A non-invasive coronary blood flow evaluation method and system based on multi-phase CTA and bayesian inference

Abstract

The application provides a non-invasive coronary blood flow evaluation method based on multi-phase CTA and Bayesian inference, comprising the following steps: acquiring multi-phase coronary CTA images of a target to be evaluated, and extracting coronary vessel deformation waveform data from the multi-phase coronary CTA images; generating posterior probability distribution estimation results of coronary blood flow rate through a pre-established Bayesian inference framework and the coronary vessel deformation waveform data; and generating coronary blood flow rate evaluation results corresponding to the target to be evaluated according to the posterior probability distribution estimation results. The application uses a Bayesian framework to simultaneously estimate multiple physiological parameters, and identifies parameter interactions through posterior correlation analysis, thereby improving inversion accuracy.

Description

Technical Field

[0001] This application relates to the technical field of non-invasive coronary blood flow assessment, specifically to a non-invasive coronary blood flow assessment method and system based on multi-phase CTA and Bayesian inference. Background Technology

[0002] Coronary artery blood flow velocity directly reflects myocardial perfusion and coronary circulation function. Abnormal changes in it are not only a direct consequence of coronary artery stenosis leading to restricted blood flow, but also closely related to pathological conditions such as microvascular dysfunction. Accurate assessment of coronary artery blood flow velocity is of great significance for realizing non-invasive functional assessment based on coronary computed tomography angiography (CTA).

[0003] Conventional single-phase CTA, as a mature non-invasive anatomical imaging technique, can clearly present the degree of stenosis and morphological characteristics of blood vessels, but it cannot provide hemodynamic functional information. In clinical practice, many hemodynamic abnormalities (such as microvascular dysfunction) are easily missed in simple morphological assessments, leading to delays in clinical intervention. To overcome this deficiency, multi-phase CTA data obtained using retrospective ECG gating technology, which tracks the dynamic changes in the cross-sectional area of ​​the coronary artery lumen during the cardiac cycle to invert blood flow velocity, has become an important research direction for non-invasive blood flow function assessment. This method is based on the laws of conservation of mass and momentum, arguing that the periodic changes in the cross-sectional area of ​​the vessel directly reflect the changes in intraluminal blood flow velocity, providing a new pathway for achieving non-invasive functional assessment.

[0004] However, retrieving blood flow velocity from vascular deformation data faces significant technical challenges: this inversion process is a typical multi-parameter coupled inverse problem. Instantaneous vascular deformation is determined by various physiological and physical parameters such as blood flow velocity, intraluminal pressure, and the elastic modulus of the vessel wall. This leads to different parameter combinations potentially producing similar vascular deformation waveforms, resulting in a non-uniqueness of solutions, mathematically manifested as severe ill-posedness. Existing solutions mainly fall into two categories: one is simulation methods based on computational fluid dynamics (CFD), which rely on single-phase CTA data and empirical assumptions, resulting in high computational costs and low individualization; the other is traditional optimization inversion methods, such as gradient descent or grid search, which are prone to getting trapped in local optima and cannot quantify parameter uncertainties, leading to insufficient reliability of the results.

[0005] Therefore, there is an urgent need for a non-invasive coronary blood flow assessment method that can systematically solve the problems of parameter coupling and uncertainty quantification, so as to improve the accuracy, robustness and clinical applicability of the assessment. Summary of the Invention

[0006] In view of the aforementioned problems, this application is proposed to provide a non-invasive coronary blood flow assessment method and system based on multi-temporal CTA and Bayesian inference to overcome or at least partially solve the aforementioned problems, comprising: A non-invasive coronary blood flow assessment method based on multi-phase CTA and Bayesian inference includes the following steps: Acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; Using a pre-established Bayesian inference framework and the coronary artery deformation waveform data, the posterior probability distribution estimation result of the coronary blood flow velocity is generated. Based on the posterior probability distribution estimation results, a coronary blood flow velocity assessment result corresponding to the target to be evaluated is generated.

[0007] Furthermore, the step of acquiring multi-temporal coronary CTA images of the target to be evaluated and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images specifically includes: Acquire a sequence of continuous coronary CTA images during the cardiac cycle; Coronary CTA images within a preset motion artifact threshold are selected from the continuous coronary CTA image sequence and used as the multi-temporal coronary CTA images; Coronary vessel deformation waveform data were extracted from the multi-temporal coronary CTA images.

[0008] Furthermore, the step of extracting coronary artery deformation waveform data from the multi-temporal coronary CTA images specifically includes: A coordinate system was established, with the end-diastolic coronary CTA image as the reference frame, and the coronary CTA images of each time phase were aligned in the coordinate system. By combining coronary artery bifurcation point marking with a gray-scale mutual information maximization strategy, spatial offsets caused by cardiac pulsation and respiratory motion in coronary CTA images at different time phases are compensated. The coronary artery centerline is extracted from coronary CTA images at different time phases, and the smoothness of the centerline is optimized by confidence point detection; Nodes are set at fixed intervals along the centerline to locate vessel boundary points in coronary CTA images at various time phases; Calculate the normal cross-sectional diameter at each boundary point to generate coronary vessel deformation waveform data of the lumen cross-sectional area changing over time.

[0009] Furthermore, the step of generating a posterior probability distribution estimate of coronary blood flow velocity using a pre-established Bayesian inference framework and the coronary vessel deformation waveform data specifically includes the following steps: Set an inversion parameter vector, which includes instantaneous blood flow velocity, instantaneous intraluminal pressure, and vessel wall elastic modulus; A physical model of vascular dynamics was established based on the laws of conservation of mass and momentum. The vascular dynamics physical model is solved numerically, and the predicted cross-sectional area time series is output. The coronary artery deformation waveform data is used as observation data. The probability of the observation data is quantified when given inversion parameters, and the posterior probability distribution estimation result is generated.

