Classification method, system, electronic device and medium based on non-invasive pressure volume ring

Abstract

The present invention discloses a classification method, system, electronic device and medium based on non-invasive pressure-volume loop, which relates to the field of artificial intelligence. The method comprises: obtaining a cardiogram of at least one cardiac cycle of a subject to be tested; the cardiogram is a standardized transthoracic echocardiogram; extracting the endocardial boundary of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, and obtaining the left ventricular volume curve of the subject to be tested; determining the left ventricular pressure-volume loop of the subject to be tested by using a time-varying elastic model according to the left ventricular volume curve of the subject to be tested; obtaining the cardiac fluid dynamics parameters of the subject to be tested from the left ventricular pressure-volume loop of the subject to be tested; constructing a classification prediction model by using a classification algorithm; inputting the cardiac fluid dynamics parameters of the subject to be tested into a trained classification prediction model to determine the prediction classification result. The present invention can establish a non-invasive pressure-volume loop according to a cardiac ultrasound image, and perform classification through the non-invasive pressure-volume loop.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a classification method, system, electronic equipment and medium based on a non-invasive pressure-volume ring. Background Art

[0002] Heart failure is a widespread public health problem that carries a considerable risk of morbidity and mortality, greatly threatening physical health. Heart failure, or HF for short, is a comprehensive disease caused by the heart's inability to pump enough blood to the body, usually caused by diabetes, hypertension or other heart diseases. Safe, reliable and rapid diagnosis and risk assessment of heart failure are essential to provide timely and cost-effective treatment.

[0003] As a rapid and accurate method, echocardiography can help noninvasively detect and characterize hemodynamic and morphological abnormalities in heart failure. Pressure-volume loops reflect the cardiac cycle and ventricular properties, and pressure-volume loops have been the gold standard for direct, real-time assessment of cardiac function. The loops contain a wealth of information about cardiac performance, and their analysis can elucidate the pathophysiological mechanisms of heart failure, including preserved heart failure, myocardial and valvular heart disease, etc. Ventriculoarterial coupling, mechanical potential energy, ventricular efficiency, and other parameters obtained from pressure-volume loops can be used to assess cardiac function and provide guidance for clinical diagnosis. However, analysis of pressure-volume loops is often overlooked in the clinic, and it is difficult to guide patient management because both invasive and noninvasive methods are difficult to obtain.

[0004] Currently, the diagnosis of heart failure is mainly based on comprehensive evaluation by experts based on clinical symptoms, signs, laboratory tests and imaging results. However, due to the uneven distribution of medical resources, manual diagnosis is time-consuming and inefficient, and experts' diagnosis of heart failure is difficult to meet clinical needs. Summary of the invention

[0005] The purpose of the present invention is to provide a classification method, system, electronic device and medium based on non-invasive pressure-volume loop, which can establish a non-invasive pressure-volume loop according to cardiac ultrasound images and perform classification through the non-invasive pressure-volume loop.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A classification method based on non-invasive pressure-volume loops, the method comprising:

[0008] Acquiring a cardiogram of at least one cardiac cycle of the subject to be tested; the cardiogram is a standardized transthoracic echocardiogram;

[0009] Extracting the endocardial border of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, to obtain a left ventricular volume curve of the subject to be tested;

[0010] According to the left ventricular volume curve of the subject to be tested, using a time-varying elastic model, determining the left ventricular pressure-volume loop of the subject to be tested;

[0011] Acquiring cardiac fluid dynamics parameters of the subject from the left ventricular pressure-volume loop of the subject;

[0012] Use classification algorithms to build classification prediction models;

[0013] The cardiac fluid dynamics parameters of the subject are input into the trained classification prediction model to determine the prediction classification result.

