Cardiovascular disease assessment method based on CT enhanced scanning examination

Abstract

The invention discloses a cardiovascular disease assessment method and system based on CT enhanced scanning examination, relates to the technical field of CT enhanced scanning, and discloses the cardiovascular disease assessment method and system based on CT enhanced scanning examination, which obtains original projection data through dynamic CT enhanced scanning, and calculates permeability function data and time delay data; a vascular wall inflammation score and a microcirculation dysfunction score are generated by using a dual-channel deep learning fusion model, so that synchronous quantitative evaluation of vascular wall inflammation activity and microcirculation functions is realized, and the vascular wall inflammation activity state and the microcirculation dysfunction degree can be synchronously quantified; therefore, the accuracy of cardiovascular disease risk assessment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of CT enhanced scanning, in particular to a cardiovascular disease evaluation method and system based on CT enhanced scanning. BACKGROUND

[0002] The clinical evaluation of cardiovascular disease has long been limited to single-dimensional diagnostic methods, making it difficult to fully capture the complexity of pathophysiological processes. In the prior art, coronary angiography can directly show the morphological characteristics of vascular stenosis, but completely ignores the inflammatory activity of the vascular wall. Inflammation, as a key driver of plaque vulnerability, directly affects the risk of acute cardiovascular events. At the same time, microcirculation dysfunction, as the pathological basis of myocardial ischemia, lacks effective tools for quantitative evaluation in clinical practice, and doctors can only rely on indirect symptoms or invasive detection to make inferences, resulting in a significant blind spot in the diagnosis of myocardial ischemia. Although computed tomography technology has been applied in the field of cardiovascular imaging, conventional methods are limited to the analysis of anatomical structures in static images, and cannot analyze the spatiotemporal evolution of contrast agents in blood vessels during dynamic enhanced scanning, including key information such as hemodynamic parameters, permeability characteristics, and microcirculation delay effects. This technical limitation prevents existing systems from simultaneously quantifying vascular wall inflammation and microcirculation function, making it difficult to establish a dynamic correlation model of multi-dimensional parameters, hindering the precise stratification of cardiovascular disease risk and the development of early intervention strategies.

[0003] The above content is only used to assist in understanding the technical solutions of the present application and does not represent the acknowledgement of the above content as prior art. SUMMARY

[0004] The main purpose of the present application is to provide a cardiovascular disease evaluation method and system based on CT enhanced scanning, aiming to improve the accuracy of cardiovascular disease risk assessment.

[0005] To achieve the above-mentioned purpose, the present application provides a cardiovascular disease evaluation method based on CT enhanced scanning, which comprises: obtaining the original projection data of the cardiovascular region by a dynamic CT enhanced scanning device; performing dynamic permeability index calculation processing on the original projection data to generate permeability function data; performing microcirculation delay time distribution calculation processing on the original projection data to generate time delay data; inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate vascular wall inflammation score data and microcirculation dysfunction score data; generating a cardiovascular disease risk assessment result based on the vascular wall inflammation score data and the microcirculation dysfunction score data.

[0006] In an embodiment, the method further comprises: obtaining projection data and corresponding biopsy verification data of historical cases; inputting the projection data of the historical cases into the dual-channel deep learning fusion model for processing to generate prediction score data; calculating a loss function value between the prediction score data and the biopsy verification data; updating model parameters through back propagation based on the loss function value for processing until the model converges to obtain the dual-channel deep learning fusion model.

[0007] In an embodiment, the step of performing dynamic permeability index calculation processing on the original projection data to generate permeability function data comprises: extracting projection intensity time series data of multiple position points along the vessel centerline from the original projection data; performing time derivative calculation processing on the projection intensity time series data to generate projection intensity change rate data; summing the projection intensity change rate data of the same position point at different projection angles to generate projection intensity total change rate data; dividing the projection intensity total change rate data by baseline projection intensity data to generate permeability function data.

[0008] In an embodiment, the step of obtaining original projection data of a cardiovascular region by a dynamic CT enhancement scanning device comprises: performing multi-phase dynamic scanning processing on the cardiovascular region using a multi-row CT device; collecting original sinogram data as original projection data at a preset slice thickness after injecting iodine contrast agent; The method further comprises: collecting static projection data before iodine contrast agent injection; extracting initial intensity values of vessel centerline position points from the static projection data as baseline projection intensity data.

[0009] In an embodiment, the step of performing microcirculation delay time distribution calculation processing on the original projection data to generate time delay data comprises: extracting projection intensity time series data of multiple position points along the vessel centerline from the original projection data; performing peak detection processing on the projection intensity time series data to generate time point data reaching peak intensity; averaging the time point data of the same position point at all projection angles to generate time delay data.

[0010] In an embodiment, before the step of inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate the blood vessel wall inflammation score data and the microcirculation dysfunction score data, the method further comprises: performing spatial smoothing processing on the time delay data to generate optimized time delay data; performing time dimension normalization processing on the permeability function data to generate optimized permeability function data; using the optimized time delay data to replace the time delay data, and using the optimized permeability function data to replace the permeability function data.