[0010] Furthermore, the step of establishing a vascular dynamics physical model based on the laws of conservation of mass and momentum specifically includes the following steps: The inlet boundary of the vascular dynamics physical model is set to a known flow rate time series, and the outlet boundary is set to a known pressure time series. Based on clinical coronary patient data, a time-series marginal prior distribution of instantaneous blood flow velocity and instantaneous intravascular pressure was constructed using the kernel density estimation method; Based on typical values ​​of coronary artery elastic modulus in the literature, a uniform distribution of vessel wall elastic modulus was established. We define the conjugate weak information inverse Gamma distribution of the observation noise variance and update it using Gibbs sampling.

[0011] Furthermore, the step of using the coronary artery deformation waveform data as observation data, quantifying the probability of the observation data given the inversion parameters, and generating the posterior probability distribution estimation result specifically includes the following steps: Construct the joint posterior probability distribution of the inversion parameter vector and the noise variance according to Bayes' theorem; The joint posterior probability distribution is approximated, and a Gaussian process surrogate model is constructed during the approximation process. The Gaussian process surrogate model is used to simulate the physical model of vascular dynamics. Perform a preset number of sampling iterations, discard the early combustion period samples, and output the posterior probability distribution sample chain; The convergence of the sample chain is tested to generate the posterior probability distribution estimation result.

[0012] Further, the step of generating a coronary blood flow velocity assessment result corresponding to the target to be assessed based on the posterior probability distribution estimation result specifically includes the following steps: The mean of the sample chain of the posterior probability distribution is used as the point estimate of the blood flow velocity to minimize the mean square error. Based on the posterior probability distribution, the highest posterior density confidence interval of blood flow velocity at each time point is calculated. The evaluation results are output, which include blood flow rate time series, uncertainty interval, and statistical information of key parameters.

[0013] A non-invasive coronary blood flow assessment system based on multi-phase CTA and Bayesian inference includes: The data acquisition module is used to acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; The sampling estimation module is used to generate a posterior probability distribution estimation result of coronary blood flow velocity using a pre-established Bayesian inference framework and the coronary vessel deformation waveform data. The result output module is used to generate a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

[0014] A computer electronic device includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When executed by the processor, the computer program implements the steps of a non-invasive coronary blood flow assessment method based on multi-phase CTA and Bayesian inference as described above.

[0015] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference as described above.

[0016] This application has the following advantages: In the embodiments of this application, considering that the inversion process in the prior art is a typical multi-parameter coupled inverse problem, the instantaneous deformation of blood vessels is jointly determined by various physiological and physical parameters such as blood flow velocity, intraluminal pressure, and elastic modulus of the blood vessel wall. This leads to the possibility that different parameter combinations may produce similar blood vessel deformation waveforms, resulting in a non-uniqueness problem of solutions. This application provides a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference, including the following steps: acquiring multi-temporal coronary CTA images of the target to be assessed, and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images; generating a posterior probability distribution estimation result of coronary blood flow velocity through a pre-established Bayesian inference framework and the coronary vessel deformation waveform data; and generating a coronary blood flow velocity assessment result corresponding to the target to be assessed based on the posterior probability distribution estimation result. By constructing a Bayesian inference framework, unknown parameters such as blood flow velocity, intraluminal pressure, and vessel wall elastic modulus are treated as random variables, and a joint posterior probability distribution is built. This solves the problem of non-uniqueness of inversion solutions caused by multi-parameter coupling, achieving the effect of systematically quantifying parameter uncertainty and allowing doctors to intuitively assess diagnostic confidence. Through multi-temporal CTA image preprocessing, dynamic vascular deformation waveform data within the cardiac cycle are extracted, eliminating the need for invasive monitoring data. This addresses the problem that single-temporal CTA can only provide static anatomical information and that current technologies rely on empirical assumptions, achieving the effect of non-invasive acquisition of individualized hemodynamic functional information, reducing patient risk and medical costs. Uncertainty quantification analysis outputs point estimates of blood flow velocity and the highest posterior density confidence interval, solving... Traditional methods only provide single numerical values ​​and lack reliable references. This approach achieves the effect of providing doctors with comprehensive assessment basis and optimizing clinical intervention decisions for coronary artery disease. By adopting the MCMC sampling strategy combined with the Gaussian process surrogate model, the parameter space is explored globally and the number of calls to expensive physical models is reduced. This solves the problems of traditional optimization algorithms being prone to getting trapped in local optima and having high computational costs. This achieves stable and reliable inversion results, improved computational efficiency, and meets the needs of real-time clinical decision-making. By integrating data-driven prior distribution settings with vascular dynamics physical models, the results are ensured to have both physical rationality and clinical applicability. This solves the problem of individualized assessment bias caused by the simplification of assumptions in existing technologies and achieves the effect of improving the accuracy and clinical applicability of coronary blood flow velocity assessment. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, the drawings used in the description of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1This is a flowchart of a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference provided in one embodiment of this application; Figure 2 This is a diagram showing the results of feasibility verification through numerical simulation and retrospective analysis of clinical data, provided in one embodiment of this application. Figure 3 This is a structural block diagram of a non-invasive coronary blood flow assessment system based on multi-temporal CTA and Bayesian inference provided in one embodiment of this application; Figure 4 This is a schematic diagram of the structure of a computer electronic device provided in an embodiment of this application; 1. Computer electronic device; 2. External device; 3. Processing unit; 4. Bus; 5. Network adapter; 6. I / O interface; 7. Display; 8. Memory; 9. Random access memory; 10. Cache memory; 11. Storage system; 12. Program / utility; 13. Program module. Detailed Implementation

[0019] To make the objectives, features, and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] The inventors, through analysis of existing technologies, discovered the following: 1. The process of inverting blood flow velocity from vascular deformation data is a typical multi-parameter coupled inverse problem. Instantaneous vascular deformation is jointly determined by various physiological and physical parameters such as blood flow velocity, intraluminal pressure, and vascular wall elastic modulus. This leads to different parameter combinations potentially producing similar vascular deformation waveforms, resulting in a non-uniqueness of solutions. 2. Computational fluid dynamics (CFD)-based simulation methods rely on single-phase CTA data and empirical assumptions, resulting in high computational costs and low individualization. 3. Traditional optimization inversion methods, such as gradient descent or grid search, are prone to getting trapped in local optima and cannot quantify parameter uncertainties, leading to insufficient reliability of results.