[0014] Optionally, the training process of the classification prediction model includes:

[0015] Obtaining a cardiogram of at least one cardiac cycle of the subject; the cardiogram is a standardized transthoracic echocardiogram;

[0016] According to the set categories, the cardiogram of at least one cardiac cycle of the subject is labeled with categories to obtain standardized transthoracic echocardiograms of various categories;

[0017] Extracting the endocardial border of the ventricle from the standardized transthoracic echocardiograms of each category to obtain the left ventricular volume curve of each category;

[0018] According to the left ventricular volume curves of the various categories, using a time-varying elastic model, obtaining left ventricular pressure-volume loops of the various categories;

[0019] Acquire cardiac fluid dynamic parameters of various categories from the left ventricular pressure-volume loops of various categories;

[0020] The classification prediction model is trained with the cardiac fluid dynamics parameters of each category as input and the probability of the corresponding category as output to obtain a trained classification prediction model.

[0021] Optionally, taking the cardiac fluid dynamics parameters of each category as input and the probability of the corresponding category as output, training the classification prediction model to obtain the trained classification prediction model specifically includes:

[0022] Dividing the cardiac fluid dynamics parameters of each category into a training set and a validation set;

[0023] Construct multiple initial classification prediction models based on multiple classification algorithms;

[0024] Taking the training set as input, training the multiple initial classification prediction models to obtain multiple trained initial classification prediction models;

[0025] Inputting the validation set into the multiple trained initial classification prediction models to obtain the category probabilities output by the multiple trained initial classification prediction models;

[0026] Obtaining prediction classification results of the multiple trained initial classification prediction models according to the category probabilities output by the multiple trained initial classification prediction models;

[0027] From the multiple trained initial classification prediction models, a trained initial classification prediction model that outputs the predicted classification result with the highest category probability is selected as the trained classification prediction model.

[0028] Optionally, the time-varying elastic model is:

[0029]

[0030] Among them, E(t) is the change of myocardial stiffness at time t represented by time-varying elasticity, P(t) is the ventricular pressure at time t, V(t) is the ventricular volume at time t, and V0 is the ventricular volume at zero pressure.

[0031] Optionally, determining the left ventricular pressure-volume loop of the subject to be tested by using a time-varying elastic model according to the left ventricular volume curve of the subject to be tested specifically includes:

[0032] converting the time-varying elastic model into a double Hill equation;

[0033] The double Hill equation is solved according to the left ventricular volume curve of the subject to be tested to obtain the left ventricular pressure-volume loop of the subject to be tested.

[0034] Optionally, inputting the cardiac fluid dynamics parameters of the subject to be tested into a trained classification prediction model to determine the prediction classification result specifically includes:

[0035] Inputting the cardiac fluid dynamics parameters of the subject to be tested into a trained classification prediction model to obtain the probability of the category output by the trained classification prediction model;

[0036] The category with the highest probability is determined as the predicted classification result.

[0037] The present invention also provides a classification system based on a non-invasive pressure-volume ring, which is applied to the above-mentioned classification method based on a non-invasive pressure-volume ring, and the system comprises:

[0038] An echocardiogram acquisition module, used for acquiring a cardiogram of at least one cardiac cycle of the subject to be tested; the cardiogram is a standardized transthoracic echocardiogram;

[0039] A left ventricular volume curve extraction module, used for extracting the endocardial boundary of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, so as to obtain the left ventricular volume curve of the subject to be tested;

[0040] A left ventricular pressure-volume loop determination module, used to determine the left ventricular pressure-volume loop of the subject to be tested based on the left ventricular volume curve of the subject to be tested and using a time-varying elastic model;

[0041] A cardiac fluid dynamics parameter acquisition module, used for acquiring cardiac fluid dynamics parameters of the subject to be tested from the left ventricular pressure-volume loop of the subject to be tested;

[0042] A model building module is used to build a classification prediction model using a classification algorithm;

[0043] The prediction and classification result determination module is used to input the cardiac fluid dynamics parameters of the subject to be tested into the trained classification prediction model to determine the prediction and classification result.

[0044] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the above-mentioned classification method based on the non-invasive pressure-volume ring.

[0045] The present invention also provides a computer-readable storage medium storing a computer program, wherein the computer program implements the above-mentioned classification method based on non-invasive pressure-volume ring when executed by a processor.

