. The target segmentation branch network is used to output the vascular mask image of the corresponding coronary artery inlet vessel in the target detection box A, and the vascular mask image of the corresponding stenotic vessel in the target detection box B.

3. The method for processing fractional flow reserve based on angiography image analysis according to claim 1, characterized in that, The step of performing coronary artery inlet diameter analysis on the blood vessel mask image of the target detection box A to generate the corresponding coronary artery inlet diameter specifically includes: Use the blood vessel mask image of the target detection box A as the current blood vessel mask image; The current blood vessel mask image is used to identify the blood vessel edges and the blood vessel center line to generate the corresponding current blood vessel edges and the current center line; The multiple pixels included in the current center line are sequentially labeled as the corresponding first pixel X. 1,j Where 1≤j≤m1, m1 is the total number of pixels on the current center line; the first pixel X 1,j＝1 The last first pixel is the blood flow inflow point of the blood vessel corresponding to the current blood vessel mask image. The blood outflow point of the blood vessel corresponding to the current blood vessel mask image; After each of the first pixel points X 1,j Multiple first intersecting line segments are obtained by constructing intersecting line segments with the current blood vessel edge; and the shortest first intersecting line segment is selected as the length of the first intersecting line segment with the current first pixel X. 1,j The corresponding first pixel point blood vessel diameter L 1,j ; For all the first pixel points, the blood vessel diameter L 1,j The coronary artery inlet diameter is generated by averaging.

4. The method for processing fractional flow reserve based on angiography image analysis according to claim 1, characterized in that, The respective target points P i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n Specifically, this includes: In the first image, the starting position of the blood vessel in the blood vessel mask image of the target detection box A is taken as the coronary artery inlet position; and according to the blood flow direction, the distance from the coronary artery inlet position to each of the target points P is calculated. i The blood flow path is used as the connection between each of the target points P. i The corresponding vascular branch C i ; Each of the aforementioned vascular branches C i All the first stenotic segment diameter sequences of the target detection boxes B that pass through are included in the corresponding first sequence set; and in each first sequence set, the first stenotic segment diameters of all the first stenotic segment diameter sequences are sorted according to the blood flow direction to obtain the corresponding first branch stenotic segment diameter sequence. From each of the aforementioned vascular branches C i Among all the target detection boxes B that have passed through, the maximum value is selected as the corresponding first branch maximum narrowing rate; This will be related to each of the aforementioned vascular branches C i The corresponding target point P i The diameter of the target point is taken as the diameter of the corresponding first branch target point; For each of the aforementioned vascular branches C i The coronary artery inlet diameter, the corresponding first branch stenosis segment diameter sequence, the first branch target point diameter, and the first branch maximum stenosis rate are combined to form the corresponding branch feature data sequence S. i ; All branch feature data sequences S are processed in ascending order of target index i. i The data is sorted to obtain the branch feature data set D. in (S1…S i …S n ).

5. The method for processing fractional flow reserve based on angiography image analysis according to claim 1, characterized in that, The branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n Specifically, this includes: The branch feature data set D in (S1…S i …S n The data is divided into n one-dimensional segments; each segment corresponds to a branch feature data sequence S. i ; Each of the first data segments is input into the artificial neural network (ANN) model for computation to obtain the corresponding branch operation result U. i ; The results of all branch operations, U, are processed in ascending order of target index i. i Sort the results to obtain the set D of the branch operation results. out (U1…U i …U n ).

6. The method for processing fractional flow reserve based on angiography image analysis according to claim 5, characterized in that, The artificial neural network (ANN) model includes an input layer, one or more hidden layers, and an output layer; the input layer includes multiple input layer nodes; each hidden layer includes multiple hidden layer nodes; and the output layer includes one output layer node. The input layer is used to input each segment of the first data segment into the corresponding input layer node as the corresponding node output value; Each of the hidden layer nodes in the first hidden layer is fully connected to all the input layer nodes to form a corresponding first fully connected network; The first hidden layer is used to input the node output values ​​of all the input layer nodes in the first fully connected network into a preset fully connected linear operation function to generate a corresponding first result, and input the first result into a preset activation function to generate a corresponding second result, and use the second result as the node output value of the current hidden layer node. Each hidden node in the next hidden layer is fully connected to all hidden nodes in the previous hidden layer to form a corresponding second fully connected network. The next hidden layer is used to input the node output values ​​of all hidden nodes in the previous hidden layer in the corresponding second fully connected network into a preset fully connected linear operation function to generate a corresponding third result, and input the third result into a preset activation function to generate a corresponding fourth result, and use the fourth result as the node output value of the current hidden layer node. The output layer nodes of the output layer are fully connected to all the hidden layer nodes of the last hidden layer to form a corresponding third fully connected network; The output layer is used to input the node output values ​​of all hidden layer nodes in the last hidden layer of the corresponding third fully connected network into a preset fully connected linear operation function to generate a corresponding fifth result, and input the fifth result into a preset activation function to generate a corresponding sixth result, and use the sixth result as the branch operation result U. i ; The activation function used by the hidden layer and the output layer is the ReLU function by default.

