Automatic monitoring method and monitoring system for coronary artery blood flow

Abstract

The invention relates to an automatic monitoring method and a monitoring system for coronary artery blood flow, which specifically comprises the following steps of: performing blood vessel segmentation on each frame of image, and extracting a mask, a center line and a diameter; feature points meeting the conditions that the normal distance from the center line is smaller than a threshold value and the local energy entropy meets a preset value are screened in the first-frame blood vessel area for tracking; tracking the points among frames by adopting a pyramid optical flow method, and constructing a non-isotropic weighting matrix solution equation when optical flow displacement is calculated to obtain an initial displacement vector field; an outlier vector in the initial displacement vector field is recognized and corrected, and a corrected velocity field is obtained through weighted average of neighborhood non-outlier vectors; establishing a coupling state space model for the feature points, and solving by taking the corrected velocity field as input to obtain a velocity field considering the blood flow conduction effect; and combining the normal component of the velocity field and the local diameter of the corresponding blood vessel to obtain the coronary artery blood flow.

Description

Technical Field

[0001] This application belongs to the field of blood flow monitoring, and in particular relates to an automatic monitoring method and system for coronary artery blood flow. Background Technology

[0002] Currently, standard clinical methods for assessing coronary artery blood flow function, such as fractional flow reserve (FFR) and instantaneous flow rate without waveform (iFR), are invasive procedures requiring the use of pressure guidewires or Doppler guidewires. These procedures are not only complex and costly but also carry risks of complications such as vasospasm, perforation, or dissection, limiting their widespread application in routine screening and repeated examinations. Blood flow quantification based on X-ray coronary angiography (XCA) image sequences offers a novel approach. Optical flow methods estimate the velocity field by analyzing the motion of pixels in the image sequence, providing a computational basis for quantifying blood flow velocity. Compared to traditional density measurement methods based on contrast agent filling time, such as the TIMI frame rate method, it provides richer and more refined hemodynamic information. Feature point selection within the vascular region is fundamental to subsequent tracking accuracy. Due to image noise, vessel wall pulsation, and the influence of non-vascular structures, randomly selected feature points or those based on simple gradient responses may have poor stability and are prone to drift or loss during tracking, leading to deviations in the velocity field. Furthermore, classic optical flow algorithms, such as the Lucas-Kanade method, are typically based on the isotropic motion assumption, meaning that all pixels in the neighborhood contribute equally to the optical flow calculation. This assumption does not match the actual situation of coronary blood flow, where blood mainly moves at high speed along the vessel centerline, exhibiting significant anisotropic characteristics. Traditional algorithms struggle to capture this constrained flow field pattern. Influenced by factors such as image quality, respiratory motion, and cardiac pulsation, optical flow calculations inevitably generate outlier velocity vectors. If these erroneous vectors are not processed, they will severely contaminate the overall velocity field, leading to distorted blood flow calculation results. Most existing methods estimate the velocity of feature points independently, ignoring the physical nature of blood as a fluid transported within the vessel—that is, the upstream blood flow state affects the downstream. They fail to utilize this spatiotemporal coupling relationship to constrain and correct the velocity field, limiting the physiological realism and accuracy of the calculation results. Summary of the Invention

[0003] This invention addresses the problem of inaccurate results in automated coronary artery blood flow monitoring by proposing an automated method for monitoring coronary artery blood flow, comprising the following steps: A sequence of coronary angiography images is acquired. Blood vessels are segmented in each frame of the image, and the blood vessel region mask, blood vessel centerline, and local diameter of the blood vessel at each point are extracted. Candidate feature points are extracted within the blood vessel region of the first frame image. Points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range are selected as feature points to be tracked. For the feature points to be tracked, the pyramid optical flow method is used for tracking between adjacent frames of the image sequence. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. Identify outlier vectors in the initial velocity field, i.e. vectors whose directional angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range, and replace the outlier vectors with the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field; A coupled state space model is established for each feature point to be tracked, and the modified velocity field is used as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the vessel centerline at the previous moment. The velocity field considering the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the vessel at the corresponding position, the cross-sectional area of ​​the vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

[0004] Optionally, the step of extracting candidate feature points within the blood vessel region of the first frame image and selecting points whose normal distance to the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as feature points to be tracked includes: The first distance threshold is a preset multiple of the local diameter of the blood vessel; The energy entropy of the local image region is expressed by the formula... Calculation, where The probability of gray level i appearing within the region is given by the preset entropy range, which is used to select points containing significant structural or texture information.