[0046] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0047] The present invention provides a classification method based on non-invasive pressure-volume loop, comprising: obtaining a cardiogram of at least one cardiac cycle of a subject to be tested; the cardiogram is a standardized transthoracic echocardiogram; extracting the endocardial boundary of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, and obtaining the left ventricular volume curve of the subject to be tested; determining the left ventricular pressure-volume loop of the subject to be tested according to the left ventricular volume curve of the subject to be tested using a time-varying elastic model; obtaining the cardiac fluid dynamics parameters of the subject to be tested from the left ventricular pressure-volume loop of the subject to be tested; constructing a classification prediction model using a classification algorithm; inputting the cardiac fluid dynamics parameters of the subject to be tested into the trained classification prediction model to determine the predicted classification result. The present invention extracts the endocardial boundary of the ventricle from the ultrasonic image of the subject to be tested, determines the non-invasive pressure-volume loop using a time-varying elastic model, and obtains the cardiac fluid dynamics parameters of the subject to be tested, and classifies the cardiac fluid dynamics parameters of the subject to be tested using a classification algorithm to obtain a classification result. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 A flow chart of a classification method based on non-invasive pressure-volume loop provided by the present invention;

[0050] Figure 2 is a schematic diagram of a non-invasive pressure-volume loop;

[0051] Figure 3 A module diagram of a classification system based on a non-invasive pressure-volume loop provided by the present invention.

[0052] Explanation of symbols: 1- echocardiography acquisition module, 2- left ventricular volume curve extraction module, 3- left ventricular pressure-volume loop determination module, 4- cardiac fluid dynamics parameter acquisition module, 5- model construction module, 6- prediction and classification result determination module. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] The purpose of the present invention is to provide a classification based on a non-invasive pressure-volume loop, which can establish a non-invasive pressure-volume loop according to a cardiac ultrasound image and perform classification through the non-invasive pressure-volume loop.

[0055] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Embodiment 1

[0057] like Figure 1 As shown, the present invention provides a classification method based on non-invasive pressure-volume ring, the method comprising:

[0058] Step S1: obtaining a cardiogram of at least one cardiac cycle of the subject; the cardiogram is a standardized transthoracic echocardiogram.

[0059] In practical applications, the collected 2D-A4C views should ensure sufficient image quality (ie, no obvious border loss) and include at least one complete cardiac cycle.

[0060] Step S2: extracting the endocardial border of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, and obtaining the left ventricular volume curve of the subject to be tested.

[0061] In practical applications, the endocardial boundary of the ventricle is extracted frame by frame from the cardiac ultrasound image sequence. cycle ) is determined by the complete cardiac cycle, including systole and diastole. End-systolic time (T ES ) is defined as the time between the minimum LV volume and the end-diastolic time at the maximum LV volume (T ED ). The left ventricular end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV) and volume-time curve are automatically obtained from cardiac ultrasound images. The maximum value of the reference pressure defines the left ventricular systolic pressure peak (LVP systole ), the last data point defines the peak left ventricular diastolic pressure (LVP distole ).

[0062] Step S3: According to the left ventricular volume curve of the subject to be tested, the left ventricular pressure-volume loop of the subject to be tested is determined by using a time-varying elastic model. Specifically, the time-varying elastic model is:

[0063]

[0064] Among them, E(t) is the change of myocardial stiffness at time t represented by time-varying elasticity, P(t) is the ventricular pressure at time t, V(t) is the ventricular volume at time t, and V0 is the ventricular volume at zero pressure.

