Intelligent myocardial ischemia diagnosis system, computer readable storage medium and electronic equipment

Abstract

The invention discloses an intelligent myocardial ischemia diagnosis system which comprises a coronary blood flow state calculation module, a high-frame-rate cardiac ultrasonic data acquisition module and a coronary lesion judgment module. The coronary artery blood flow state calculation module establishes and trains a machine learning model, and calculates the coronary artery blood flow state by using the machine learning model. The high-frame-rate cardiac ultrasound data acquisition module acquires high-frame-rate cardiac ultrasound data, and the coronary artery lesion judgment module judges whether myocardial ischemia exists or not according to the coronary artery blood flow state calculated by the machine learning model. When the myocardial ischemia exists, judging whether the lesion blood vessel causing the myocardial ischemia is a coronary artery large vessel, a coronary artery microvessel or both of the coronary artery large vessel and the coronary artery microvessel. According to the method, rich coronary blood flow time information of the high-frame-rate cardiac ultrasound data is utilized, and the accuracy of predicting whether myocardial ischemia exists or not and the cause of the myocardial ischemia is high.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence-assisted diagnosis and treatment technology, and particularly to the intelligent diagnosis of myocardial ischemia. More specifically, it relates to an intelligent diagnostic system for myocardial ischemia based on high frame rate cardiac ultrasound data, as well as a computer-readable storage medium and electronic device. Background Technology

[0002] Coronary microvascular disease (CMVD) refers to a clinical syndrome characterized by exertional angina or objective evidence of myocardial ischemia caused by structural and / or functional abnormalities of the precoronary arterioles and arterioles under the influence of various pathogenic factors. With the continuous development of artificial intelligence (AI) technology, the diagnosis and treatment of diseases increasingly utilize AI-assisted methods such as big data and machine learning. AI enables the effective use of previously overlooked or unusable diagnostic information, greatly contributing to the predictability and accuracy of disease diagnosis. For example, Chinese invention patent CN107205725B proposes a method for assessing myocardial infarction using ultrasound strain imaging.

[0003] The same applies to the diagnosis and treatment of myocardial ischemia, and various artificial intelligence-assisted diagnostic systems have emerged. For example, Chinese invention patent application CN115804620A proposes a method for generating a bullseye diagram segment model and subsequently indirectly reflecting the relationship between coronary artery patterns and myocardial segments. However, this method has limitations in terms of sensitivity and accuracy in detecting myocardial ischemia.

[0004] Chinese invention patent application CN115211897A proposes a myocardial ischemia diagnostic system based on ultrasound data. This system uses echocardiography data to build a model for diagnosis, reflecting not only whether myocardial ischemia exists, but also the degree and extent of ischemia, and determining whether the ischemia is caused by epicardial coronary artery or coronary microvascular lesions. This significantly improves the diagnostic efficiency of myocardial ischemia while maintaining diagnostic accuracy. However, the traditional myocardial acoustic contrast imaging method used in this patent application has a low frame rate. When performing data fitting analysis, the time resolution of the rising limb of the curve reflecting the contrast agent's entry into the myocardium and its steady-state distribution is insufficient, making it difficult to accurately reflect the actual situation at this stage, thus limiting the accuracy of its analysis and diagnosis. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] One of the technical problems that this invention aims to solve is that existing intelligent diagnostic systems for myocardial ischemia based on ultrasound data have insufficient temporal resolution in the rising branch of the curve of the contrast agent entering the myocardium and reaching steady-state distribution during myocardial acoustic contrast imaging, resulting in a lack of sufficient spatiotemporal information of the rising branch to establish an accurate intelligent diagnostic model.

[0007] Another technical problem that this invention aims to solve is that existing intelligent diagnostic systems for myocardial ischemia cannot fully utilize the spatiotemporal information of myocardial motion and mechanics, resulting in existing intelligent diagnostic models being unable to accurately assess coronary artery blood flow reserve parameters.

[0008] (II) Technical Solution

[0009] To address the aforementioned technical problems, this invention proposes an intelligent diagnostic system for myocardial ischemia, comprising:

[0010] The coronary blood flow state calculation module is used to establish and train a machine learning model and use the machine learning model to calculate the coronary blood flow state. The machine learning model uses high frame rate cardiac ultrasound data as input data, and the coronary arteries include large coronary vessels and coronary microvessels.

[0011] High frame rate cardiac ultrasound data acquisition module, used to acquire high frame rate cardiac ultrasound data;

[0012] The coronary artery lesion judgment module is used to determine whether myocardial ischemia exists based on the coronary artery blood flow status calculated by the machine learning model, and to determine whether the lesion causing myocardial ischemia is a large coronary artery, a microcoronary artery, or both when myocardial ischemia exists.

[0013] According to a preferred embodiment of the present invention, the high frame rate cardiac ultrasound data includes high frame rate myocardial acoustic contrast data, which includes blood perfusion parameters at different time stages within the cardiac cycle.

[0014] According to a preferred embodiment of the present invention, the blood perfusion parameters at different time stages within the cardiac cycle include the blood perfusion rate during systole, the blood perfusion rate during diastole, the blood perfusion duration during systole, and / or the blood perfusion duration during diastole.

[0015] According to a preferred embodiment of the present invention, the diastolic phase includes at least two time phases, which are divided according to at least one of blood flow velocity, the upward or downward trend of blood flow velocity, and the duration of the upward or downward trend of blood flow velocity.

[0016] According to a preferred embodiment of the present invention, the systolic phase includes one time stage, which is a period of steady increase in blood perfusion rate; the diastolic phase includes three time stages, namely a first diastolic phase, a second diastolic phase, and a third diastolic phase, wherein the first diastolic phase is a period of rapid increase in blood perfusion rate, the second diastolic phase is a period of slow decrease in blood perfusion rate, and the third diastolic phase is a period of rapid decrease in blood perfusion rate.

[0017] According to a preferred embodiment of the present invention, the high frame rate cardiac ultrasound data acquisition module is further configured to segment the high frame rate cardiac ultrasound data into high frame rate cardiac ultrasound data of myocardial segments; the input data of the machine learning model established by the coronary blood flow state calculation module is the high frame rate cardiac ultrasound data of each myocardial segment.