[0005] Optionally, constructing the non-isotropic weighted matrix includes: The weights of the anisotropic weighted matrix are determined by a two-dimensional Gaussian function. The standard deviation of the two-dimensional Gaussian function in the local tangent direction along the blood vessel centerline is greater than its standard deviation in the normal direction. When solving the optical flow equation, pixel displacement information along the blood flow direction is given priority.

[0006] Optionally, identifying outlier vectors in the initial velocity field includes: By determining whether any displacement vector deviates significantly from the average direction and average magnitude of other displacement vectors in the spatial neighborhood; The replacement is performed using the inverse distance weighting method, which utilizes non-outlier vectors within the neighborhood for interpolation replacement.

[0007] Optionally, establishing a coupled state-space model for each feature point to be tracked includes: Define the state vector of the i-th feature point at time k. This includes its position and velocity information; Establish the state transition equation: Where A and B are state transition matrices, This represents the state of nearby feature points upstream of the vessel centerline. This is process noise; Establish the observation equation: ,in Let H be the velocity observations from the corrected velocity field, and H be the observation matrix. To observe noise; The coupled state-space model is solved using a Kalman filter.

[0008] Optionally, the calculation of coronary blood flow includes: At the location of each feature point to be tracked, the local cross-sectional area A of the blood vessel is calculated using the local diameter D of the blood vessel; The velocity vector in the velocity field is decomposed to obtain the velocity component perpendicular to the cross-section. ; Through formula Calculate instantaneous blood flow; The average coronary blood flow is obtained by integrating the instantaneous blood flow over time within a cardiac cycle and then dividing it by the duration of the cardiac cycle.

[0009] In addition, this application also proposes an automatic monitoring system for coronary artery blood flow, comprising the following modules: The feature point determination module is used to acquire a sequence of coronary angiography images, segment the blood vessels in each frame, extract the blood vessel region mask, the blood vessel centerline, and the local diameter of the blood vessel at each point; extract candidate feature points within the blood vessel region of the first frame image, and select points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as the feature points to be tracked. The matrix construction module is used to track the feature points to be tracked between adjacent frames of the image sequence using the pyramid optical flow method. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. The vector recognition module is used to identify outlier vectors in the initial velocity field, namely vectors whose direction angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range. The outlier vectors are replaced by the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field. The model building and flow calculation module is used to establish a coupled state space model for each feature point to be tracked, and to take the modified velocity field as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the center line of the blood vessel at the previous moment. The velocity field taking into account the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the blood vessel at the corresponding position, the cross-sectional area of ​​the blood vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

[0010] Optionally, the step of extracting candidate feature points within the blood vessel region of the first frame image and selecting points whose normal distance to the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as feature points to be tracked includes: The first distance threshold is a preset multiple of the local diameter of the blood vessel; The energy entropy of the local image region is expressed by the formula... Calculation, where The probability of gray level i appearing within the region is given by the preset entropy range, which is used to select points containing significant structural or texture information.

[0011] Optionally, constructing the non-isotropic weighted matrix includes: The weights of the anisotropic weighted matrix are determined by a two-dimensional Gaussian function. The standard deviation of the two-dimensional Gaussian function in the local tangent direction along the blood vessel centerline is greater than its standard deviation in the normal direction. When solving the optical flow equation, pixel displacement information along the blood flow direction is given priority.

[0012] Optionally, identifying outlier vectors in the initial velocity field includes: By determining whether any displacement vector deviates significantly from the average direction and average magnitude of other displacement vectors in the spatial neighborhood; The replacement is performed using the inverse distance weighting method, which utilizes non-outlier vectors within the neighborhood for interpolation replacement.

[0013] Optionally, establishing a coupled state-space model for each feature point to be tracked includes: Define the state vector of the i-th feature point at time k. This includes its position and velocity information; Establish the state transition equation: Where A and B are state transition matrices, This represents the state of nearby feature points upstream of the vessel centerline. This is process noise; Establish the observation equation: ,in Let H be the velocity observations from the corrected velocity field, and H be the observation matrix. To observe noise; The coupled state-space model is solved using a Kalman filter.