[0021] Reference Figure 1 This paper illustrates a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference, according to an embodiment of this application, including the following steps: S110. Acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; S120. Using the pre-established Bayesian inference framework and the coronary artery deformation waveform data, generate the posterior probability distribution estimation result of the coronary blood flow velocity. S130. Generate a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

[0022] In the embodiments of this application, the inversion process in the prior art is a typical multi-parameter coupled inverse problem. The instantaneous deformation of blood vessels is determined by a variety of physiological and physical parameters such as blood flow rate, intraluminal pressure, and elastic modulus of the blood vessel wall. This results in different combinations of parameters producing similar blood vessel deformation waveforms, leading to a non-uniqueness problem of solutions. This application provides a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference. By constructing a Bayesian inference framework, unknown parameters such as blood flow velocity, intravascular pressure, and vessel wall elastic modulus are treated as random variables, and a joint posterior probability distribution is built. This solves the problem of non-uniqueness of inversion solutions caused by multi-parameter coupling, achieving the effect of systematically quantifying parameter uncertainty and allowing doctors to intuitively assess diagnostic confidence. By extracting vascular dynamic deformation waveform data within the cardiac cycle through multi-temporal CTA image preprocessing, without relying on invasive monitoring data, it solves the problem that single-temporal CTA can only provide static anatomical information and that existing technologies rely on empirical assumptions. This achieves the effect of non-invasively obtaining individualized hemodynamic functional information, reducing patient risk and medical costs. By outputting blood flow velocity point estimates and highest posterior density confidence intervals through uncertainty quantification analysis, it solves the problem that traditional methods only provide single values ​​and lack reliable references, achieving the effect of providing doctors with comprehensive assessment basis and optimizing clinical intervention decisions for coronary heart disease.

[0023] The following will further explain a non-invasive coronary blood flow assessment method based on multi-phase CTA and Bayesian inference in this exemplary embodiment.

[0024] In one embodiment of this application, the specific process of "acquiring multi-temporal coronary CTA images of the target to be evaluated and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images" described in step S110 can be further explained in conjunction with the following description.

[0025] As described in the following steps Acquire a sequence of continuous coronary CTA images during the cardiac cycle; Coronary CTA images within a preset motion artifact threshold are selected from the continuous coronary CTA image sequence and used as the multi-temporal coronary CTA images; Coronary vessel deformation waveform data were extracted from the multi-temporal coronary CTA images.

[0026] It should be noted that multiphase coronary CTA images refer to a sequence of CTA images acquired continuously within a single cardiac cycle using retrospective ECG gating technology, comprising multiple phases (e.g., at 10% RR intervals). Coronary vessel deformation waveform data specifically refers to the curve A(t) extracted from this image sequence, reflecting the periodic change of the cross-sectional area of ​​the coronary vessel lumen over time; it is the core input for subsequent hemodynamic inversion.

[0027] As an example, acquiring continuous coronary CTA image sequences can be achieved using a 256-slice or higher spiral CT scanner; motion artifact screening can be achieved based on image sharpness evaluation functions or machine learning classifiers; and deformation waveform data extraction can be completed using commercial medical image processing software (such as Philips IntelliSpace Portal) or a self-developed image analysis module.

[0028] In one specific implementation, a Siemens dual-source CT scanner was used, employing a retrospective ECG-gated mode, to acquire 10 temporal images from 0% to 90% of the RR interval (10% interval), generating the original image sequence. Subsequently, a pre-trained convolutional neural network (CNN) model was invoked to score the quality of each frame, automatically discarding images with scores below a threshold (e.g., 0.8), resulting in a high-quality multi-temporal image set.

[0029] Next, the following processing flow is performed using an open-source toolkit (such as SimpleITK): using the end-diastolic image (usually 75% phase) as a fixed reference image, the remaining phase images are aligned with it using a B-spline non-rigid registration algorithm. During the registration process, the bifurcation points of the left anterior descending coronary artery and the circumflex artery are used as key markers, and the normalized mutual information between the two images is maximized to optimize the registration accuracy, thereby compensating for the displacement caused by heartbeat and respiratory movements.

[0030] After registration, a deep learning model based on the U-Net architecture was used to automatically segment the coronary artery tree and extract its centerline. An analysis node was set every 0.4 mm along the centerline. In each temporal image, rays were emitted along the normal direction at the nodes. The boundary points of the inner and outer walls of the blood vessel were located by analyzing the gray-level gradient peaks on the rays, and the lumen radius was calculated, thereby obtaining the cross-sectional area. Finally, the area data of all nodes in all temporal phases were integrated to form a waveform A(t) representing the deformation of the target blood vessel segment.