[0018] According to a preferred embodiment of the present invention, the cardiovascular lesion judgment module is further used to calculate the ischemia degree information of each myocardial segment, and calculate the probability that the main cause of myocardial ischemia is coronary macrovascular lesion and / or coronary microvascular lesion based on the coronary artery blood flow status and the ischemia degree information of each myocardial segment.

[0019] According to a preferred embodiment of the present invention, it further includes a myocardial ischemia analysis and display module for displaying the ischemia degree information of each myocardial segment.

[0020] According to a preferred embodiment of the present invention, the coronary artery blood flow status is a score of at least one of the blood flow parameters of the coronary macrovessels and coronary microvessels, wherein the blood flow parameters of the coronary macrovessels and / or coronary microvessels include: the degree of stenosis of the coronary macrovessels and / or coronary microvessels, the rate of blood flow in the coronary macrovessels and / or coronary microvessels, and / or the degree of blood flow resistance in the coronary macrovessels and / or coronary microvessels.

[0021] According to a preferred embodiment of the present invention, the high frame rate cardiac ultrasound data acquisition module includes a high frame rate beamforming unit, and further includes a high frame rate myocardial acoustic angiography unit, a high frame rate coronary blood flow and spectral Doppler module unit and / or a high frame rate myocardial motion analysis unit that can cooperate with the high frame rate beamforming unit.

[0022] The high frame rate myocardial acoustic contrast unit is used to present the entire process of contrast agent entering the myocardium, steady-state distribution, redistribution, and elution from the myocardium in real time at a high frame rate.

[0023] The high frame rate coronary artery blood flow and spectral Doppler module unit is used to acquire cardiac chamber and valve blood flow information at a high frame rate, and to display, acquire and analyze coronary artery blood flow.

[0024] The high frame rate myocardial motion analysis unit is used to acquire spatiotemporal information on myocardial motion and / or mechanical characteristics.

[0025] According to a preferred embodiment of the present invention, the coronary blood flow state calculation module includes a data access unit, a data storage unit, and a model unit;

[0026] The data access unit is used to receive input data from the outside for model training and / or model computation;

[0027] The data storage unit is used to store data used during model training and computation;

[0028] The model unit is used to build and train a machine learning model, and to use the machine learning model to calculate the coronary artery blood flow status.

[0029] According to a preferred embodiment of the present invention, the model unit includes a feature extraction subunit, a model building and training subunit, and a model computation subunit.

[0030] The feature extraction subunit is used to obtain whole-heart structural features, systolic and diastolic function features, motion and strain features, and coronary artery structural blood flow features based on the high frame rate cardiac ultrasound data. The above feature data are connected to generate fused features.

[0031] The model building and training subunit is used to build and train machine learning models.

[0032] The model calculation subunit is used to input parameter data into the machine learning model established and trained by the model building and training subunit to calculate the predicted value of coronary artery blood flow status.

[0033] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0034] A machine learning model is established and trained, and the machine learning model is used to calculate the coronary artery blood flow status. The machine learning model uses high frame rate cardiac ultrasound data as input data, and the coronary arteries include large coronary vessels and coronary microvessels.

[0035] Acquire high frame rate cardiac ultrasound data;

[0036] Based on the acquired high-frame-rate cardiac ultrasound data and the coronary artery blood flow status calculated by the machine learning model, it is determined whether myocardial ischemia exists, and if myocardial ischemia exists, it is determined whether the lesion causing myocardial ischemia is a large coronary artery, a microcoronary artery, or both.

[0037] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps:

[0038] A machine learning model is established and trained, and the machine learning model is used to calculate the coronary artery blood flow status. The machine learning model uses high frame rate cardiac ultrasound data as input data, and the coronary arteries include large coronary vessels and coronary microvessels.

[0039] Acquire high frame rate cardiac ultrasound data;

[0040] Based on the acquired high-frame-rate cardiac ultrasound data and the coronary artery blood flow status calculated by the machine learning model, it is determined whether myocardial ischemia exists, and if myocardial ischemia exists, it is determined whether the lesion causing myocardial ischemia is a large coronary artery, a microcoronary artery, or both.

[0041] (III) Technical Effects

[0042] This invention is an intelligent diagnostic system for myocardial ischemia based on high frame rate cardiac ultrasound data. Compared with existing technologies, it can obtain richer data on coronary artery blood flow status and has a high accuracy in predicting the presence and cause of myocardial ischemia.

[0043] This invention can build a machine learning model based solely on ultrasound data. Compared to existing technologies, the training data required is more readily available and simpler, while still achieving high prediction accuracy. Attached Figure Description

[0044] Figure 1 This is a modular architecture diagram of the intelligent diagnostic system for myocardial ischemia of the present invention.

[0045] Figure 2 This is a functional unit architecture diagram of the high frame rate cardiac ultrasound data acquisition module of the intelligent diagnostic system for myocardial ischemia of the present invention.

[0046] Figure 3 This is a functional unit architecture diagram of the coronary blood flow status calculation module of the intelligent diagnostic system for myocardial ischemia of the present invention.

[0047] Figure 4 This is a module architecture diagram of an intelligent diagnostic system for myocardial ischemia, including a myocardial ischemia analysis and presentation module.

[0048] Figure 5 It is coronary artery blood flow data obtained from ordinary ultrasound, showing the relationship between blood flow velocity and time.

[0049] Figure 6The coronary artery blood flow data obtained by high frame rate ultrasound in one embodiment of the present invention also shows the relationship between blood flow velocity and time in the blood vessels.

[0050] Figure 7 This is a comparative diagram of blood flow status data of coronary artery lesions and / or microvascular lesions obtained according to an embodiment of the present invention.

[0051] Figure 8 This is a graph of myocardial perfusion function.

[0052] Figure 9 It is an actual high frame rate cardiac ultrasound image and the corresponding AIF curve.