[0011] In an embodiment, the step of inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate the blood vessel wall inflammation score data and the microcirculation dysfunction score data comprises: in a first channel, converting the permeability function data into a three-dimensional spatial feature map, processing through a three-dimensional convolutional neural network to generate spatial feature data; in a second channel, converting the time delay data into time series data, processing through a Transformer encoder to generate time feature data; inputting the spatial feature data and the time feature data into an attention fusion layer for processing to generate fusion feature data; inputting the fusion feature data into a fully connected layer for processing to generate the blood vessel wall inflammation score data and the microcirculation dysfunction score data.

[0012] In an embodiment, the step of inputting the spatial feature data and the time feature data into an attention fusion layer for processing to generate fusion feature data comprises: calculating a correlation matrix of the spatial feature data and the time feature data; generating spatial feature weight data and time feature weight data based on the correlation matrix; using the spatial feature weight data to perform weighted processing on the spatial feature data to generate weighted spatial feature data; using the time feature weight data to perform weighted processing on the time feature data to generate weighted time feature data; concatenating the weighted spatial feature data and the weighted time feature data to generate fusion feature data.

[0013] In an embodiment, the step of generating a cardiovascular disease risk assessment result based on the blood vessel wall inflammation score data and the microcirculation dysfunction score data comprises: generate high-risk vulnerable plaque early warning data when the blood vessel wall inflammation score data exceeds a preset inflammation threshold value; obtain blood flow reserve score data calculated in advance by an external device; weight and fuse the microcirculation dysfunction score data and the blood flow reserve score data to generate corrected myocardial ischemia risk assessment data.

[0014] In addition, to achieve the above-mentioned purpose, the application further provides a cardiovascular disease assessment system based on CT enhanced scan examination, which comprises a memory, a processor, and a cardiovascular disease assessment program based on CT enhanced scan examination stored on the memory and capable of running on the processor, and the cardiovascular disease assessment program based on CT enhanced scan examination is configured to implement the steps of the cardiovascular disease assessment method based on CT enhanced scan examination.

[0015] The cardiovascular disease assessment method and system based on CT enhanced scan examination provided by the application can realize synchronous quantitative assessment of blood vessel wall inflammation activity and microcirculation function, can quantitatively assess the state of blood vessel wall inflammation activity and the degree of microcirculation dysfunction, and thus can improve the accuracy of cardiovascular disease risk assessment. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the specification, serve to explain the principles of the application.

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative labor.

[0018] Figure 1 a flowchart provided by an embodiment of the cardiovascular disease assessment method based on CT enhanced scan examination of the application; Figure 2 a structural schematic diagram provided by an embodiment of the cardiovascular disease assessment system based on CT enhanced scan examination of the application.

[0019] Explanation of the reference signs: 10, memory; 20, processor.

[0020] The objectives, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0021] The technical solutions in the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. The components of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0022] It should be understood that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used for differentiation, and cannot be understood as indicating or implying relative importance.

[0023] In the prior art, although coronary angiography can intuitively present the morphological characteristics of vascular stenosis, it completely ignores the inflammatory activity state of the vascular wall, and the inflammatory response as the core driving factor of plaque vulnerability directly affects the risk of acute cardiovascular events. At the same time, as the pathological basis of myocardial ischemia, the quantitative evaluation of microcirculation dysfunction lacks effective tools in clinical practice, and doctors can only rely on indirect symptoms or invasive detection for inference, resulting in a significant blind area in the diagnosis of myocardial ischemia. Although computed tomography technology has been applied in the field of cardiovascular imaging, the conventional method is limited to the analysis of the anatomical structure of static images, and cannot analyze the spatiotemporal evolution law of the contrast agent in the blood vessel during the dynamic enhancement scanning process, including the key information of hemodynamic parameters, permeation behavior characteristics and microcirculation delay effect. This technical limitation makes the existing system unable to realize the synchronous quantification of vascular wall inflammation and microcirculation function, and it is more difficult to establish a dynamic correlation model of multi-dimensional parameters, which hinders the precise stratification of cardiovascular disease risk and the development of early intervention strategies.

[0024] Based on this, the embodiments of the present application provide a cardiovascular disease evaluation method based on CT enhancement scanning examination, referring to Figure 1 , the cardiovascular disease evaluation method based on CT enhancement scanning examination comprises steps S100-S500, wherein: Step S100: acquiring original projection data of a cardiovascular region through a dynamic CT enhancement scanning device; Step S200: Perform dynamic permeability index calculation on the original projection data to generate permeability function data; Step S300: Perform micro-circulation delay time distribution calculation on the original projection data to generate time delay data; Step S400: Input the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate vascular wall inflammation score data and microcirculation dysfunction score data; Step S500: Generate cardiovascular disease risk assessment results based on the vascular wall inflammation score data and the microcirculation dysfunction score data.

[0025] In this embodiment, the dynamic CT contrast-enhanced scanning device can be understood as a device capable of continuously acquiring imaging data of the cardiovascular region. Its main function is to capture the time-series information of hemodynamic changes after contrast agent injection. For example, the device can achieve continuous imaging of the target area through helical scanning or multi-phase scanning, thereby providing raw projection data containing the time dimension. Furthermore, the dynamic permeability index calculation can be implemented in various ways, such as by performing numerical difference or curve slope analysis on the projection intensity time-series data to extract the intensity change rate information, and then normalizing it in conjunction with the baseline intensity value to generate permeability function data. Specifically, this processing method can reflect the dynamic changes in vascular wall permeability. In addition, the microcirculation delay time distribution calculation can be achieved by detecting the location of feature points in the projection intensity time series, such as by identifying the time points of intensity peaks or specific thresholds, and then performing statistical analysis in conjunction with multi-angle data to generate time delay data. Thus, this step can quantify the delay characteristics of microcirculation blood flow transmission. The design of a dual-channel deep learning fusion model can be implemented using various architectures. For example, in the first channel, a two-dimensional convolutional network can be used to extract spatial features, and in the second channel, a recurrent neural network can be used to analyze time series features. Feature fusion can be achieved through weighted summation or concatenation, ultimately generating vascular wall inflammation score data and microcirculation dysfunction score data.