[0031] Specifically, the U-Net model training data comes from public datasets (such as CAD-PE) and approximately 500 anonymized coronary CTA images provided by partner hospitals. The coronary artery masks and centerlines were manually annotated by doctors as the gold standard. The data underwent standardization (window width and level adjustment, pixel size normalization) and enhancement (rotation, scaling, and noise addition). The model adopts the classic U-Net architecture. The encoder uses VGG16 convolutional layers to extract features, and the decoder recovers spatial details through upsampling and skip connections. The loss function is a weighted sum of Dice loss and cross-entropy loss. The optimizer is Adam, with an initial learning rate of 1e-4. Training was conducted for 200 epochs with a batch size of 8. The final average Dice coefficient on the independent test set exceeded 0.92, and the centerline extraction error was less than 0.3 mm.

[0032] In one embodiment of this application, the specific process of "extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images" can be further explained in conjunction with the following description.

[0033] As described in the following steps A coordinate system was established, with the end-diastolic coronary CTA image as the reference frame, and the coronary CTA images of each time phase were aligned in the coordinate system. By combining coronary artery bifurcation point marking with a gray-scale mutual information maximization strategy, spatial offsets caused by cardiac pulsation and respiratory motion in coronary CTA images at different time phases are compensated. The coronary artery centerline is extracted from coronary CTA images at different time phases, and the smoothness of the centerline is optimized by confidence point detection; Nodes are set at fixed intervals along the centerline to locate vessel boundary points in coronary CTA images at various time phases; Calculate the normal cross-sectional diameter at each boundary point to generate coronary vessel deformation waveform data of the lumen cross-sectional area changing over time.

[0034] It should be noted that the confidence point detection is a post-processing technique used to optimize the smoothness of the centerline. It identifies and removes outlier or noise points by evaluating the local geometric characteristics (such as curvature and neighborhood consistency) of each point on the centerline. Maximizing grayscale mutual information is a commonly used similarity measure in image registration, used to measure the statistical dependence between two images, and is particularly suitable for medical image registration with multimodal or contrast variations.

[0035] As an example, spatial offset compensation can also be achieved using deep learning-based registration networks (such as VoxelMorph); centerline extraction can also be achieved using traditional image morphology methods based on skeletonization; and vascular boundary points can be located using active contour models (Snake models) or level set methods.

[0036] In a specific implementation, firstly, a three-dimensional Cartesian coordinate system is established, with the end-diastolic image as a reference. The Elastix registration toolbox is used, with adaptive stochastic gradient descent as the optimizer, normalized mutual information as the similarity measure, and a multi-resolution strategy is combined to complete the accurate non-rigid registration of each phase image to the reference frame.

[0037] Subsequently, a pre-trained U-Net model was applied to segment blood vessels in the registered images of each time phase. The segmented binary mask was then thinned using a three-dimensional thinning algorithm to extract the initial centerline. The average curvature and orientation consistency of each point on the centerline within the window of 5 points before and after it were calculated. Points with abnormally steep increases in curvature (exceeding the threshold) or whose orientation was significantly inconsistent with the preceding and following segments were marked as low-confidence points and replaced by linear interpolation to obtain a smooth and continuous final centerline.

[0038] Next, a series of nodes were sampled along the smooth centerline with a step size of 0.4 mm. For each node, 120 rays were uniformly emitted from the center point outwards in a 360-degree arc along its cross-section. The Canny edge detection operator was used to find the two points with the largest gray-level gradients on each ray, corresponding to the inner and outer walls, respectively. By fitting these boundary points, the average lumen radius r at that node was obtained, and the cross-sectional area was then calculated. By traversing all time phases and all nodes, a deformation waveform matrix showing the area changing over time can be constructed, serving as direct observation data for subsequent inversion.

[0039] In one embodiment of this application, the specific process of "generating the posterior probability distribution estimation result of coronary blood flow velocity through the pre-established Bayesian inference framework and the coronary vessel deformation waveform data" in step S120 can be further explained in conjunction with the following description.

[0040] As described in the following steps Set an inversion parameter vector, which includes instantaneous blood flow velocity, instantaneous intraluminal pressure, and vessel wall elastic modulus; A physical model of vascular dynamics was established based on the laws of conservation of mass and momentum. The vascular dynamics physical model is solved numerically, and the predicted cross-sectional area time series is output. The coronary artery deformation waveform data is used as observation data. The probability of the observation data is quantified when given inversion parameters, and the posterior probability distribution estimation result is generated.

[0041] It should be noted that the Bayesian inference framework described is a complete computational architecture that combines physical models with statistical inference. Its core is to treat unknown physiological parameters as random variables and use Bayes' theorem to combine prior knowledge about the parameters with observational data to obtain the posterior probability distribution of the parameters. The estimated posterior probability distribution is not a single numerical value, but rather describes the probability distribution of all unknown parameters (such as blood flow velocity Q(t), pressure p(t), and elastic modulus E) under given data, typically represented as a chain of sampled data.

[0042] In a specific implementation, firstly, the inversion parameter vector is defined. Here, Q(t) and p(t) are discrete time series (e.g., 20 points sampled within a cardiac cycle). Based on historical clinical data (e.g., catheter measurement data from 100 patients), kernel density estimation is performed on the Q and p values ​​at each time point to form the marginal prior distributions P(Q(t)) and P(p(t)) of their time series. The prior for the elastic modulus E of the vessel wall is set as a uniform distribution U(1.0e6, 5.0e6) Pa based on literature. The prior for the observation noise variance σ² is set as an inverse Gamma distribution σ²~Inv-Gamma(a,b), where the hyperparameters a=0.5ns and b=0.5ns. Here, ns=1 represents the prior precision. This is a prior variance estimate calculated based on the residuals of the training data, where d is the parameter dimension. This setting ensures conjugacy and simplifies the posterior calculation. The noise variance is updated through Gibbs sampling. .

[0043] The vascular dynamics physical model employs a one-dimensional axisymmetric elastic thin-walled tube model combined with simplified Navier-Stokes equations. Its governing equations are derived from the conservation of mass and momentum. The inlet boundary is set as Q(t) to be inverted, and the outlet boundary is set as p(t) to be inverted. The model is solved discretically using the finite difference method, outputting the predicted cross-sectional area time series. , where x is the inversion parameter vector.