[0014] Optionally, the calculation of coronary blood flow includes: At the location of each feature point to be tracked, the local cross-sectional area A of the blood vessel is calculated using the local diameter D of the blood vessel; The velocity vector in the velocity field is decomposed to obtain the velocity component perpendicular to the cross-section. ; Through formula Calculate instantaneous blood flow; The average coronary blood flow is obtained by integrating the instantaneous blood flow over time within a cardiac cycle and then dividing it by the duration of the cardiac cycle.

[0015] This invention ensures the stability and representativeness of tracking points by combining energy entropy and normal distance to screen feature points. Furthermore, in optical flow calculation, an anisotropic weighted matrix is ​​used to assign higher weights to pixels along the vascular tangent, making velocity estimation more closely reflect the physical nature of blood flow along the vessel and improving the accuracy of the initial velocity field. By effectively identifying and correcting outlier vectors, the robustness of the velocity field to noise and artifacts is enhanced. Using a coupled state-space model, the velocity estimation of each feature point is correlated with the upstream blood flow state, constructing a globally consistent velocity field that considers blood flow conduction effects. This overcomes the limitations of independent calculation at each point in traditional methods, resulting in higher realism and reliability of the final calculated coronary artery blood flow. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of blood vessel segmentation and geometric parameter extraction. Figure 2 A schematic diagram illustrating the selection of feature points to be tracked; Figure 3 This is a schematic diagram of the non-isotropic weighted optical flow method; Figure 4 A schematic diagram for outlier vector identification and correction; Figure 5 This is a schematic diagram of a coupled state-space model; Figure 6 A schematic diagram for calculating coronary artery blood flow. Detailed Implementation

[0017] To facilitate understanding of the technical solution of this application, the application will be further described below with reference to the accompanying drawings.

[0018] The terms "first" and "second," etc., used in the specification, claims, and drawings of this application are used only to distinguish different objects and not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0019] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0020] In the first embodiment, this application proposes an automatic monitoring method for coronary artery blood flow, comprising the following steps: S1, acquire the coronary angiography image sequence, segment the blood vessels in each frame of the image, and extract the blood vessel region mask, the blood vessel centerline, and the local diameter of the blood vessel at each point; extract candidate feature points in the blood vessel region of the first frame of the image, and select the points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as the feature points to be tracked. The input X-ray coronary angiography image is segmented end-to-end at the pixel level using a deep learning network model such as U-Net to generate a binarized vascular region mask. A skeleton extraction algorithm, such as a morphological thinning algorithm, is applied to the vascular region mask to obtain a single-pixel-width vascular centerline. For each point on the centerline, a search is performed within the vascular region mask along its normal direction, and the Euclidean distance between the two boundary points is measured as the local diameter of the vascular at that point. Figure 1 As shown.

[0021] The Shi-Tomasi corner detection or Harris corner detection algorithm is used to extract candidate feature points with strong gradients within the blood vessel mask region of the first frame. For each candidate feature point, the nearest Euclidean distance to the center line of the blood vessel is calculated as the normal distance. A first distance threshold is set to one-quarter of the average diameter of the blood vessel to eliminate points near or outside the vessel wall. An image block of, for example, 15x15 pixels is taken centered on this point, and the gray-level histogram of the image block is calculated. Based on the gray-level histogram, the energy entropy is calculated using the Shannon entropy formula. Points with entropy values ​​between 4.5 and 7.0 are retained, and regions with overly simple or complex textures are filtered out. The resulting set of points is the feature point to be tracked.

[0022] S2, For the feature points to be tracked, the pyramid optical flow method is used for tracking between adjacent frames of the image sequence. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. In the standard pyramid Lucas-Kanade optical flow framework, a two-dimensional Gaussian weighted matrix is ​​constructed for a local window around each tracking point. The covariance matrix of the Gaussian function is anisotropic, with its principal axis aligned with the local tangent direction of the blood vessel centerline corresponding to the tracking feature point. Furthermore, the standard deviation along the tangent direction is much smaller than the standard deviation along the normal direction, for example, the standard deviation along the tangent is 1 pixel while that along the normal is 5 pixels. This results in pixels along the blood flow direction receiving higher weights when solving the optical flow equation. The weighted least squares problem obtains the optical flow displacement by solving an equation of the form A transpose WAx equal to A transpose Wb. In one embodiment, higher weight means that the weight of pixels along the local tangent direction of the blood vessel centerline is greater than the weight of pixels not along the local tangent direction of the blood vessel centerline.