[0026] This application addresses the shortcomings of traditional cardiovascular disease assessment methods, which rely on single indicators and lack integration of dynamic hemodynamic information. It proposes a technical solution based on the fusion of dynamic CT enhanced scanning and deep learning. By processing permeability function data and time delay data separately, it achieves a quantitative assessment of vascular wall inflammation and the degree of microcirculatory dysfunction. Simultaneously, it combines the two sets of scoring data to generate a comprehensive risk assessment result, thus overcoming the limitations of existing technologies that rely solely on morphological indicators.

[0027] The embodiment of the application obtains the original projection data of the cardiovascular region through a dynamic CT enhancement scanning device, which can continuously capture time sequence information after contrast agent injection, thereby providing a basic data source containing hemodynamic changes for subsequent analysis. Further, the original projection data is processed by dynamic permeability index calculation to generate permeability function data, wherein based on the projection intensity time sequence along the vessel centerline, the dynamic intensity information is converted into function data quantifying the permeability of the vessel wall by calculating and normalizing the time derivative, which can reflect the abnormal vascular permeability caused by inflammation in real time. At the same time, the original projection data is also used for microcirculation delay time distribution calculation processing to generate time delay data, specifically, the peak arrival time is detected from the projection intensity time sequence, and the distribution data characterizing the microcirculation blood flow delay is generated by multi-angle average processing, which realizes the quantitative evaluation of microcirculation dysfunction. Subsequently, the permeability function data and the time delay data are input into a dual-channel deep learning fusion model for processing, wherein the first channel extracts spatial features through a three-dimensional convolutional neural network, the second channel analyzes time sequence features through a Transformer encoder, and the fusion of the information of the two channels is realized by using an attention mechanism, thereby generating vessel wall inflammation score data and microcirculation dysfunction score data. Finally, based on the two types of score data, cardiovascular disease risk assessment results are generated, which integrate the inflammation score and the microcirculation score to comprehensively characterize the cardiovascular pathophysiological state, thereby effectively solving the problem that the traditional method cannot simultaneously quantitatively evaluate the activity state of the vessel wall inflammation and the degree of microcirculation dysfunction, and fully utilizing the hemodynamic information in the dynamic CT enhancement scanning to realize comprehensive risk assessment.

[0028] In a feasible implementation, the method further comprises: obtaining projection data of historical cases and corresponding vascular biopsy verification data; inputting the projection data of the historical cases into the dual-channel deep learning fusion model for processing to generate predicted score data; calculating a loss function value between the predicted score data and the vascular biopsy verification data; updating the model parameters by back propagation based on the loss function value for processing until the model converges to obtain the dual-channel deep learning fusion model.

[0029] In this embodiment, the projection data of historical cases refers to the original data derived from actual dynamic CT scanning, which can be achieved by using multi-phase dynamic scanning data collected by a multi-row CT device, aiming to completely retain the timing information of blood flow dynamics. Among them, the vascular biopsy verification data can be understood as the pathological gold standard data, which can be obtained by microscope analysis or molecular biology detection of vascular tissue samples, aiming to provide clear supervision targets for model training. In practical applications, the loss function value refers to an index that quantifies the difference between the prediction results and the pathological verification, which can be achieved by using generalization methods such as mean square error, cross-entropy loss, etc., aiming to establish an objective error evaluation mechanism. In addition, back propagation refers to an optimization algorithm that can be implemented by gradient descent method or other variants, aiming to efficiently use loss information to guide model parameter adjustment.

[0030] In this embodiment, the above scheme solves the problem of lack of real pathological data calibration for the dual-channel deep learning fusion model before application by constructing a model training mechanism. Specifically, the projection data and corresponding vascular biopsy verification data of historical cases are obtained, providing a reliable data basis for model training, ensuring that the training data closely matches the real pathological state. The projection data of historical cases is input into the dual-channel deep learning fusion model for processing to generate prediction score data, which fully utilizes the vascular permeability and microcirculation delay features contained in the projection data, enabling the model to extract deep patterns from the original scan information. The loss function value between the prediction score data and the vascular biopsy verification data is calculated, establishing an objective error evaluation mechanism by quantifying the difference, avoiding the bias introduced by subjective experience. Based on the loss function value, the model parameters are updated through back propagation until the model converges, and this iteration process ensures that the model has stable and reliable prediction ability, and finally the obtained model can accurately output the quantitative scores of vascular wall inflammation and microcirculation dysfunction. In this way, not only does it effectively solve the problem of unreliable risk assessment results caused by untargeted model training, but also significantly improves the accuracy and reliability of cardiovascular disease risk assessment.