[0044] The likelihood function P(d|x,σ²) quantifies the probability of observed data d given a parameter x. The likelihood assumption is that the observation errors are independently distributed. The likelihood function assumes that the observation errors follow an independent Gaussian distribution, i.e. ,in For observation data, For the inversion parameter vector, The predicted cross-sectional area output by the physical model. Let Variance be the noise variance. According to Bayes' theorem, the posterior distribution is... The evidence term (normalization constant) is: Due to the high dimensionality of the parameter space and the difficulty in calculating the evidence terms, the posterior distribution needs to be approximated by MCMC sampling. The posterior distribution is proportional to the product of the likelihood and the prior. Since the posterior distribution is high-dimensional and non-standard, a Markov chain Monte Carlo (MCMC) method is used for sampling approximation.

[0045] In one embodiment of this application, the specific process of "establishing a physical model of vascular dynamics based on the laws of conservation of mass and momentum" can be further explained in conjunction with the following description.

[0046] As described in the following steps The inlet boundary of the vascular dynamics physical model is set to a known flow rate time series, and the outlet boundary is set to a known pressure time series. Based on clinical coronary patient data, a time-series marginal prior distribution of instantaneous blood flow velocity and instantaneous intravascular pressure was constructed using the kernel density estimation method; Based on typical values ​​of coronary artery elastic modulus in the literature, a uniform distribution of vessel wall elastic modulus was established. We define the conjugate weak information inverse Gamma distribution of the observation noise variance and update it using Gibbs sampling.

[0047] It should be noted that kernel density estimation is a nonparametric estimation method that starts from the data sample itself and does not require assumptions about specific distribution patterns. It is used to construct a smooth probability density function. Conjugate prior refers to the prior distribution having the same functional form as the likelihood function, thus ensuring that the posterior distribution also belongs to the same distribution family, greatly simplifying the calculation of Bayesian inference.

[0048] In one specific implementation, the vascular segment is modeled as a linearly elastic thin-walled cylindrical tube. Its governing equations combine the one-dimensional continuity and momentum equations of fluid mechanics with the elastic constitutive equations of the tube wall (i.e., the relationship between tube wall stress and strain). In the prior setting, Q(t) and p(t) waveforms from 100 patients are extracted from a clinical database, and for each time point... Kernel density estimation is performed using a Gaussian kernel function to obtain the prior probability density function of the parameter at that time point. This constitutes a time-varying prior distribution sequence, which reflects physiological variability better than a fixed distribution. The prior range of E is set according to coronary physiology literature. The inverse Gamma conjugate prior parameter setting of the noise variance ensures that the posterior distribution is still an inverse Gamma distribution, which facilitates direct updating via Gibbs sampling: in each MCMC iteration, given the current parameter x, a new σ² value is directly extracted from the conditional posterior distribution σ²~Inv-Gamma(a,b).

[0049] In one embodiment of this application, the specific process of "using the coronary artery deformation waveform data as observation data, quantifying the probability of the observation data when given inversion parameters, and generating the posterior probability distribution estimation result" can be further explained in conjunction with the following description.

[0050] As described in the following steps Construct the joint posterior probability distribution of the inversion parameter vector and the noise variance according to Bayes' theorem; The joint posterior probability distribution is approximated, and a Gaussian process surrogate model is constructed during the approximation process. The Gaussian process surrogate model is used to simulate the physical model of vascular dynamics. Perform a preset number of sampling iterations, discard the early combustion period samples, and output the posterior probability distribution sample chain; The convergence of the sample chain is tested to generate the posterior probability distribution estimation result.

[0051] It should be noted that the Gaussian process surrogate model is a strategy that uses a small number of high-fidelity model run results to train a fast approximation model. This approximation model can predict the model output and its uncertainty under new parameter inputs, thereby significantly reducing the number of calls to the computationally expensive original physics model during sampling. Delayed acceptance is a variant of MCMC that determines whether to accept a new proposal in two stages. The first stage uses a fast but imprecise surrogate model for screening. Only when a proposal passes the first stage is the accurate model called for the final judgment, thereby improving sampling efficiency.

[0052] In a specific implementation, MCMC sampling is first implemented using Python's emcee or PyMC3 library. To accelerate MCMC sampling, a Gaussian process (GP) surrogate model is introduced to simulate the expensive physical model, reducing computational costs. The training process of the GP model is as follows: Training data generation: Sample the parameter vector θ from the prior distribution, run the physical model to obtain the corresponding output Apred(t,θ), and construct the training set. Gaussian process modeling: Define the covariance function using the squared exponential kernel function, and set the GP prior as... Training data generation: Sample the parameter vector from the prior distribution... Run the physical model to obtain the corresponding output. Construct a training set. Gaussian process modeling: Define the covariance function using the squared exponential kernel function, and set the GP prior as... The mean function m( Set the kernel to zero, and the kernel function is: ,here, It is the signal variance. The length scale is a hyperparameter, which is estimated by maximizing the logarithmic marginal likelihood optimization function.

[0053] Subsequently, the Delayed Acceptance Metropolis-Hastings (DA-MH) MCMC algorithm is executed: 1) Initialize the chain. 2) In each iteration, generate a new parameter proposal x from the multivariate Gaussian proposal distribution. 3) First stage (Delayed Acceptance): Call the trained GP model to calculate the approximate likelihood corresponding to proposal x. Based on this, a preliminary acceptance probability is calculated. If the random number is greater than α1, the proposal is rejected directly, and the process jumps back to step 2); if it passes, the process proceeds to the second stage. 4) Second stage: The expensive original physics model is invoked to calculate the true likelihood corresponding to proposal x. The final acceptance probability is calculated based on the results of the first stage. 5) For the noise variance σ², in each iteration, based on the current parameters and residuals, new values ​​are directly drawn from its complete conditional posterior distribution using a Gibbs sampling step. The total number of iterations is set to 15,000, with the first 5,000 discarded as a "burn-in" period, and the posterior sample chain obtained from the remaining 10,000 iterations. The convergence of the chain is tested using the Geweke diagnostic method to ensure reliable sampling. Ultimately, these 10,000 samples constitute the posterior distribution. The approximation is the posterior probability distribution estimate.