[0053] Figure 10 This is a comparison chart of the traditional coronary artery blood flow spectrum and the coronary artery blood flow spectrum of this invention.

[0054] Figures 11 to 19 This is a diagram showing cardiac ultrasound data according to an embodiment of the present invention.

[0055] Figure 20 This is a diagram illustrating the results calculated by the diagnostic system for myocardial ischemia in a first case according to an embodiment of the present invention.

[0056] Figures 21 to 23 This is a coronary angiography image of a first case according to an embodiment of the present invention.

[0057] Figures 24 to 32 This is a diagram showing cardiac ultrasound data according to an embodiment of the present invention.

[0058] Figure 33 This is a diagram illustrating the results calculated by the diagnostic system for myocardial ischemia in a first case according to an embodiment of the present invention.

[0059] Figures 34 to 36 This is a coronary angiography image of a first case according to an embodiment of the present invention. Detailed Implementation

[0060] Coronary arteries include epicardial coronary arteries (diameter > 400 μm), anterior arterioles (diameter < 400 μm), arterioles (diameter < 100 μm), and capillaries (diameter < 10 μm). Among these, the epicardial coronary arteries are primarily responsible for blood flow, while the anterior arterioles and arterioles belong to the coronary microvessels. For ease of description, in this invention, the epicardial coronary arteries are referred to as the large coronary vessels, and the anterior arterioles and arterioles are referred to as the coronary microvessels.

[0061] The overall idea of ​​this invention is to propose a system that uses high-frame-rate cardiac ultrasound data to build a machine learning model, ultimately achieving intelligent diagnosis of myocardial ischemia. It should be noted that the system of this invention refers to a hardware-based electronic system with a certain computing capability. It can be a single device such as a desktop computer, laptop, tablet, or mobile terminal, or a service network consisting of a server or server cluster and clients. As a service network, it can be deployed locally or in the cloud. To achieve the intelligent diagnosis of this invention, relevant software is typically installed on the server or terminal device to perform related operations, thereby enabling the hardware on which the system of this patent is based to realize the relevant functions.

[0062] Although the present invention does not exclude the use of software to implement the calculation steps, the present invention as a whole excludes those solutions that are not based on any hardware but only contain pure software, and therefore should meet the requirements of the subject matter of patent protection.

[0063] For ease of explanation, the system of this invention can be divided into different modules. It should be noted that the modules in this invention should also be understood as hardware modules with specific functions. A hardware module can execute software, but it is not limited to software modules. It should also be understood that different modules can be implemented by different hardware or by the same hardware.

[0064] Figure 1 This is a modular architecture diagram of the intelligent diagnostic system for myocardial ischemia of the present invention, as shown below. Figure 1 As shown, the system of the present invention includes a high frame rate cardiac ultrasound acquisition module 1, a coronary artery blood flow state calculation module 2, and a coronary artery lesion judgment module 3. One of the key modules is the coronary artery blood flow state calculation module 2, which is used to establish and train a machine learning model and use this machine learning model to calculate the coronary artery blood flow state.

[0065] As previously stated, in this invention, the coronary arteries include the large coronary vessels and the small coronary vessels; therefore, the coronary artery blood flow status refers to the blood flow status of the large coronary vessels and the small coronary vessels. More specifically, the coronary artery blood flow status refers to the blood flow parameters of the large coronary vessels and the small coronary vessels. Furthermore, for ease of calculation, these parameters can be converted into scoring values.

[0066] According to a preferred embodiment of the present invention, the scoring values ​​of the coronary artery blood flow parameters include at least one of the following: the degree of stenosis of the coronary macrovessels and coronary microvessels, the rate of blood flow in the coronary macrovessels and coronary microvessels, and the degree of blood flow resistance in the coronary macrovessels and coronary microvessels.

[0067] Unlike existing technologies, the machine learning model of this invention directly uses high-frame-rate cardiac ultrasound data as input data, whereas previously it was necessary to use CT and MRI technologies together. Clearly, the input data of this invention is more singular, easier to obtain, and more repeatedly examined, and it eliminates the need to fuse CT and MRI data for computation, which greatly improves the convenience of the machine learning model architecture.

[0068] To acquire high frame rate cardiac ultrasound data, such as Figure 1 As shown, the system of the present invention includes a high frame rate cardiac ultrasound data acquisition module 1. Figure 2 This is a functional unit architecture diagram of the high frame rate cardiac ultrasound data acquisition module 1 of the intelligent diagnostic system for myocardial ischemia of the present invention, as shown below. Figure 2 As shown, the high frame rate cardiac ultrasound data acquisition module 1 includes multiple functional units: a high frame rate beamforming unit 11, a high frame rate myocardial acoustic contrast unit 12, a high frame rate coronary artery blood flow and spectral Doppler module unit 13, and a high frame rate myocardial motion analysis unit 14. The high frame rate myocardial acoustic contrast unit 12, the high frame rate coronary artery blood flow and spectral Doppler module unit 13, and the high frame rate myocardial motion analysis unit 14 are all functional units that cooperate with the high frame rate beamforming unit 11. In this invention, a unit or functional unit refers to a sub-module that constitutes a functional module and has relatively independent functions; it can be a hardware unit or a unit combining hardware and software. The high frame rate beamforming unit 11 is a beamforming unit that supports high frame rate myocardial acoustic contrast, and typically includes an array of multiple elements. Each time an ultrasound beam is transmitted, all or some of the elements of the high frame rate beamforming unit 11 participate in the transmission of the ultrasound beam. These array elements involved in the ultrasonic beam emission are excited by the emission pulse and emit ultrasonic waves. These ultrasonic waves superimpose during propagation to form a synthetic ultrasonic beam that is emitted to the scanned object.