7. A processing device for analyzing fractional flow reserve based on angiographic images, characterized in that, The device includes: an acquisition module, an image target detection and semantic segmentation module, a quantitative coronary angiography analysis and processing module, a target processing module, a vascular branch processing module, and a fractional flow reserve processing module; The acquisition module is used to acquire angiography images as the first image; The image target detection and semantic segmentation module is used to perform target detection and semantic segmentation processing on the first image based on a preset image target detection and semantic segmentation model, thereby obtaining a target detection box A and one or more target detection boxes B on the first image; each of the target detection boxes A and B includes a segment of vascular mask image, the target type corresponding to the target detection box A is the coronary inlet vessel type, and the target type corresponding to the target detection box B is the stenotic vessel type; The quantitative coronary angiography analysis and processing module is used to perform coronary artery inlet diameter analysis on the vascular mask image of the target detection box A to generate the corresponding coronary artery inlet diameter; and to perform stenosis segment diameter and stenosis rate analysis on the vascular mask image of each target detection box B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate. The target processing module is used to use the fractional blood flow reserve analysis points marked by the user on the first image as the corresponding target points P. i and will the user's target P i The blood vessel diameter measurement result at the location is used as the corresponding target point diameter; where 1≤i≤n, and n is the total number of measurement points; The vascular branch processing module is used to process each of the target points P. i The corresponding vascular branch C is generated by identifying the vascular branch located in the first image. i And according to each of the said vascular branches C i The corresponding first stenotic segment diameter sequence and the first stenosis rate, and the corresponding target point P i The target diameter and the coronary artery inlet diameter are used to extract vascular branch features to generate corresponding branch feature data sequences S. i And for all the obtained branch feature data sequences S i The branch feature data set D is obtained by sorting. in (S1…S i …S n ); The fractional flow reserve processing module is used to process the branch feature data set D in (S1…S i …S n Input a preset artificial neural network (ANN) model, perform calculations, and output a set of branch calculation results D. out (U1…U i …U n ); and the results of each branch operation in the set U i As the corresponding target point P i Output the results of fractional blood flow reserve analysis; Specifically, the step of analyzing the stenosis segment diameter and stenosis rate of the vascular mask image of each of the target detection boxes B to generate the corresponding first stenosis segment diameter sequence and first stenosis rate includes: Use the blood vessel mask image of the current target detection box B as the current blood vessel mask image; The current blood vessel mask image is used to identify the blood vessel edges and the blood vessel center line to generate the corresponding current blood vessel edges and the current center line; The multiple pixels included in the current center line are sequentially labeled as the corresponding second pixel X. 2,h Where 1≤h≤m2, m2 is the total number of pixels on the current center line; the first second pixel X 2,h＝1 The last second pixel is the blood flow inflow point of the blood vessel corresponding to the current blood vessel mask image. The blood outflow point of the blood vessel corresponding to the current blood vessel mask image; After each of the second pixel points X 2,h Multiple second intersection segments are obtained by constructing line segments that intersect with the current blood vessel edge; and the shortest second intersection segment is selected as the length of the line segment intersecting with the current second pixel X. 2,h The corresponding second pixel point blood vessel diameter L 2,h ; Additionally, based on the blood vessel diameter L of the second pixel... 2,h＝1 Second pixel blood vessel diameter The construction can reflect the first and second pixel points X 2,h＝1 To the last second pixel The linear change function f(h) represents the linear relationship between blood vessel changes. f(h)=L 2,h＝1 +k*(h-1), And according to the linear transformation function f(h), for each of the second pixel points X 2,h The corresponding linearly varying diameter length is used to calculate and generate the corresponding linear diameter L′ of the second pixel. 2,h , Then, based on the blood vessel diameter L of the second pixel... 2,h and the linear diameter L′ of the second pixel 2,h Calculate the X of each of the second pixels. 2,h The corresponding second pixel narrowing rate R 2,h , The blood vessel diameter L of the second pixel 2,h As the corresponding first stenotic segment diameter, and sorting all the first stenotic segment diameters in ascending order of the second pixel index h, a sequence of first stenotic segment diameters corresponding to the current blood vessel mask image is generated; and the stenosis rate R of all the obtained second pixels is used as the first stenotic segment diameter. 2,h The maximum value is selected as the first stenosis rate corresponding to the current vascular mask image.

8. An electronic device, characterized in that, include: Memory, processor, and transceiver; The processor is configured to be coupled to the memory, read and execute instructions in the memory to implement the method according to any one of claims 1-6; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1-6.

Images