[0054] In one embodiment of this application, the specific process of "generating a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result" in step S130 can be further explained in conjunction with the following description.

[0055] As described in the following steps The mean of the sample chain of the posterior probability distribution is used as the point estimate of the blood flow velocity to minimize the mean square error. Based on the posterior probability distribution, the highest posterior density confidence interval of blood flow velocity at each time point is calculated. The evaluation results are output, which include blood flow rate time series, uncertainty interval, and statistical information of key parameters.

[0056] It should be noted that the highest posterior density confidence interval is an interval estimate used in Bayesian statistics to describe parameter uncertainty. It encompasses the region with the highest probability density in the posterior distribution, similar to the confidence interval concept in frequentist economics but with a different interpretation. The point estimate is usually the posterior mean because it is the optimal estimate in the sense of minimizing the mean squared error.

[0057] As an example, point estimates can also use the posterior median or mode (MAP estimate); uncertainty intervals can also use quantile intervals (such as 2.5% to 97.5% quantiles); the evaluation results can be presented as PDF reports, interactive charts, or directly integrated into the hospital PACS system.

[0058] In one specific implementation, all samples concerning the target parameter—instantaneous blood flow velocity Q(t)—are extracted from the converged MCMC posterior sample chain. For each time point in the time series... Calculate all points at that point. The arithmetic mean of the samples is used as the final point estimate of the blood flow velocity at that moment. Next, for each ti, calculate the value at that point. The 95% highest posterior density interval for the sample: This involves finding an interval such that the area under the posterior probability density function curve within this interval is 95%, and the density at any point within the interval is no less than the density at any point outside the interval. This can be obtained numerically by estimating the kernel density of the sample. Simultaneously, point estimates and confidence intervals for instantaneous intravascular pressure p(t) and vessel wall elastic modulus E are calculated. Finally, a structured assessment report is generated, including: 1) a point estimate curve of blood flow velocity Q(t) throughout the cardiac cycle, overlaid with its 95% HPD interval band (shaded). 2) a summary statistical table of key parameters, including... (Maximum blood flow rate) (Minimum blood flow velocity), mean blood flow velocity, coronary blood flow reserve (if calculable), and their corresponding uncertainty ranges. 3) Posterior correlation matrix heatmap, showing the interdependencies among parameters such as Q, p, and E. 4) Brief clinical interpretation tips based on point estimation and uncertainty.

[0059] To verify the feasibility and effectiveness of the method in this application, tests were conducted using numerical simulation and retrospective clinical data. The test results are as follows: Figure 2 As shown. Figure 2This diagram illustrates the comparison between the estimated posterior distribution of core parameters obtained using the method described in this application and the preset "true values" on a simulated coronary artery segment. The diagram typically includes several subplots, such as: (a) a posterior distribution band plot of blood flow velocity Q(t): where the solid line represents the posterior mean (point estimate), the shaded area represents its 95% highest posterior density confidence interval, and the dashed line or stars represent the preset true value curve used in the simulation. It can be observed that the true value curve falls almost entirely within the confidence interval band and closely matches the posterior mean curve. (b) a posterior distribution histogram of key scalar parameters (such as the elastic modulus of the vessel wall E): showing its probability density distribution, with vertical dashed lines indicating the true values, which are located within the high probability density region of the distribution. (c) statistical analysis of parameter estimation errors: possibly presented in tabular or bar chart form, quantifying the deviation between the point estimates and the true values.

[0060] Specifically, in the testing of the method in this application, the preset true values ​​of all key parameters (such as blood flow velocity Q(t), intraluminal pressure p(t), and vessel wall elastic modulus E) all fell within the 95% confidence interval of their respective posterior distributions. The average relative deviation between the point estimates (posterior mean) and the true values ​​was less than 5%. This fully demonstrates that the method has high inversion accuracy under the condition of controllable observation noise. At the same time, the posterior distribution exhibits a reasonable unimodal shape and a moderate distribution width, objectively reflecting the uncertainty inherent in parameter estimation under given data quality. This result verifies the core advantage of this application: it not only provides accurate blood flow velocity point estimates, but also systematically quantifies the reliability of the estimation results through a Bayesian framework, providing clinicians with more comprehensive and reliable functional assessment information, including "best estimate" and "uncertainty range," which is significantly better than traditional optimization methods that only provide a single value and cannot assess confidence.

[0061] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0062] Reference Figure 3 This illustration shows a non-invasive coronary blood flow assessment system based on multi-temporal CTA and Bayesian inference provided in an embodiment of this application, specifically including the following modules: Specifically, it includes: The data acquisition module 310 is used to acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; The sampling estimation module 320 is used to generate a posterior probability distribution estimation result of coronary blood flow velocity through a pre-established Bayesian inference framework and the coronary vessel deformation waveform data. The result output module 330 is used to generate a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

[0063] In one embodiment of this application, the data acquisition module 310 includes: The first data acquisition submodule is used to acquire continuous coronary CTA image sequences during the cardiac cycle; The second data acquisition submodule is used to filter out coronary CTA images within a preset motion artifact threshold from the continuous coronary CTA image sequence as the multi-temporal coronary CTA images; The third data acquisition submodule is used to extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images.