[0069] The high-frame-rate myocardial acoustic contrast unit 12 features a low-mechanical-index imaging mode with adjustable mechanical index and is equipped with high-mechanical-index pulse delivery. After observing the acoustic contrast agent within the cardiac field of view and delivering high-mechanical-index pulses, it allows for real-time observation of the entire process of contrast agent entry into the myocardium, steady-state distribution, redistribution, and myocardial washout at a high frame rate. Traditional myocardial acoustic contrast imaging suffers from low frame rates, resulting in insufficient temporal resolution in the rising limb of the curve reflecting the contrast agent's entry into the myocardium and steady-state distribution during data fitting analysis, making it difficult to accurately reflect information from this stage. The high-frame-rate myocardial acoustic contrast unit 12 of this invention solves this problem, providing richer spatiotemporal information about this rising limb.

[0070] In a preferred embodiment, the high frame rate myocardial acoustic contrast data of the present invention can support the calculation of blood perfusion parameters at different stages of the cardiac cycle. For example, blood perfusion parameters at different stages of the cardiac cycle include blood perfusion rates during systole and diastole, as well as blood perfusion durations during systole and diastole.

[0071] The high-frame-rate coronary blood flow and spectral Doppler unit 13 can not only acquire information on blood flow in the heart chambers and valves in traditional echocardiography at a high frame rate, but also display, acquire and analyze coronary blood flow, enabling more accurate assessment of coronary blood flow reserve parameters, such as CFR (coronary flow reserve) and CFVR (coronary flow velocity reserve).

[0072] The high frame rate myocardial motion analysis unit 14 can obtain richer spatiotemporal information on myocardial motion and mechanical characteristics. Myocardial motion characteristic analysis, such as strain, reflects the relative degree and magnitude of changes in myocardial contraction, relaxation, or rotation during different phases of the cardiac cycle. The spatiotemporal information on myocardial motion and mechanical characteristics includes both temporal and spatial information. The spatial information includes, but is not limited to: 1. the spatial distribution of myocardial blood flow from the epicardium and coronary arteries through successive vessels into the myocardium; 2. the distribution of myocardial blood flow in different segments of the ventricle from the base to the apex; 3. the distribution of motion characteristics in different segments of the myocardium. The temporal information includes, but is not limited to: 1. information on the changes in motion characteristics of different segments of the myocardium during the cardiac cycle; 2. the temporal changes in myocardial blood flow from the aorta into the myocardium and from the epicardium into the myocardium.

[0073] As can be seen from the functions of the above units, the high frame rate cardiac ultrasound data of the present invention includes high frame rate myocardial acoustic contrast data and at least one of the following: high frame rate coronary artery blood flow data, high frame rate myocardial strain data, whole heart structural feature data, and systolic and diastolic function data.

[0074] Therefore, the input data of the coronary blood flow state calculation module 2 of the present invention is high frame rate cardiac ultrasound data including the above data. This rich multi-dimensional data provides many details of coronary blood flow and can provide a data foundation for the establishment of machine models.

[0075] Figure 3 This is a functional unit architecture diagram of the coronary blood flow state calculation module of the intelligent diagnostic system for myocardial ischemia of the present invention. More specifically, as shown in the diagram... Figure 3 As shown, the coronary blood flow state calculation module 2 of the present invention includes functional units: a data access unit 21, a data storage unit 22, and a model unit 23.

[0076] The data access unit 21 is configured to receive input data from an external source for model training or model computation. As previously mentioned, the input data includes at least the high frame rate cardiac ultrasound data. In a preferred embodiment, the input data also includes clinical data, etc.

[0077] Data storage unit 22 is used to store input data, model data, output data, and other data used during model training and computation. Model data includes training datasets, test datasets, model parameters, etc.

[0078] Model unit 23 is the core unit of coronary blood flow state calculation module 2. It can be further divided into multiple sub-units according to different functions, including feature extraction sub-unit 231, model building and training sub-unit 232, model calculation sub-unit 233, etc.

[0079] Feature extraction subunit 231 is configured to obtain whole-heart structural features, systolic and diastolic function features, motion and strain features, and coronary artery structural blood flow features based on high-frame-rate cardiac ultrasound data. These feature data are then concatenated to generate fused features. Furthermore, the myocardial structural blood flow features include myocardial structural features and high-frame-rate perfusion features of the myocardium.

[0080] It should be noted that deep learning, machine learning, and other methods can be applied in the above feature extraction process.

[0081] Traditional machine learning methods include Principal Component Analysis (PCA) for feature dimensionality reduction in ultrasound. Deep learning methods are primarily based on architectures using Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). These include: 1. 3D Convolutional Neural Networks (3D CNNs), which perform convolution operations simultaneously in time and space, effectively processing the spatiotemporal features of cardiac ultrasound videos and extracting useful structural and functional features of the heart; 2. Convolutional-Long Short-Term Memory Networks (CNN-LSTM), which captures the temporal features of ultrasound through a Recurrent Neural Network (LSTM) while simultaneously capturing the spatial features of the video; 3. Temporal Attention Convolutional Networks (TACNs), which add a temporal attention mechanism to the traditional convolutional neural network, enhancing the model's ability to capture temporal information.

[0082] Furthermore, coronary structural blood flow characteristic data includes coronary artery characteristics and coronary blood flow reserve parameter calculation characteristics, including but not limited to CFR and CFVR.

[0083] The model building and training subunit 232 is used to build and train a machine learning model, wherein the machine learning model can be an existing classification model or a neural network model.

[0084] Traditional machine learning classification models include Support Vector Machines (SVM), Random Forest Classifier, and XGBoost (Extreme Gradient Boosting) classifier. In addition to the models mentioned above that can be directly used for classification, neural network models can also utilize simple feed-forward neural network architectures.

[0085] To train a model, a training dataset is needed, and sometimes validation and test datasets are also required. The data format of the validation and test datasets is similar to that of the training dataset. The training dataset includes multiple training samples and typically requires a large amount of data. One of the key differences between this invention and existing technologies is the training sample data in the training dataset. The training sample data includes input parameter data and label data (annotated data).

[0086] The input parameter data of this invention includes at least high frame rate cardiac ultrasound data, and preferably also includes clinical data. As mentioned above, the high frame rate cardiac ultrasound data includes at least one of the following: high frame rate coronary artery blood flow data, high frame rate myocardial blood flow data, high frame rate myocardial strain data, whole-heart structural feature data, and systolic and diastolic function data.