[0031] In a feasible implementation, the step of performing dynamic permeability index calculation on the original projection data to generate permeability function data includes: extracting projection intensity time series data of multiple position points along the blood vessel centerline from the original projection data; performing time derivative calculation on the projection intensity time series data to generate projection intensity change rate data; summing the projection intensity change rate data of the same position point at different projection angles to generate projection intensity total change rate data; and dividing the projection intensity total change rate data by baseline projection intensity data to generate permeability function data.

[0032] In this embodiment, the projection intensity time series data refers to a data set reflecting the change of contrast agent concentration in the blood vessel region over time obtained by continuous acquisition, which can be realized by multi-phase dynamic CT scanning technology. Among them, the time derivative calculation process can adopt numerical methods such as finite difference method or polynomial fitting derivation, the purpose is to highlight the instantaneous change characteristics of contrast agent concentration, so as to more accurately reflect the dynamic change process of vascular wall permeability. The total change rate data of projection intensity is obtained by integrating the change rate data of the same position point at different projection angles, which can be realized by simple arithmetic sum or weighted sum, etc. The purpose is to suppress the influence of single-angle artifacts and random noise, and improve the stability and representativeness of the data. The baseline projection intensity data refers to the initial intensity value extracted from the static projection data collected before injecting contrast agent, which serves as a reference benchmark for subsequent normalization processing.

[0033] In this embodiment, the scheme first extracts the projection intensity time series data of multiple position points along the blood vessel center line, ensures that the analysis focuses on the core area of the blood vessel rather than the surrounding tissue, and completely captures the dynamic enhancement process of contrast agent. Then, the projection intensity change rate data is generated by time derivative calculation processing, which highlights the instantaneous change rate of intensity, because the vascular permeability is directly related to the change rate of contrast agent concentration rather than the static intensity value, so as to more truly reflect the abnormal vascular wall permeability caused by inflammation. Then, the projection intensity total change rate data is generated by summing the projection intensity change rate data of the same position point at different projection angles, which integrates multi-view three-dimensional information and significantly improves the stability and representativeness of the change rate data. Finally, the permeability function data is generated by dividing the projection intensity total change rate data by the baseline projection intensity data, the normalization process eliminates the influence of device parameter differences and patient individual basic state, so that the permeability function data has cross-sample comparability, provides standardized input for the subsequent blood vessel wall inflammation scoring model, and ensures the accuracy of risk assessment.

[0034] Based on the above technical scheme, first, the problem of lack of standardized method in the processing of original projection data is solved, and the calculation accuracy of the permeability function data is improved through the systematic data processing flow. Secondly, through the multi-angle data integration and normalization processing, the noise interference is effectively suppressed, and the problems of ineffective integration of data at different projection angles and lack of baseline intensity normalization are overcome, thereby significantly improving the reliability of the blood vessel wall inflammation score.

[0035] In an implementation, the step of acquiring the original projection data of the cardiovascular region by the dynamic CT enhancement scanning device comprises: performing multi-phase dynamic scanning on the cardiovascular region by using a multi-row CT device; collecting original sinogram data as the original projection data with a preset slice thickness after injecting iodine contrast agent; the method further comprises: collecting static projection data before injecting the iodine contrast agent; and extracting an initial intensity value of a blood vessel center line position point from the static projection data as baseline projection intensity data.

[0036] In the embodiment, the multi-row CT device refers to a computer tomography device with a multi-row detector array, which can be implemented by using a CT device with 64 rows, 128 rows or more rows, aiming to provide dynamic scanning capability with high time resolution and spatial resolution. The preset slice thickness refers to a slice thickness parameter set in advance according to scanning requirements, which can be implemented by adjusting the slice thickness setting in the scanning protocol, aiming to ensure the spatial resolution consistency of the collected data. The static projection data refers to the blood vessel region image data acquired without introducing the contrast agent, which can be implemented by setting a contrast agent-free stage in the scanning sequence, aiming to provide a pure baseline reference.

[0037] In the embodiment, the technical solution can continuously capture the blood flow dynamics evolution process by performing multi-phase dynamic scanning on the cardiovascular region by using the multi-row CT device, providing complete time dimension data basis for subsequent analysis. The original sinogram data is collected as the original projection data with a preset slice thickness after injecting the iodine contrast agent, which ensures the stability of the spatial resolution and avoids data distortion caused by inconsistent slice thickness. The key lies in the operation of collecting the static projection data before injecting the iodine contrast agent, which effectively isolates the variables introduced by the contrast agent based on the physiological state of the blood vessel without contrast agent interference before injection, thereby obtaining a pure initial image of the blood vessel. Furthermore, the initial intensity value of the blood vessel center line position point is extracted from the static projection data as the baseline projection intensity data, which directly measures the core area of the blood vessel and avoids the interference of surrounding tissue noise, so that the baseline value accurately corresponds to the original intensity of the blood vessel itself, providing a reliable basis for the permeability function calculation, ensuring the reliability of the subsequent permeability function calculation, and finally supporting the accuracy of the blood vessel wall inflammation score.

[0038] In an implementation, the step of performing microcirculation delay time distribution calculation processing on the original projection data to generate time delay data comprises: extracting projection intensity time series data of multiple position points along the blood vessel center line from the original projection data; performing peak value detection processing on the projection intensity time series data to generate time point data reaching the peak intensity; and performing average processing on the time point data of the same position point under all projection angles to generate time delay data.