[0023] S3, identify outlier vectors in the initial velocity field, i.e. vectors whose direction angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range, and replace the outlier vectors with the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field; For each displacement vector in the initial velocity field, its spatial neighborhood is defined as the 10 displacement vectors closest to it in Euclidean distance. The median of the direction and magnitude of all vectors in the spatial neighborhood is calculated. If the angle between the direction of the current vector and the median direction is greater than 30 degrees, or the ratio of its magnitude to the median magnitude is greater than 2.0 or less than 0.5, it is determined to be an outlier vector. For all identified outliers, their original values ​​are replaced with the inverse distance weighted average of all non-outliers in their neighborhood to complete the correction of the velocity field.

[0024] S4. Establish a coupled state space model for each feature point to be tracked, and use the modified velocity field as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the vessel centerline at the previous moment. The velocity field considering the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the vessel at the corresponding position, the cross-sectional area of ​​the vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

[0025] A linear dynamic system is constructed, where the two-dimensional velocity vector of each feature point serves as its state variable. The state transition equation is defined as follows: the velocity state of point i at the current moment is the weighted sum of its own velocity state at the previous moment and the velocity states of all neighboring feature points j within a certain upstream distance along the vessel centerline at the previous moment, with the weights inversely proportional to the upstream distance. The observation equation uses the velocity vector of the corresponding point in the corrected velocity field as the observation value of the current state. By recursively solving the coupled state-space model using a Kalman filter, forward filtering and backward smoothing are performed on the entire image sequence to obtain the optimal state estimate of each feature point at each moment, i.e., the velocity field.

[0026] The cross-sectional area of ​​the blood vessel at any point on the centerline is calculated based on the local diameter of the blood vessel at that corresponding location, using the following formula: D is the diameter; obtain the velocity vector V of this point in the velocity field, and calculate the projection component of the velocity vector along the tangent direction of the blood vessel centerline at this point, that is, the effective velocity component perpendicular to the cross-section. The instantaneous blood flow Q at that point is expressed by the formula. The final average coronary blood flow can be obtained by integrating or averaging the instantaneous blood flow over one or more cardiac cycles.

[0027] In an optional embodiment, the step of extracting candidate feature points within the blood vessel region of the first frame image and selecting points whose normal distance to the blood vessel centerline is less than a first distance threshold and whose local image region's energy entropy is within a preset entropy value range as feature points to be tracked includes: The first distance threshold is a preset multiple of the local diameter of the blood vessel; The energy entropy of the local image region is expressed by the formula... Calculation, where The probability of gray level i appearing within the region is given by the preset entropy range, which is used to select points containing significant structural and / or texture information.

[0028] Setting the first distance threshold to a preset multiple of the local diameter of the blood vessel ensures that the distribution of feature points within the vessel is both uniform and representative. For example, if the local diameter of a coronary artery is measured to be 30 pixels, the preset multiple can be set to 0.4, requiring that the distance between any two feature points to be tracked is no less than 12 pixels. This avoids feature points being too concentrated in a small area, which could lead to local deviations in velocity field calculations; and it also prevents feature points from being too sparse, making it impossible to accurately capture the subtle changes in blood flow.

[0029] Energy entropy is used to filter feature points and select regions rich in image information. Energy entropy measures the complexity of the gray-level distribution in a local image region. For example, in a 16×16 pixel local image region, if the region is a uniform blood cavity with minimal gray-level variation, the calculated energy entropy may be lower than 2.5. Conversely, if the region contains blood vessel walls, calcified plaques, or stent structures, its gray-level distribution is complex, and the energy entropy may be higher than 4.0. Therefore, a preset entropy range, such as 2.8-5.5, can be set to select feature points that are neither pure noise nor overly smooth, but truly contain traceable structural information, such as... Figure 2 As shown, this improves the stability and accuracy of subsequent motion estimation.