[0087] The label data of this invention refers to the data on coronary artery blood flow status mentioned above, including blood flow parameters of the coronary macrovessels and coronary microvessels. For ease of training, the blood flow parameters of the coronary macrovessels and coronary microvessels are converted into scoring values. The label data constituting the training dataset comes from data observed during patient surgery in historical medical records. As a preferred embodiment of this invention, the scoring values ​​of the coronary artery blood flow parameters include at least one of the following: the degree of stenosis of the coronary macrovessels and coronary microvessels, the rate of blood flow in the coronary macrovessels and coronary microvessels, and the degree of blood flow resistance in the coronary macrovessels and coronary microvessels. As another extended embodiment of this invention, the coronary artery blood flow parameters can also be other parameters from historical medical records that can evaluate the status of coronary arteries or blood flow.

[0088] The model building and training subunit 232 inputs each data point (including parameter data and label data) from the training dataset into the machine learning model. Based on the weights of the obtained parameter data in the model, scores are assigned to the parameter data. Training stops when the model reaches a predetermined metric, and validation and testing are then performed. When to stop model training can be determined by considering factors such as validation set error, loss function value, and overfitting.

[0089] The model calculation subunit 233 uses the machine learning model established and trained by the model building and training subunit 232 to input parameter data into the model to calculate predicted values ​​of coronary artery blood flow status, including predicted values ​​of blood flow parameters of coronary macrovessels and coronary microvessels. Specifically, these are predicted values ​​of the degree of stenosis of coronary macrovessels and coronary microvessels, the speed of blood flow in coronary macrovessels and coronary microvessels, and the degree of resistance in coronary macrovessels and coronary microvessels.

[0090] The system of this invention also includes a coronary artery lesion judgment module 3, which determines whether myocardial ischemia exists based on the coronary artery blood flow status obtained by the coronary blood flow status calculation module 3. Specifically, it is used to determine whether myocardial ischemia exists based on the coronary artery blood flow status calculated by the machine learning model established and trained by the model unit 23 of the coronary blood flow status calculation module 2, and, if myocardial ischemia exists, to determine whether the lesion causing myocardial ischemia is a large coronary vessel, a microcoronary vessel, or both. Specifically, the coronary artery lesion judgment module 3 provides the severity of coronary heart disease corresponding to the predicted value of the coronary artery blood flow status and the probability that the existing myocardial ischemia is caused by large coronary vessel lesions and / or microcoronary vessel lesions, and uses the obtained severity and probability as diagnostic indicators for myocardial ischemia. That is, the coronary vascular lesion judgment module 3 calculates the probability that the main cause of myocardial ischemia is large coronary vessel lesions and / or microcoronary vessel lesions based on the coronary artery and microvascular blood flow status. As mentioned above, the predicted value is a prediction of the degree of stenosis of the coronary macrovessels and coronary microvessels, the speed of blood flow in the coronary macrovessels and coronary microvessels, and the degree of resistance in the coronary macrovessels and coronary microvessels, and the predicted value is a score.

[0091] In a preferred embodiment of the present invention, the high frame rate cardiac ultrasound data acquisition module 1 is further used to segment the high frame rate cardiac ultrasound data into high frame rate cardiac ultrasound data of myocardial segments. Therefore, the input parameter data for the machine learning model established by the coronary blood flow state calculation module 2 is the high frame rate cardiac ultrasound data of each myocardial segment. That is, the input parameter data in each data item of the training dataset used in the present invention for training the model is preferably high frame rate coronary artery blood flow data, high frame rate myocardial blood flow data, and high frame rate myocardial strain data, which are combined based on the characteristics of the whole heart structure and the myocardial segment. The label data consists of the blood flow parameters of the coronary large vessels and coronary microvessels corresponding to the myocardial segments.

[0092] The feature extraction subunit 231 of the model unit 23 in the coronary blood flow state calculation module 2 is used to further obtain myocardial layer feature data, myocardial layer structural blood flow feature data, and myocardial layer motion and strain characteristics for each segment. Thus, the trained model can predict the degree of stenosis of the large and small coronary vessels, the speed of blood flow in the large and small coronary vessels, and the resistance of the large and small coronary vessels in each myocardial segment, thereby predicting the type and specific location of myocardial ischemia.

[0093] That is, the coronary artery lesion judgment module 3 is also used to calculate the ischemia degree information of each myocardial segment, and calculate the probability that the main cause of myocardial ischemia is coronary large vessel lesion and / or coronary micro vessel lesion based on the coronary artery blood flow status and the ischemia degree information of each myocardial segment.

[0094] As a preferred embodiment of the present invention, the intelligent diagnostic system for myocardial ischemia further includes a myocardial ischemia analysis and presentation module, which is used to display the ischemia degree information of each myocardial segment.

[0095] Figure 4 This is a module architecture diagram of an intelligent diagnostic system for myocardial ischemia, including a myocardial ischemia analysis and presentation module. (Example) Figure 4 As shown, the coronary artery lesion judgment module 3 sends the judgment result of whether myocardial ischemia exists and the judgment result of the location of myocardial ischemia to the myocardial ischemia analysis presentation and display module 4 for display. As described, the judgment result preferably includes the judgment result of whether myocardial ischemia exists and whether the diseased vessel causing myocardial ischemia is a large coronary vessel, a microcoronary vessel, or both, with the probability of large coronary vessel lesion and / or microcoronary vessel lesion. Thus, the myocardial ischemia analysis presentation and display module 4 can display the probability of large coronary vessel lesion and / or microcoronary vessel lesion. Preferably, different input parameters and probabilities can be displayed in a pseudo-color format. The display mode can be one of the following, but not limited to the following: The first is based on the industry-recognized 17-segment division method, i.e., the target map or bullseye map, and the above information is superimposed on the map. The second method is a three-dimensional display method, which uses a screen, projection, or AR / VR to display the spatial morphology and segments of the heart, and then overlays the myocardial ischemia information calculated by this patent. At the same time, not only the myocardium and ischemia information of the heart are displayed, but also the coronary arteries that supply specific areas and segments are also displayed in a fused three-dimensional manner, and the large and small coronary vessels that cause ischemia can be displayed together.