[0039] In this embodiment, the projection intensity time series data refers to the projection intensity information collected at multiple position points along the vessel centerline during dynamic CT enhancement scanning. It can be achieved by time segmentation and spatial positioning of the original projection data, aiming to ensure that the analysis focuses on the core blood flow path of the vessel rather than the edge area, thereby avoiding the influence of surrounding tissue signal interference and vessel wall noise. The peak detection process can be understood as a technical means for identifying local maximum values in the projection intensity time series data, which can be achieved by using the derivative-based extreme value detection method or the threshold-based peak extraction algorithm, aiming to accurately capture the moment when the contrast agent reaches the highest concentration in the vessel. In addition, the average processing of the time point data refers to the arithmetic average or weighted average operation of the peak time of the same position point under different projection angles, which can be achieved by simple statistical calculation or more complex multi-source data fusion algorithm, aiming to effectively offset the time series fluctuations caused by the CT device geometry or random noise.

[0040] In this embodiment, the technical solution realizes accurate quantification of microcirculation delay time through a series of ordered steps. First, the projection intensity time series data of multiple position points along the vessel centerline is extracted from the original projection data, which ensures that the obtained data can stably reflect the contrast agent passing process, laying a reliable foundation for subsequent delay calculation. Then, the time point data reaching the peak intensity is generated by performing peak detection processing on the projection intensity time series data, which directly relates to blood flow velocity and microcirculation perfusion efficiency, avoiding time sequence deviation caused by relying only on average intensity and other ambiguous features. Finally, the time point data of the same position point under all projection angles is averaged to generate time delay data, which eliminates angle-dependent error by utilizing the complementarity of multi-source data, significantly improving the robustness of time delay estimation. Through the above technical solution, not only the time delay data fluctuation problem caused by the difference in projection angle is solved, but also a high-precision microcirculation function state quantification index for cardiovascular disease risk assessment is provided.

[0041] In a feasible implementation, before the step of inputting the permeability function data and the time delay data into the dual-channel deep learning fusion model for processing to generate vessel wall inflammation score data and microcirculation dysfunction score data, the method further comprises: performing spatial smoothing processing on the time delay data to generate optimized time delay data; performing time dimension normalization processing on the permeability function data to generate optimized permeability function data; using the optimized time delay data to replace the time delay data, and using the optimized permeability function data to replace the permeability function data.

[0042] In this embodiment, the time delay data refers to the time point data generated after the peak detection processing of the projection intensity time series data extracted from the original projection data at multiple position points along the vessel centerline. It can be realized by spatial smoothing techniques such as Gaussian filtering, mean filtering, etc. The purpose is to suppress the spatial distribution discontinuity problem caused by local noise and individual differences, so as to more accurately reflect the true delay characteristics of microcirculation. Among them, the permeability function data can be understood as the data generated by the ratio of the total change rate of projection intensity to the baseline projection intensity. It can be realized by time dimension normalization processing, such as mapping the time series to a unified interval or standardizing the time starting point and rate difference, in order to eliminate the time scale inconsistency problem caused by the difference in contrast agent injection speed and hemodynamics between different patients, so that the data is more comparable.

[0043] In this embodiment, during the dynamic CT enhancement scan, the time delay data may cause spatial distribution mutation due to scanning noise and individual vascular anatomy differences. Directly using the original data will affect the stability of the model input. Therefore, by performing spatial smoothing processing on the time delay data, such as applying Gaussian filtering technology, local fluctuations can be effectively suppressed, and the time delay distribution is more continuous. At the same time, the time dynamic characteristics of the permeability function data are individualized in time scale due to the influence of contrast agent injection speed and patient hemodynamics. By time dimension normalization processing, such as mapping the time series to a unified interval, the time starting point and rate difference can be eliminated, which is convenient for the model to focus on pathological characteristics rather than time variables. On this basis, the optimized time delay data is used to replace the time delay data, and the optimized permeability function data is used to replace the permeability function data, which ensures that the data input into the dual-channel deep learning fusion model is optimized, avoids the direct transmission of noise and scale difference, and improves the model's ability to extract key features of vascular wall inflammation and microcirculation dysfunction, solves the problem of spatial distribution discontinuity of original time delay data caused by scanning noise and individual differences, and the scale difference of permeability function data in time dimension, makes the model input more stable, and finally improves the reliability of cardiovascular disease risk assessment.

[0044] In an available implementation, the step of inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model to generate the blood vessel wall inflammation score data and the microcirculation dysfunction score data includes: in a first channel, converting the permeability function data into a three-dimensional spatial feature map, processing through a three-dimensional convolutional neural network to generate spatial feature data; in a second channel, converting the time delay data into time series data, processing through a Transformer encoder to generate time feature data; inputting the spatial feature data and the time feature data into an attention fusion layer for processing to generate fusion feature data; inputting the fusion feature data into a fully connected layer for processing to generate the blood vessel wall inflammation score data and the microcirculation dysfunction score data.