[0030] In an optional embodiment, constructing the non-isotropic weighted matrix includes: The weights of the anisotropic weighted matrix are determined by a two-dimensional Gaussian function. The standard deviation of this Gaussian function along the local tangent direction of the vessel centerline is greater than its standard deviation along the normal direction. When solving the optical flow equation, pixel displacement information along the blood flow direction is given priority. Constructing the anisotropic weighted matrix makes the optical flow calculation more consistent with the physical characteristics of blood flow, i.e., blood mainly flows along the axial direction of the vessel. In the traditional Lucas-Kanade optical flow method, a circular neighborhood weight is usually used, assuming that pixels in all directions contribute equally to the center point displacement. This embodiment corrects this assumption by introducing an elliptical two-dimensional Gaussian function as the weight. For example, in a straight segment of a vessel, the local tangent direction of its centerline is horizontal. In this case, the standard deviation of the two-dimensional Gaussian function in the horizontal direction can be set to 2.5, and the standard deviation in the vertical direction (normal direction) can be set to 0.8. Figure 3 As shown. Furthermore, when calculating the displacement of the center point, neighboring pixels in the horizontal direction are assigned a much greater weight than pixels in the vertical direction. This suppresses spurious motion perpendicular to the blood flow direction caused by image noise and / or minute pulsations of the blood vessel wall, resulting in a more accurate velocity vector aligned with the true blood flow direction.

[0031] In an optional embodiment, identifying outlier vectors in the initial velocity field includes: By determining whether any displacement vector deviates significantly from the average direction and average magnitude of other displacement vectors in the spatial neighborhood; The replacement is performed using the inverse distance weighting method, which utilizes non-outlier vectors within the neighborhood for interpolation replacement.

[0032] Identifying and replacing outlier vectors in the initial velocity field ensures the smoothness and physical realism of the blood flow velocity field. For example, within a 5×5 local region containing 25 velocity vectors, the average direction and average magnitude of the remaining 24 vectors (excluding the center vector) are calculated. Assume the calculated average direction is 45 degrees and the average magnitude is 10 pixels per frame. If the direction of the center vector under test is 150 degrees and the magnitude is 2 pixels per frame, showing a significant difference from the average of the neighborhood, then the center vector under test will be identified as an outlier. This significant deviation can be quantified by setting a threshold, such as an angle deviation exceeding 40 degrees or a magnitude difference exceeding 50 percentage points of the average magnitude.

[0033] Once a vector is identified as an outlier, it is corrected using inverse distance weighting. Continuing the example above, all non-outlier normal vectors are searched within its neighborhood. Assume the nearest normal vector A is 3 pixels away, with a velocity value of 9.5 pixels per frame and an orientation of 48 degrees; another, more distant normal vector B is 6 pixels away, with a velocity value of 10.5 pixels per frame and an orientation of 42 degrees. According to inverse distance weighting, the greater the distance, the smaller the weight; vector A will have a much larger weight than vector B. By weighted averaging these neighboring normal vectors, a new velocity vector consistent with the surrounding blood flow pattern can be interpolated to the outlier location, eliminating erroneous measurements. Figure 4 As shown.

[0034] In an optional embodiment, establishing a coupled state-space model for each feature point to be tracked includes: Define the state vector of the i-th feature point at time k. This includes its position and velocity information; Establish the state transition equation: Where A and B are state transition matrices, This represents the state of nearby feature points upstream of the vessel centerline. This is process noise; Establish the observation equation: ,in Let H be the velocity observations from the corrected velocity field, and H be the observation matrix. To observe noise; The coupled state-space model is solved using a Kalman filter.

[0035] Establishing a coupled state-space model for each feature point is an effective way to incorporate the continuity and spatiotemporal correlation of blood flow into the tracking process. Traditional Kalman filtering, when tracking a single point, only considers the point's state at the previous moment during the state transition. In this method, the coupling term refers to the state of the upstream neighboring feature points. This introduces physical constraints on blood flow in space. For example, for a feature point i in a blood vessel, its motion depends not only on its own inertia but also on the strong influence of the motion state of its immediate upstream feature point i-1 at the previous moment, because blood flows continuously. The role of the state transition matrix B is to quantify this pushing or pulling effect of upstream blood flow on downstream.

[0036] At each time step, the filter first uses the state transition equation to predict the prior position and velocity of feature point i at the current time k. For example, based on the velocity of the previous frame and the state of the upstream point, it predicts that the point will move to a new position. An actual observation of the velocity at that point is then obtained from the corrected optical flow field. ,like Figure 5 As shown, the Kalman filter combines the predicted value with this noisy observation to provide a posterior optimal estimate. If a feature point is temporarily unobservable due to occlusion by the blood vessel wall, the filter can still provide a reasonable prediction based solely on the state transition equation until the point becomes visible again, achieving smooth and continuous tracking.