[0096] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0097] Figure 5 and Figure 6 This is a comparison chart of coronary artery blood flow data obtained from ordinary ultrasound data and high frame rate ultrasound data according to an embodiment of the present invention. Figure 5 The image shows the blood flow velocity in the blood vessels as obtained by conventional ultrasound. Figure 6 The image shows the blood flow velocity in the blood vessels as obtained through high frame rate ultrasound. From Figure 5 and Figure 6 The comparison shows that high frame rate myocardial acoustic contrast data, due to its richer information acquisition, provides more temporal and spatial information about blood flow perfusion. In the figure, the horizontal axis represents time, and the vertical axis represents signal contrast intensity.

[0098] Figure 7 This is a comparative diagram of blood flow status data of coronary macrovascular lesions and / or coronary microvascular lesions obtained according to an embodiment of the present invention.

[0099] Figure 7 Figure A shows the blood flow status obtained by high-frame-rate echocardiography when there is coronary artery disease. In the figure, P1 and P2 represent the pressures at the proximal and distal ends of the heart, respectively. It can be seen from this figure that the overall blood flow velocity of each coronary microvessel is affected.

[0100] Figure 7 Image B shows the blood flow status obtained by high frame rate ultrasound when there is coronary microvascular disease. In the image, P1 and P2 represent the pressure at the proximal and distal ends, respectively; a and b represent two different microvessels. It can be seen from the image that only the blood flow velocity of individual microvessels is affected.

[0101] The following is an analysis method based on high frame rate cardiac ultrasound data.

[0102] Figure 8 Figure (a) is a theoretical schematic diagram of the arterial input function (AIF) of myocardial perfusion. The function relating the concentration of acoustic contrast agent in the myocardium to its concentration in the feeding arteries is called the arterial input function (AIF). Figure 8 As shown in Figure (a), the horizontal axis represents time t and the vertical axis represents the arterial input function (AIF). Corresponding input pulses can exist at different time points. As shown in the figure, there are two AIF pulses at the two time points t1 and t2.

[0103] Figure 8 Figure (b) is a theoretical schematic diagram of the impulse response function (IRF) of the arterial input function (AIF) of myocardial perfusion. The impulse response function (IRF) represents the relationship between changes in the concentration of acoustic contrast agent in myocardial tissue and the ventricles / aorta. Figure 8As shown in Figure (b), compared to the diameter of the ventricle / aorta, the inflow of acoustic contrast agent at the arterial inlet supplying the myocardium can be regarded as a pulsed injection due to the narrowing of the lumen and the effects of cardiac systole. The concentration changes of acoustic contrast agent in myocardial tissue and ventricle / aorta will occur over time at a certain dynamic ratio, with pulsed injections starting from time t1 and t2.

[0104] Figure 8 Figure (c) illustrates the changes in the arterial input function (AIF) of myocardial perfusion during the cardiac cycle. In actual myocardial perfusion imaging studies, such as... Figure 8 As shown in Figure (c), the arterial input function AIF can be considered as a combination of a series of time-shifted and amplitude-scaled pulse injections.

[0105] Figure 8 Figure (d) illustrates the change in the impulse response function (IRF) corresponding to the myocardial perfusion arterial input function (AIF) during the cardiac cycle. Figure 8 As shown in Figure (d), the IRF generated in the tissue by an actual myocardial perfusion pulse injection can be viewed as a combination of IRFs generated by a series of pulse injections with time shift and amplitude scaling, which is called the tissue enhancement curve. Therefore, after obtaining the actual AIF and tissue enhancement curve, we can apply an iterative curve fitting operation of deconvolution to reconstruct the IRF from the AIF and tissue enhancement curve.

[0106] Because acoustic contrast agents cannot enter cells and remain stable in low-mechanical-index imaging modes, their interaction with tissues is essentially linear and time-invariant. Intravenously injected contrast agents diffuse into the myocardial tissue via the aorta and then flow out through the veins. This system can be considered a linear system, and a key characteristic of linear systems is the superposition principle. Therefore, the process of myocardial perfusion in MCE (myocardial acoustic contrast imaging) can be represented as:

[0107]

[0108] C AIF : TAC in the heart chambers / TAC in the aorta, i.e., arterial input function, TAC represents the time-concentration curve;

[0109] C T TAC in myocardial tissue reflects the concentration of UCA in the ROI (region of interest). UCA refers to ultrasound contrast agent.

[0110] FR(t): The impulse response function of blood flow F, which reflects the dynamic process of UCA through the tissue.

[0111] For each impulse response function, the average time of the UCA through the tissue can be expressed in the following form, i.e., the average transit time (MTT) of R(t) is the area under the curve of R(t):

[0112]

[0113] in, It is the weighted time integral of the response function, representing the sum of the contributions of UCA at each time point; It is the total area of ​​the response function, representing the total response of the contrast agent over the entire time span.

[0114] In the deconvolution model, C AIF and C T The blood flow rate (FR) can be obtained from ultrasound images and then processed using a deconvolution algorithm. The plateau height of the function represents the blood flow rate (F). Using the central volume law, the area under the FR(t) curve represents the blood volume.

[0115] V = F * MTT

[0116] Therefore, it is only necessary to use C AIF (t) and C T By calculating FR(t), parameters such as myocardial blood flow (MBF), blood volume (MBV), and contrast agent mean transit time (MTT) can be calculated.

[0117] Figure 9 This is an actual high frame rate cardiac ultrasound image and its corresponding AIF curve. Figure 9 In the upper right corner of the left image, a high-frame-rate ultrasound image of the aorta is shown. The upper curve in the left image is the AIF curve at point Ao in the high-frame-rate ultrasound image, and the lower curve is the acoustic AIF curve of the myocardial region. Similarly, by placing the region of interest (ROI) in different myocardial layers, high-frame-rate myocardial perfusion information parameters for different myocardial layers can be obtained.