[0064] In one embodiment of this application, the third data acquisition submodule includes: The image alignment unit is used to establish a coordinate system, using the end-diastolic coronary CTA image as a reference frame, and aligning the coronary CTA images of each time phase in the coordinate system. The spatial offset compensation unit is used to combine coronary artery bifurcation point marking with a gray-scale mutual information maximization strategy to compensate for the spatial offset of coronary CTA images at different time phases caused by cardiac pulsation and respiratory motion. The coronary artery centerline extraction unit is used to extract the coronary artery centerline from coronary CTA images at various time phases and optimize the smoothness of the centerline through confidence point detection; The vessel boundary point localization unit is used to set nodes at fixed intervals along the center line to locate vessel boundary points in coronary CTA images at various time phases. The deformation waveform generation unit is used to calculate the normal cross-sectional diameter of each boundary point and generate coronary vessel deformation waveform data of the cross-sectional area of ​​the lumen changing over time.

[0065] In one embodiment of this application, the sampling estimation module 320 includes: The first sampling estimation submodule is used to set the inversion parameter vector, which includes instantaneous blood flow velocity, instantaneous intraluminal pressure and vessel wall elastic modulus; The second sampling estimation submodule is used to establish a vascular dynamics physical model based on the laws of conservation of mass and momentum. The third sampling estimation submodule is used to numerically solve the vascular dynamics physical model and output the predicted cross-sectional area time series. The fourth sampling estimation submodule is used to take the coronary artery deformation waveform data as observation data, quantify the probability of the observation data when given inversion parameters, and generate the posterior probability distribution estimation result.

[0066] In one embodiment of this application, the second sampling estimation submodule includes: The model boundary condition setting unit is used to set the inlet boundary of the vascular dynamics physical model to a known flow time series and the outlet boundary to a known pressure time series. The first prior distribution construction unit is used to construct the time series marginal prior distribution of instantaneous blood flow velocity and instantaneous intravascular pressure based on clinical coronary patient data using the kernel density estimation method; The second prior distribution construction unit is used to set a uniform distribution of the elastic modulus of the vessel wall based on typical values ​​of the coronary vessel elastic modulus in the literature. The third prior distribution construction unit is used to set the conjugate weak information inverse Gamma distribution of the observation noise variance and update it through Gibbs sampling.

[0067] In one embodiment of this application, the fourth sampling estimation submodule includes: The posterior probability distribution construction unit is used to construct the joint posterior probability distribution of the inversion parameter vector and the noise variance according to Bayes' theorem. An approximate sampling unit is used to approximate the joint posterior probability distribution and construct a Gaussian process surrogate model during the approximate sampling process. The Gaussian process surrogate model is used to simulate the vascular dynamics physical model. The sample chain output unit is used to perform a preset number of sampling iterations, discard the early combustion period samples, and output the posterior probability distribution sample chain. The posterior estimation result generation unit is used to perform convergence testing on the sample chain and generate the posterior probability distribution estimation result.

[0068] In one embodiment of this application, the result output module 330 includes: The first result output submodule is used to take the mean of the sample chain of the posterior probability distribution as the point estimate of blood flow velocity and minimize the mean square error. The first result output submodule is used to calculate the highest posterior density confidence interval of blood flow velocity at each time point based on the posterior probability distribution. The first result output submodule is used to output the evaluation results, which include blood flow rate time series, uncertainty interval and key parameter statistics.

[0069] Reference Figure 4 The illustration shows a computer electronic device for implementing a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to this application, which may specifically include the following: The aforementioned computer electronic device 1 is manifested in the form of a general-purpose computing device. The components of the computer electronic device 1 may include, but are not limited to: one or more processors or processing units 3, memory 8, and a bus 4 connecting different system components (including memory 8 and processing unit 3).

[0070] Bus 4 represents one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Audio / Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0071] Computer electronic device 1 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer electronic device 1, including volatile and non-volatile media, removable and non-removable media.

[0072] Memory 8 may include computer system readable media in the form of volatile memory, such as random access memory 9 and / or cache memory 10. Computer electronic device 1 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 11 may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard disk drive"). Although Figure 4 As not shown, a disk drive for reading and writing to a removable non-volatile disk (such as a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (such as a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 4 via one or more data media interfaces. The memory may include at least one program product having a set (e.g., at least one) of program modules 13 configured to perform the functions of the embodiments of this application.

[0073] A program / utility 12 having a set (at least one) of program modules 13 may be stored, for example, in memory. Such program modules 13 include—but are not limited to—an operating system, one or more application programs, other program modules 13, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 13 typically perform the functions and / or methods described in the embodiments of this application.

[0074] The computer electronic device 1 can also communicate with one or more external devices 2 (e.g., keyboard, pointing device, display 7, camera, etc.), and with one or more devices that enable an operator to interact with the computer electronic device 1, and / or with any device that enables the computer electronic device 1 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed through the I / O interface 6. Furthermore, the computer electronic device 1 can also communicate with one or more networks (e.g., local area network (LAN)), wide area network (WAN), and / or public networks (e.g., the Internet) through the network adapter 5. Figure 4 As shown, network adapter 5 communicates with other modules of computer electronic device 1 via bus 4. It should be understood that, although... Figure 4 Not shown, it may be combined with other hardware and / or software modules, including but not limited to: microcode, device drivers, redundant processing unit 3, external disk drive array, RAID system, tape drive and data backup storage system 11, etc.

[0075] The processing unit 3 executes various functional applications and data processing by running programs stored in memory 8, such as implementing a non-invasive coronary blood flow assessment method based on multi-phase CTA and Bayesian inference provided in the embodiments of this application.