[0037] In an optional embodiment, the calculation of coronary blood flow includes: At the location of each feature point to be tracked, the local cross-sectional area A of the blood vessel is calculated using the local diameter D of the blood vessel; The velocity vector in the velocity field is decomposed to obtain the velocity component perpendicular to the cross-section. ; Through formula Calculate instantaneous blood flow; The average coronary blood flow is obtained by integrating the instantaneous blood flow over time within a cardiac cycle and then dividing it by the duration of the cardiac cycle.

[0038] This embodiment provides a method for quantitatively assessing blood flow based on a two-dimensional angiography image sequence. Specifically, at each stably tracked feature point, the local cross-sectional area A is calculated using an area formula, assuming the blood vessel cross-section is circular, based on the measured local diameter D of the blood vessel in the image. For example, if the local diameter of a blood vessel is 3.0 mm, its cross-sectional area A is approximately 7.07 square millimeters.

[0039] Because the direction of the blood flow velocity vector V may have a slight angle with the normal direction of the blood vessel cross-section, velocity decomposition is necessary. The projected component of the velocity vector V in the direction perpendicular to the cross-section is then calculated. This is the velocity that represents the effective throughput. For example, if the final velocity V is 150 millimeters per second and its direction makes an angle of 10 degrees with the normal direction of the cross section, then the effective velocity component... Approximately 147.7 millimeters per second, such as Figure 6 As shown. Multiplying this velocity by the area A yields the instantaneous blood flow at that location at that moment. The unit is milliliters per second. Within a complete cardiac cycle, for example, spanning 40 frames, 40 instantaneous blood flow values ​​are obtained. Integrating these values ​​over time and dividing by the total duration of the cardiac cycle yields a stable, clinically diagnostic average coronary blood flow value, such as 90 milliliters per minute.

[0040] In a second embodiment, this application proposes an automatic monitoring system for coronary artery blood flow, comprising the following modules: The feature point determination module is used to acquire a sequence of coronary angiography images, segment the blood vessels in each frame, extract the blood vessel region mask, the blood vessel centerline, and the local diameter of the blood vessel at each point; extract candidate feature points within the blood vessel region of the first frame image, and select points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as the feature points to be tracked. The matrix construction module is used to track the feature points to be tracked between adjacent frames of the image sequence using the pyramid optical flow method. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. The vector recognition module is used to identify outlier vectors in the initial velocity field, namely vectors whose direction angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range. The outlier vectors are replaced by the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field. The model building and flow calculation module is used to establish a coupled state space model for each feature point to be tracked, and to take the modified velocity field as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the center line of the blood vessel at the previous moment. The velocity field taking into account the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the blood vessel at the corresponding position, the cross-sectional area of ​​the blood vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

[0041] The embodiments described in this application are described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0042] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0043] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0044] This application can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0045] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0046] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. The protection scope of this application should be determined by the protection scope of the claims.

Claims

1. An automatic monitoring method for coronary artery blood flow, characterized in that, Includes the following steps: A sequence of coronary angiography images is acquired. Blood vessels are segmented in each frame of the image, and the blood vessel region mask, blood vessel centerline, and local diameter of the blood vessel at each point are extracted. Candidate feature points are extracted within the blood vessel region of the first frame image. Points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range are selected as feature points to be tracked. For the feature points to be tracked, the pyramid optical flow method is used for tracking between adjacent frames of the image sequence. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. Identify outlier vectors in the initial velocity field, i.e. vectors whose directional angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range, and replace the outlier vectors with the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field; A coupled state space model is established for each feature point to be tracked, and the modified velocity field is used as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the vessel centerline at the previous moment. The velocity field considering the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the vessel at the corresponding position, the cross-sectional area of ​​the vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

2. The method according to claim 1, characterized in that, The process of extracting candidate feature points within the blood vessel region of the first frame image, and selecting points whose normal distance to the blood vessel centerline is less than a first distance threshold and whose local image region's energy entropy is within a preset entropy value range, as the feature points to be tracked, includes: The first distance threshold is a preset multiple of the local diameter of the blood vessel; The energy entropy of the local image region is expressed by the formula... Calculation, where The probability of gray level i appearing within the region is given by the preset entropy range, which is used to select points containing significant structural or texture information.

3. The method according to claim 1, characterized in that, The construction of the non-isotropic weighted matrix includes: The weights of the anisotropic weighted matrix are determined by a two-dimensional Gaussian function. The standard deviation of the two-dimensional Gaussian function in the local tangent direction along the blood vessel centerline is greater than its standard deviation in the normal direction. When solving the optical flow equation, pixel displacement information along the blood flow direction is given priority.