[0045] In this embodiment, the three-dimensional spatial feature map refers to a three-dimensional matrix form data structure constructed based on the spatial distribution characteristics of the permeability function data, which can be realized in the form of multi-dimensional array or tensor, and the purpose is to fully capture the three-dimensional structure information related to blood vessel wall inflammation. The three-dimensional convolutional neural network is a deep learning model specially used for processing three-dimensional data, which can extract the local correlation in three-dimensional space through local receptive field, so as to ensure the integrity of spatial features. The time series data refers to an ordered data set formed by arranging the time delay data according to the time dimension, which can be realized in the form of sliding window or time step segmentation, and the purpose is to effectively represent the time sequence characteristics of the dynamic changes of microcirculation. The Transformer encoder is a deep learning model based on self-attention mechanism, which can extract global patterns in time series by modeling long-range dependencies, thereby overcoming the gradient vanishing problem of traditional time series models when processing long sequences. The attention fusion layer is a dynamic weighted fusion mechanism, which can generate weights by calculating the correlation matrix between different features, thereby realizing adaptive focusing on key information. The fully connected layer is a nonlinear combination module, which can generate the final output by weighted sum and activation function processing of input features, and the purpose is to integrate multi-dimensional pathological information to generate reliable score results.

[0046] In the above scheme, the first channel converts the permeability function data into a three-dimensional spatial feature map through processing, and extracts spatial feature data using a three-dimensional convolutional neural network. This process fully utilizes the characteristics of permeability function data reflecting the spatial distribution of blood vessel wall permeability, and can capture three-dimensional structural information related to inflammation of the blood vessel wall, avoiding the defects of two-dimensional convolution that ignore the depth dimension. The second channel converts the time delay data into time series data through processing, and extracts time feature data with the help of a Transformer encoder. This process targets the time series characteristics of time delay data representing microcirculation dynamics, and can effectively model long-distance dependencies to extract time dynamic patterns of microcirculation dysfunction. Subsequently, the spatial feature data and the time feature data are processed by an attention fusion layer to calculate the correlation matrix of the two and generate dynamic weights, realizing adaptive focusing on key information and avoiding feature redundancy or information loss caused by simple concatenation. Finally, the fusion feature data is processed by a fully connected layer for nonlinear combination to generate blood vessel wall inflammation score data and microcirculation dysfunction score data. Since the input features have fully integrated spatial and temporal pathological information, the score results can more truly reflect the complex state of cardiovascular diseases.

[0047] In addition, in the above scheme, the technical means for obtaining permeability function data and time delay data are combined to form a complete dual-channel deep learning fusion model. Through fine processing of spatial features and time features, and application of adaptive fusion mechanism, the problem of lack of targeted processing mechanism and adaptive fusion ability in the model is solved, thereby significantly improving the reliability of the score results.

[0048] In one possible implementation, the step of inputting the spatial feature data and the time feature data into the attention fusion layer for processing to generate fusion feature data includes: calculating a correlation matrix of the spatial feature data and the time feature data; generating spatial feature weight data and time feature weight data based on the correlation matrix; performing weighted processing on the spatial feature data using the spatial feature weight data to generate weighted spatial feature data; performing weighted processing on the time feature data using the time feature weight data to generate weighted time feature data; and concatenating the weighted spatial feature data and the weighted time feature data to generate fusion feature data.

[0049] In this embodiment, the correlation matrix refers to a mathematical representation for quantifying the strength of association between spatial feature data and temporal feature data, which can be achieved through Pearson correlation coefficient, cosine similarity, or mutual information, etc. The purpose is to capture the intrinsic relationship between the two features in a numerical way. The spatial feature weight data and the temporal feature weight data can be understood as the weight distribution results dynamically generated according to the correlation matrix, which can be directly mapped through the normalized correlation values, or converted through a nonlinear activation function (such as Sigmoid or Softmax) to ensure the rationality and adaptability of the weight distribution. The weighted spatial feature data and the weighted temporal feature data refer to the spatial feature and the temporal feature after weight adjustment, respectively, which aims to highlight the key pathological information while suppressing irrelevant noise. The concatenation operation can be achieved by stacking in the channel dimension, which aims to seamlessly integrate the optimized spatial and temporal features to form a comprehensive feature representation.

[0050] In this embodiment, the above technical solution realizes the adaptive fusion of spatial features and temporal features through the attention mechanism, thereby effectively solving the problem of insufficient extraction of key information in the feature fusion process. In actual operation, first, the correlation matrix of the spatial feature data and the temporal feature data is calculated to quantify the mutual dependence between the two, avoiding the subjective limitations brought by fixed weights. Subsequently, the spatial feature weight data and the temporal feature weight data generated based on the correlation matrix can dynamically emphasize the spatial structure features of the blood vessel wall inflammation area or the temporal dynamic features of the microcirculation delay time according to the characteristics of different cases. By weighting the spatial feature data and the temporal feature data respectively, the spatial details of the blood vessel wall related to inflammation activity and the abnormal time patterns in microcirculation dysfunction are strengthened. Finally, by concatenating the weighted spatial feature data and the temporal feature data, a comprehensive feature representation is formed, which not only contains spatial anatomical details but also reflects temporal dynamic characteristics, providing high-fidelity input for subsequent generation of blood vessel wall inflammation score and microcirculation dysfunction score. On this basis, the above technical solution and the three-dimensional convolutional neural network and the Transformer encoder in the dual-channel deep learning fusion model form an organic whole. By combining the spatial feature data extracted by the three-dimensional convolutional neural network and the temporal feature data generated by the Transformer encoder, the attention fusion layer further improves the sensitivity of the model to key pathological indicators, thereby significantly enhancing the accuracy and reliability of cardiovascular disease risk assessment.