[0076] That is, when the processing unit 3 executes the above procedure, it achieves the following: acquiring multi-temporal coronary CTA images of the target to be evaluated, and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images; generating a posterior probability distribution estimation result of coronary blood flow velocity through a pre-established Bayesian inference framework and the coronary vessel deformation waveform data; and generating a coronary blood flow velocity evaluation result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

[0077] In the embodiments of this application, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference as provided in all embodiments of this application.

[0078] That is, when the program is executed by the processor, it performs the following: acquiring multi-temporal coronary CTA images of the target to be evaluated, and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images; generating a posterior probability distribution estimation result of coronary blood flow velocity using a pre-established Bayesian inference framework and the coronary vessel deformation waveform data; and generating a coronary blood flow velocity evaluation result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

[0079] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

[0080] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including—but not limited to—electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0081] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the operator's computer, partially on the operator's computer, as a standalone software package, partially on the operator's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the operator's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider). The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably.

[0082] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0083] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device 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 terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0084] The above provides a detailed description of the non-invasive coronary blood flow assessment method and system based on multi-phase CTA and Bayesian inference provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and its core ideas. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference, characterized in that, Including the following steps: Acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; Using a pre-established Bayesian inference framework and the coronary artery deformation waveform data, the posterior probability distribution estimation result of the coronary blood flow velocity is generated. Based on the posterior probability distribution estimation results, a coronary blood flow velocity assessment result corresponding to the target to be evaluated is generated.

2. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 1, characterized in that, The steps of acquiring multi-temporal coronary CTA images of the target to be evaluated and extracting coronary vessel deformation waveform data from the multi-temporal coronary CTA images specifically include: Acquire a sequence of continuous coronary CTA images during the cardiac cycle; Coronary CTA images within a preset motion artifact threshold are selected from the continuous coronary CTA image sequence and used as the multi-temporal coronary CTA images; Coronary vessel deformation waveform data were extracted from the multi-temporal coronary CTA images.

3. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 2, characterized in that, The step of extracting coronary artery deformation waveform data from the multi-temporal coronary CTA images specifically includes: A coordinate system was established, with the end-diastolic coronary CTA image as the reference frame, and the coronary CTA images of each time phase were aligned in the coordinate system. By combining coronary artery bifurcation point marking with a gray-scale mutual information maximization strategy, spatial offsets caused by cardiac pulsation and respiratory motion in coronary CTA images at different time phases are compensated. The coronary artery centerline is extracted from coronary CTA images at different time phases, and the smoothness of the centerline is optimized by confidence point detection; Nodes are set at fixed intervals along the centerline to locate vessel boundary points in coronary CTA images at various time phases; Calculate the normal cross-sectional diameter at each boundary point to generate coronary vessel deformation waveform data of the lumen cross-sectional area changing over time.

4. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 1, characterized in that, The step of generating a posterior probability distribution estimate of coronary blood flow velocity using a pre-established Bayesian inference framework and the coronary vessel deformation waveform data specifically includes the following steps: Set an inversion parameter vector, which includes instantaneous blood flow velocity, instantaneous intraluminal pressure, and vessel wall elastic modulus; A physical model of vascular dynamics was established based on the laws of conservation of mass and momentum. The vascular dynamics physical model is solved numerically, and the predicted cross-sectional area time series is output. The coronary artery deformation waveform data is used as observation data. The probability of the observation data is quantified when given inversion parameters, and the posterior probability distribution estimation result is generated.

5. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 4, characterized in that, The steps for establishing a vascular dynamics physical model based on the laws of conservation of mass and momentum specifically include the following steps: The inlet boundary of the vascular dynamics physical model is set to a known flow rate time series, and the outlet boundary is set to a known pressure time series. Based on clinical coronary patient data, a time-series marginal prior distribution of instantaneous blood flow velocity and instantaneous intravascular pressure was constructed using the kernel density estimation method; Based on typical values ​​of coronary vessel elastic modulus in the literature, a uniform distribution of vessel wall elastic modulus is established. We define the conjugate weak information inverse Gamma distribution of the observation noise variance and update it using Gibbs sampling.

6. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 5, characterized in that, The step of using the coronary artery deformation waveform data as observation data, quantifying the probability of the observation data given inversion parameters, and generating the posterior probability distribution estimation result specifically includes the following steps: Construct the joint posterior probability distribution of the inversion parameter vector and the noise variance according to Bayes' theorem; The joint posterior probability distribution is approximated, and a Gaussian process surrogate model is constructed during the approximation process. The Gaussian process surrogate model is used to simulate the physical model of vascular dynamics. Perform a preset number of sampling iterations, discard the early combustion period samples, and output the posterior probability distribution sample chain; The convergence of the sample chain is tested to generate the posterior probability distribution estimation result.

7. The non-invasive coronary blood flow assessment method based on multi-temporal CTA and Bayesian inference according to claim 6, characterized in that, The step of generating a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result specifically includes the following steps: The mean of the sample chain of the posterior probability distribution is used as the point estimate of the blood flow velocity to minimize the mean square error. Based on the posterior probability distribution, the highest posterior density confidence interval of blood flow velocity at each time point is calculated. The evaluation results are output, which include blood flow rate time series, uncertainty interval, and statistical information of key parameters.

8. A non-invasive coronary blood flow assessment system based on multi-temporal CTA and Bayesian inference, characterized in that, include: The data acquisition module is used to acquire multi-temporal coronary CTA images of the target to be evaluated, and extract coronary vessel deformation waveform data from the multi-temporal coronary CTA images; The sampling estimation module is used to generate a posterior probability distribution estimation result of coronary blood flow velocity using a pre-established Bayesian inference framework and the coronary vessel deformation waveform data. The result output module is used to generate a coronary blood flow velocity assessment result corresponding to the target to be evaluated based on the posterior probability distribution estimation result.

9. A computer electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images