4. The method according to claim 1, characterized in that, The identification of outlier vectors in the initial velocity field includes: By determining whether any displacement vector deviates significantly from the average direction and average magnitude of other displacement vectors in the spatial neighborhood; The replacement is performed using the inverse distance weighting method, which utilizes non-outlier vectors within the neighborhood for interpolation replacement.

5. The method according to claim 1, characterized in that, The process of establishing a coupled state-space model for each feature point to be tracked includes: Define the state vector of the i-th feature point at time k. This includes its position and velocity information; Establish the state transition equation: Where A and B are state transition matrices, This represents the state of nearby feature points upstream of the vessel centerline. This is process noise; Establish the observation equation: ,in Let H be the velocity observations from the corrected velocity field, and H be the observation matrix. To observe noise; The coupled state-space model is solved using a Kalman filter.

6. The method according to claim 1, characterized in that, The calculation of coronary artery blood flow includes: At the location of each feature point to be tracked, the local cross-sectional area A of the blood vessel is calculated using the local diameter D of the blood vessel; The velocity vector in the velocity field is decomposed to obtain the velocity component perpendicular to the cross-section. ; Through formula Calculate instantaneous blood flow; The average coronary blood flow is obtained by integrating the instantaneous blood flow over time within a cardiac cycle and then dividing it by the duration of the cardiac cycle.

7. An automatic monitoring system for coronary artery blood flow, characterized in that, Includes the following modules: The feature point determination module is used to acquire a sequence of coronary angiography images, segment the blood vessels in each frame, extract the blood vessel region mask, the blood vessel centerline, and the local diameter of the blood vessel at each point; extract candidate feature points within the blood vessel region of the first frame image, and select points whose normal distance from the blood vessel centerline is less than a first distance threshold and whose energy entropy of the local image region is within a preset entropy value range as the feature points to be tracked. The matrix construction module is used to track the feature points to be tracked between adjacent frames of the image sequence using the pyramid optical flow method. When calculating the optical flow displacement, an isotropic weighted matrix is ​​constructed. The isotropic weighted matrix assigns higher weights to pixels along the local tangent direction of the blood vessel centerline. The optical flow equation is solved using the isotropic weighted matrix to obtain the displacement vector that constitutes the initial velocity field. The vector recognition module is used to identify outlier vectors in the initial velocity field, namely vectors whose direction angle or magnitude ratio with other displacement vectors in the spatial neighborhood exceeds a preset range. The outlier vectors are replaced by the weighted average of the non-outlier vectors in the neighborhood to obtain the corrected velocity field. The model building and flow calculation module is used to establish a coupled state space model for each feature point to be tracked, and to take the modified velocity field as the observation input. The state of any feature point at the current moment depends on its own state and the state of its neighboring feature points upstream along the center line of the blood vessel at the previous moment. The velocity field taking into account the blood flow conduction effect is obtained by solving the coupled state space model. Based on the velocity field and the local diameter of the blood vessel at the corresponding position, the cross-sectional area of ​​the blood vessel is calculated. Combined with the velocity component perpendicular to the cross-section, the coronary artery blood flow is calculated.

8. The system according to claim 7, characterized in that, The process of extracting candidate feature points within the blood vessel region of the first frame image, and selecting points whose normal distance to the blood vessel centerline is less than a first distance threshold and whose local image region's energy entropy is within a preset entropy value range, as the feature points to be tracked, includes: The first distance threshold is a preset multiple of the local diameter of the blood vessel; The energy entropy of the local image region is expressed by the formula... Calculation, where The probability of gray level i appearing within the region is given by the preset entropy range, which is used to select points containing significant structural or texture information.

9. The system according to claim 7, characterized in that, The construction of the non-isotropic weighted matrix includes: The weights of the anisotropic weighted matrix are determined by a two-dimensional Gaussian function. The standard deviation of the two-dimensional Gaussian function in the local tangent direction along the blood vessel centerline is greater than its standard deviation in the normal direction. When solving the optical flow equation, pixel displacement information along the blood flow direction is given priority.

10. The system according to claim 7, characterized in that, The identification of outlier vectors in the initial velocity field includes: By determining whether any displacement vector deviates significantly from the average direction and average magnitude of other displacement vectors in the spatial neighborhood; The replacement is performed using the inverse distance weighting method, which utilizes non-outlier vectors within the neighborhood for interpolation replacement.

Images