[0051] In an implementable embodiment, the step of generating a cardiovascular disease risk assessment result based on the blood vessel wall inflammation score data and the microcirculation dysfunction score data comprises: generating high-risk vulnerable plaque early warning data when the blood vessel wall inflammation score data exceeds a preset inflammation threshold value; obtaining blood flow reserve score data calculated in advance by an external device; and performing weighted fusion processing on the microcirculation dysfunction score data and the blood flow reserve score data to generate corrected myocardial ischemia risk assessment data.

[0052] In this embodiment, the preset inflammation threshold value refers to a critical value for judging the severity of blood vessel wall inflammation, which is set according to clinical experience or statistical analysis. It can be implemented by using a fixed threshold value or a dynamically adjusted threshold value, and the purpose is to provide a clear risk dividing point to trigger a timely early warning mechanism. The high-risk vulnerable plaque early warning data can be understood as a kind of symbolic output information, which is used to prompt that the blood vessel wall inflammation has reached a dangerous level that may cause acute cardiovascular events. The purpose is to make up for the deficiency of lagging behind in identifying vulnerable plaques in traditional assessment. The blood flow reserve score data refers to a functional index reflecting the influence of coronary artery stenosis on myocardial perfusion, which can be measured by a pressure guide wire or estimated by a non-invasive imaging technique. The purpose is to supplement the limitations of microcirculation dysfunction score in macro blood flow assessment. The weighted fusion processing refers to the process of designing a weight distribution mechanism based on the internal correlation of the two types of data. It can be implemented by using linear weighting, nonlinear weighting or machine learning model to optimize the weight. The purpose is to generate a corrected risk assessment that is more consistent with clinical practice.

[0053] In this embodiment, the scheme first solves the problem of lack of dynamic monitoring of the severity of vascular wall inflammation through a dynamic threshold early warning mechanism. When the vascular wall inflammation score data exceeds the preset inflammation threshold, the system can generate high-risk vulnerable plaque early warning data in a timely manner, thereby providing key decision-making basis for the prevention of acute cardiovascular events. On this basis, by introducing the blood flow reserve score data provided by the external device, the deficiency of microcirculation dysfunction score in macro blood flow evaluation is effectively made up, so that the pathophysiological continuous spectrum from microcirculation to macro blood flow is completely covered. Finally, by weighting and fusing the microcirculation dysfunction score data and the blood flow reserve score data, adaptive weight adjustment of the two types of data is realized, which not only retains the sensitivity of microcirculation score to early dysfunction, but also integrates the objective verification of blood flow reserve score to hemodynamics, and finally generates corrected myocardial ischemia risk assessment data. This process is particularly suitable for complex cases where microcirculation abnormalities do not match macro blood flow, significantly improving the accuracy and individualization level of cardiovascular disease risk assessment. In addition, the above scheme is combined with other steps in the foregoing cardiovascular disease assessment method, such as the blood vessel wall inflammation score data and microcirculation dysfunction score data generated by the dual-channel deep learning fusion model, to provide a reliable data basis for subsequent weighting and fusion processing, thereby forming a complete technical chain from data acquisition, processing to risk assessment, solving the problems of untimely identification of high-risk vulnerable plaques and insufficient accuracy of myocardial ischemia risk assessment.

[0054] In the embodiments of the present application, the cardiovascular disease assessment method based on CT enhanced scan examination obtains original projection data through dynamic CT enhanced scan, calculates permeability function data and time delay data, and generates blood vessel wall inflammation score and microcirculation dysfunction score by using a dual-channel deep learning fusion model, realizing synchronous quantitative evaluation of blood vessel wall inflammation activity and microcirculation function, and being capable of quantifying the state of blood vessel wall inflammation activity and the degree of microcirculation dysfunction, thereby improving the accuracy of cardiovascular disease risk assessment.

[0055] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the cardiovascular disease assessment method based on CT enhanced scan examination of the present application. More forms of simple transformation based on this technical concept are within the protection scope of the present application.

[0056] The present application also provides a cardiovascular disease assessment system based on CT enhanced scan examination, which is described with reference to Figure 2The CT-enhanced scan-based cardiovascular disease assessment system comprises a memory 10, a processor 20, and a CT-enhanced scan-based cardiovascular disease assessment program stored on the memory 10 and executable on the processor 20, wherein the CT-enhanced scan-based cardiovascular disease assessment program is configured to implement the steps of the CT-enhanced scan-based cardiovascular disease assessment method.

[0057] The CT-enhanced scan-based cardiovascular disease assessment system provided in the application can improve the accuracy of cardiovascular disease risk assessment by using the CT-enhanced scan-based cardiovascular disease assessment method in the above embodiments. Compared with the prior art, the CT-enhanced scan-based cardiovascular disease assessment system provided in the application has the same beneficial effects as the CT-enhanced scan-based cardiovascular disease assessment method provided in the above embodiments, and other technical features of the CT-enhanced scan-based cardiovascular disease assessment system are the same as the features disclosed in the above method, which will not be repeated here.

[0058] It should be understood that various parts of the present application can be realized by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0059] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any equivalent structural transformation or direct / indirect application in other related technical fields made by using the content of the present application specification and drawings within the technical concept of the present application is included in the patent protection scope of the present application.