[0118] In this invention, the increased frame rate provides richer coronary artery blood flow spectral information. Therefore, the high-frame-rate cardiac ultrasound data upon which this invention is based is more abundant.

[0119] Figure 10 This is a comparison chart of the traditional coronary artery blood flow spectrum and the coronary artery blood flow spectrum of this invention.

[0120] exist Figure 10 The image above shows a traditional coronary artery blood flow spectrum. It can be seen that the spectral morphology during diastole (Dia) is relatively simple, only able to depict the spectral outline. Figure 10 (The blue curve in the above figure) is used to obtain the maximum speed, average speed, and speed-time integral. Figure 10The figure below is a coronary artery blood flow spectrum based on high frame rate cardiac ultrasound data according to the present invention. The spectral information obtained by high frame rate cardiac ultrasound is richer, including the rate and time of each time period of the cardiac cycle. On the diastolic spectrum shown, four obvious curve inflection points can be seen, corresponding to four velocity change intervals (V1-V4) and corresponding time periods (T1-T4).

[0121] like Figure 10 As shown in the figure below, the blood perfusion rate and duration during systole and diastole can be observed. For example, the blood perfusion of a cardiac cycle can be divided into four phases: systole, first diastole, second diastole, and third diastole. Systole is the phase of steady increase in blood perfusion rate, characterized by a blood perfusion rate V1 and a blood perfusion duration T1; first diastole is the phase of rapid increase in blood perfusion rate, characterized by a blood perfusion rate V2 and a blood perfusion duration T2; second diastole is the phase of slow decrease in blood perfusion rate, characterized by a blood perfusion rate V3 and a blood perfusion duration T3; and third diastole is the phase of rapid decrease in blood perfusion rate, characterized by a blood perfusion rate V4 and a blood perfusion duration T4. The figure also shows that the perfusion rate V1 represents systolic blood flow (MBF), V2 represents rapid diastolic blood flow perfusion, V3 and V4 represent slow diastolic blood flow perfusion, T1 represents systolic blood flow perfusion duration, T2 represents rapid diastolic blood flow perfusion duration, and T3 and T4 represent slow diastolic blood flow perfusion duration.

[0122] Prior to this invention, the characteristics of how epicardial coronary artery or microvascular lesions affect blood perfusion rate had long been neglected, and were difficult to study in practice due to the difficulty in data acquisition. The inventors of this invention were the first to recognize that combining high-frame-rate cardiac ultrasound data with big data and machine learning techniques could effectively solve this problem. They attempted to extract features from segmented blood perfusion rates and durations within the cardiac cycle, and experimentally demonstrated that this segmented extraction of blood perfusion rates and durations within the cardiac cycle is meaningful and can significantly improve the accuracy of diagnosing the cause of myocardial ischemia. In particular, the feature extraction of segmented blood perfusion rates and durations during diastole is a preferred method.

[0123] The above is merely an example; the four different time phases can be determined by the start, end, and inflection points of a four-segment broken line fitted with the blood flow perfusion rate. However, this invention is not limited to the method of dividing different time phases within the cardiac cycle. Based on the inventors' practical experience, as a preferred embodiment, the diastolic phase within the cardiac cycle can generally be divided into at least two time phases, such as two, three, four, or five. These time phases are divided based on at least one of blood flow velocity, the upward or downward trend of blood flow velocity, and the duration of the upward or downward trend of blood flow velocity. Furthermore, as another embodiment, the blood flow perfusion rate obtained from high-frame-rate cardiac ultrasound can also be fitted into a curve, and the parameters of the curve can be used as parameters of the blood flow perfusion rate. Figures 11 to 19 This is a display of cardiac ultrasound data according to an embodiment of the present invention. This embodiment presents data from a first case. Figures 11 to 14 Traditional 2D grayscale cardiac ultrasound images; Figure 15 , Figure 16 Acoustic imaging of the myocardium; Figure 17 A diagram illustrating myocardial blood flow velocity; Figure 18 This is a diagram displaying coronary artery blood flow information and data; Figure 19 This is a graph showing the data from myocardial strain analysis.

[0124] Figure 20 This is a diagram illustrating the results calculated by the diagnostic system for myocardial ischemia in a first case according to an embodiment of the present invention.

[0125] Should Figure 20 Presented in a three-dimensional display, red gradations represent no myocardial ischemia, while blue to purple gradations represent myocardial ischemia. The closer to red, the lower the probability of myocardial ischemia; the closer to purple, the higher the probability of myocardial ischemia. This image indicates severe ischemia in the right coronary artery supply area and moderate ischemia in the left anterior descending artery supply area, with patchy areas of severe ischemia and accompanying microvascular lesions.

[0126] Figure 21 , Figure 22 and Figure 23 This is a coronary angiography image of a first case according to an embodiment of the present invention. Figure 21 , Figure 22 , Figure 23 The images presented were coronary angiography images in head-down, foot-down, and left anterior oblique + head-down positions, illustrating the right coronary artery, left circumflex artery, and left anterior descending artery. The results showed severe stenosis of the right coronary artery and mild stenosis of the left anterior descending and left circumflex arteries. The left anterior descending artery had a free flow rate (FFR) of 0.6 and an intravascular membrane flow rate (IMR) of 70.7, confirming myocardial ischemia in the blood supply areas of both the right coronary artery and the left anterior descending artery, and indicating the presence of both epicardial coronary artery and microvascular lesions. This is consistent with the calculation results of the system of this invention.

[0127] Figures 24 to 32 This is a display of cardiac ultrasound data according to an embodiment of the present invention. This embodiment presents data from a first case. Figures 24 to 26 Traditional 2D grayscale cardiac ultrasound images; Figure 27 , Figure 28 Acoustic imaging of the myocardium; Figure 29 A diagram illustrating myocardial blood flow velocity; Figure 30 , Figure 31 , Figure 32 This is a diagram showing images and data related to coronary artery blood flow.

[0128] Figure 33 This is a diagram illustrating the results calculated by the diagnostic system for myocardial ischemia in a first case according to an embodiment of the present invention.

[0129] Should Figure 33 Presented in bullseye chart format. Red levels represent no myocardial ischemia, while blue to purple levels represent myocardial ischemia. The closer to red, the lower the probability of myocardial ischemia; the closer to purple, the higher the probability of myocardial ischemia. The chart shows a patchy distribution of ischemic areas, suggesting microvascular lesions and mild ischemia, mainly concentrated in the area supplied by the left anterior descending artery.

[0130] Figure 34 , Figure 35 and Figure 36 This is a coronary angiography image of a first case according to an embodiment of the present invention. Figure 34 , Figure 35 and Figure 36 The images shown are coronary angiography images in the left anterior oblique, foot, and head positions, illustrating the right coronary artery, left circumflex artery, and left anterior descending artery. The results indicate mild stenosis in the right coronary artery, left anterior descending artery, and left circumflex artery. The left anterior descending artery has a free flow factor (FFR) of 0.88 and an intravascular membrane length (IMR) of 36.72, confirming myocardial ischemia caused by microvascular lesions. This is consistent with the calculation results of the system of this invention.

[0131] It is evident that high frame rate ultrasound can not only reflect whether myocardial ischemia is present, but also the degree and extent of ischemia, and can determine whether the ischemia is caused by epicardial coronary artery lesions or coronary microvascular lesions. While ensuring diagnostic accuracy, it greatly improves the diagnostic efficiency of myocardial ischemia.

[0132] Those skilled in the art will understand that the above embodiments of the present invention can be provided as systems, computer programs, or electronic devices. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0133] This invention is described with reference to flowchart illustrations and / or block diagrams of systems and computer programs according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0134] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0135] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0136] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0137] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent diagnostic system for myocardial ischemia, characterized in that, include: The coronary blood flow state calculation module is used to establish and train a machine learning model and use the machine learning model to calculate the coronary blood flow state. The machine learning model uses high frame rate cardiac ultrasound data as input data, and the coronary arteries include large coronary vessels and coronary microvessels. High frame rate cardiac ultrasound data acquisition module, used to acquire high frame rate cardiac ultrasound data; The coronary artery lesion judgment module is used to determine whether myocardial ischemia exists based on the coronary artery blood flow status calculated by the machine learning model, and to determine whether the lesion causing myocardial ischemia is a large coronary artery, a microcoronary artery, or both when myocardial ischemia exists.

2. The intelligent diagnostic system for myocardial ischemia as described in claim 1, characterized in that, The high frame rate cardiac ultrasound data includes high frame rate myocardial acoustic contrast data, which includes blood perfusion parameters at different time stages within the cardiac cycle.

3. The intelligent diagnostic system for myocardial ischemia as described in claim 2, characterized in that, The blood perfusion parameters at different time stages within the cardiac cycle include the blood perfusion rate during systole, the blood perfusion rate during diastole, the blood perfusion duration during systole, and / or the blood perfusion duration during diastole.

4. The intelligent diagnostic system for myocardial ischemia as described in claim 3, characterized in that, The diastolic phase includes at least two time phases, which are divided according to at least one of blood flow velocity, the trend of increasing or decreasing blood flow velocity, and the duration of increasing or decreasing blood flow velocity.

5. The intelligent diagnostic system for myocardial ischemia as described in claim 4, characterized in that, The systolic phase includes a time segment during which the blood perfusion rate steadily increases. The diastolic phase includes three time stages, namely the first diastolic phase, the second diastolic phase, and the third diastolic phase. The first diastolic phase is the period of rapid increase in blood perfusion rate, the second diastolic phase is the period of slow decrease in blood perfusion rate, and the third diastolic phase is the period of rapid decrease in blood perfusion rate.

6. The intelligent diagnostic system for myocardial ischemia as described in claim 1, characterized in that, The high frame rate cardiac ultrasound data acquisition module is also used to segment the high frame rate cardiac ultrasound data into high frame rate cardiac ultrasound data of myocardial slice segments. The input data for the machine learning model established by the coronary blood flow state calculation module is high-frame-rate cardiac ultrasound data of each myocardial segment.

7. The intelligent diagnostic system for myocardial ischemia as described in claim 6, characterized in that, The coronary artery lesion judgment module is also used to calculate the degree of ischemia in each myocardial segment, and calculate the probability that the main cause of myocardial ischemia is coronary macrovascular disease and / or coronary microvascular disease based on the coronary artery blood flow status and the degree of ischemia in each myocardial segment.

8. The intelligent diagnostic system for myocardial ischemia as described in claim 7, characterized in that, It also includes a myocardial ischemia analysis and presentation module, which is used to display the ischemia degree information of each myocardial segment.

9. The intelligent diagnostic system for myocardial ischemia as described in claim 1, characterized in that, The coronary artery blood flow status is a score of at least one of the blood flow parameters of the coronary macrovessels and coronary microvessels, and the blood flow parameters of the coronary macrovessels and / or coronary microvessels include: the degree of stenosis of the coronary macrovessels and / or coronary microvessels, the rate of blood flow in the coronary macrovessels and / or coronary microvessels, and / or the degree of blood flow resistance in the coronary macrovessels and / or coronary microvessels.

10. The intelligent diagnostic system for myocardial ischemia as described in claim 1, characterized in that, The high frame rate cardiac ultrasound data acquisition module includes a high frame rate beamforming unit, and also includes a high frame rate myocardial acoustic angiography unit, a high frame rate coronary blood flow and spectral Doppler module unit and / or a high frame rate myocardial motion analysis unit that can cooperate with the high frame rate beamforming unit. The high frame rate myocardial acoustic contrast unit is used to present the entire process of contrast agent entering the myocardium, steady-state distribution, redistribution, and elution from the myocardium in real time at a high frame rate. The high frame rate coronary artery blood flow and spectral Doppler module unit is used to acquire cardiac chamber and valve blood flow information at a high frame rate, and to display, acquire and analyze coronary artery blood flow. The high frame rate myocardial motion analysis unit is used to acquire spatiotemporal information on myocardial motion and / or mechanical characteristics.

Images