Claims

1. A method for cardiovascular disease assessment based on CT contrast enhanced scan examination, characterized in that, The method comprises: obtaining original projection data of a cardiovascular region by a dynamic CT enhancement scanning device; performing dynamic permeability index calculation processing on the original projection data to generate permeability function data; performing microcirculation delay time distribution calculation processing on the original projection data to generate time delay data; inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate blood vessel wall inflammation score data and microcirculation dysfunction score data; generating a cardiovascular disease risk assessment result based on the blood vessel wall inflammation score data and the microcirculation dysfunction score data.

2. The method for assessing cardiovascular disease based on CT enhanced scan examination according to claim 1, wherein, The method further comprises: obtaining projection data of historical cases and corresponding vascular biopsy verification data; inputting the projection data of the historical cases into the dual-channel deep learning fusion model for processing to generate predicted score data; calculating a loss function value between the predicted score data and the vascular biopsy verification data; updating model parameters by back propagation based on the loss function value for processing until the model converges to obtain the dual-channel deep learning fusion model.

3. The method for assessing cardiovascular disease based on CT enhanced scan examination according to claim 1, wherein, The step of performing dynamic permeability index calculation processing on the original projection data to generate permeability function data comprises: extracting projection intensity time series data of multiple position points along the blood vessel center line from the original projection data; performing time derivative calculation processing on the projection intensity time series data to generate projection intensity change rate data; summing the projection intensity change rate data of the same position point at different projection angles to generate projection intensity total change rate data; dividing the projection intensity total change rate data by baseline projection intensity data to generate permeability function data.

4. The method for assessing cardiovascular disease based on CT enhancement scan examination according to claim 3, wherein, The step of obtaining original projection data of a cardiovascular region by a dynamic CT enhancement scanning device comprises: performing multi-phase dynamic scanning processing on the cardiovascular region using a multi-row CT device; collecting original sinogram data as original projection data with a preset slice thickness after injecting iodine contrast agent; The method further comprises: collecting static projection data before injecting iodine contrast agent; extracting initial intensity values of blood vessel center line position points from the static projection data as baseline projection intensity data.

5. The method for assessing cardiovascular disease based on CT enhanced scan examination according to claim 1, wherein, The step of performing microcirculation delay time distribution calculation processing on the original projection data to generate time delay data comprises: extracting projection intensity time series data of multiple position points along the blood vessel center line from the original projection data; performing peak detection processing on the projection intensity time series data to generate time point data reaching peak intensity; averaging the time point data of the same position point at all projection angles to generate time delay data.

6. The method for assessing cardiovascular disease based on CT enhancement scan examination according to claim 1, wherein, Before the step of inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate blood vessel wall inflammation score data and microcirculation dysfunction score data, the method further comprises: performing spatial smoothing processing on the time delay data to generate optimized time delay data; performing time dimension normalization processing on the permeability function data to generate optimized permeability function data; The optimized time delay data is used to replace the time delay data, and the optimized permeability function data is used to replace the permeability function data.

7. The method for assessing cardiovascular disease based on CT enhanced scan examination according to claim 1, wherein, The step of inputting the permeability function data and the time delay data into a dual-channel deep learning fusion model for processing to generate blood vessel wall inflammation score data and microcirculation dysfunction score data includes: In the first channel, the permeability function data is converted into a three-dimensional spatial feature map, which is processed by a three-dimensional convolutional neural network to generate spatial feature data; In the second channel, the time delay data is converted into time series data, which is processed by a Transformer encoder to generate time feature data; The spatial feature data and the time feature data are input into an attention fusion layer for processing to generate fusion feature data; The fusion feature data is input into a fully connected layer for processing to generate blood vessel wall inflammation score data and microcirculation dysfunction score data.

8. The method for assessing cardiovascular disease based on CT enhancement scan examination according to claim 7, wherein, The step of inputting the spatial feature data and the time feature data into an attention fusion layer for processing to generate fusion feature data includes: calculating a correlation matrix of the spatial feature data and the time feature data; generating spatial feature weight data and time feature weight data based on the correlation matrix; weighting the spatial feature data using the spatial feature weight data to generate weighted spatial feature data; weighting the time feature data using the time feature weight data to generate weighted time feature data; concatenating the weighted spatial feature data and the weighted time feature data to generate fusion feature data.

9. The method for assessing cardiovascular disease based on CT enhancement scan examination according to claim 1, wherein, The step of generating a cardiovascular disease risk assessment result based on the blood vessel wall inflammation score data and the microcirculation dysfunction score data includes: when the blood vessel wall inflammation score data exceeds a preset inflammation threshold, generating high-risk vulnerable plaque warning data; obtaining blood flow reserve score data calculated in advance by an external device; weighting and fusing the microcirculation dysfunction score data and the blood flow reserve score data to generate corrected myocardial ischemia risk assessment data.

10. A cardiovascular disease assessment system based on CT contrast enhanced scan examination, characterized by, The cardiovascular disease assessment system based on CT enhanced scan examination includes a memory, a processor, and a cardiovascular disease assessment program based on CT enhanced scan examination stored on the memory and executable on the processor, which is configured to implement the steps of the cardiovascular disease assessment method based on CT enhanced scan examination according to any one of claims 1 to 9.

Images