[0065] In practice, the left ventricular pressure-volume (PV) loop is derived from the simultaneous measurement of intraventricular pressure and volume found during the cardiac cycle, e.g. Figure 2 As shown, in the PV loop, the cardiac cycle is divided into different phases. In the lower left corner (red dots), the mitral valve opens and the ventricle begins to fill rapidly during diastole. This phase ends with atrial contraction at the blue dots of the loop, accompanied by the left ventricular end-diastolic pressure (Ped) and end-diastolic volume (EDV) (lower right corner). Then, the mitral valve closes and the isovolumetric contraction phase begins. When the intraventricular pressure exceeds the aortic pressure, the aortic valve opens (upper right corner, red dots) and the ventricle expels blood until it reaches the blue dots of the loop, accompanied by the left ventricular end-systolic volume (ESV) and end-systolic pressure (Pes) (upper left corner). When the aortic valve closes, the intraventricular pressure decreases without any change in volume (isovolumetric relaxation phase). When the intraventricular pressure drops below the atrial pressure, the mitral valve opens and the cardiac cycle begins again.

[0066] As a specific implementation, S3 specifically includes:

[0067] Step S31: converting the time-varying elastic model into a double Hill equation.

[0068] In practical applications, according to Su Ga H's time-varying elastic model:

[0069] The time-varying elastic model is transformed into a double Hill equation that can be solved with a finite number of independent parameters.

[0070] Step S32: solving the double Hill equation according to the left ventricular volume curve of the subject to be tested to obtain the left ventricular pressure-volume loop of the subject to be tested.

[0071] The time-varying elastic curve can be transformed into the analytical double Hill equation:

[0072]

[0073] in, By parameter E max 、E min , τ1, τ2, n1 and n2, solve the double Hill equation. Use the Nelder-Mead simplex method to optimize the parameters τ1, τ2, n1 and n2. max and E min The left ventricular peak systolic pressure and end-diastolic pressure (LVP systole , LVP distole ). All parameters used in the time-varying elastic model are shown in Table 1. Since the volume at zero pressure is close to zero, V0 is set to zero.

[0074] Table 1 Double Hill equation parameters

[0075]

[0076]

[0077] Among them, LVP systole , LVP distole Determines the amplitude of the elastic curve, LVP systole It can be approximated by brachial artery pressure:

[0078] Among them, SBP is systolic blood pressure, DBP is diastolic blood pressure, and LVP is distole Approximately 5mmHg.

[0079] Using the time-varying model, the ventricular pressure is expressed as:

[0080] P(t)=E(t)×(V(t)-V0); According to V(t), the volume-time curve can be obtained, thereby obtaining the PV loop.

[0081] Step S4: Acquire cardiac fluid dynamics parameters of the subject from the left ventricular pressure-volume loop of the subject.

[0082] Specifically, hemodynamics and LV function were quantified by parameters such as end-diastolic and end-systolic volumes, stroke volume (SV), cardiac output (CO), left ventricular ejection fraction (LVEF), cardiac output, stroke work (SW), mechanical potential (PE), ventricular efficiency (η), mean external power (MEP), energy per ejected volume (EEV), effective arterial elastance (Ea), and ventriculo-arterial coupling (VAcoupling). End-diastolic volume, end-systolic volume, stroke volume, and ejection fraction were obtained from echocardiographic images.

[0083] Among them, stroke volume (SV) is the amount of blood pumped out of the left ventricle each time the heart contracts. SV can be calculated as the difference between the left ventricular end-diastolic volume (EDV) and the left ventricular end-systolic volume (ESV): SV = EDV-ESV.

[0084] Cardiac output (CO) is the volume of blood pumped by the heart through the systemic circulation over a period of time, measured in liters per minute. There are a variety of parameters that can be used to fully assess cardiac output, but one of the more traditional methods involves multiplying the product of heart rate (HR) and stroke volume: CO = SV·HR.

[0085] The calculation of left ventricular ejection fraction (EF) involves dividing the stroke volume by EDV and is considered an important indicator for assessing systolic and diastolic heart failure, where EF = SV / EDV.

[0086] Stroke work SW is the area enclosed by the PV loop and represents the external work of the ventricle to eject blood into the aorta and is used to assess ventricular function. Mechanical potential energy (PE) is the internal energy that the heart overcomes to eject blood. Therefore, the total mechanical energy generated by ventricular contraction is the sum of SW and PE, called pressure volume area (PVA), PVA = SW + PE. PVA has been shown to be linearly related to myocardial oxygen consumption (MVO2). It can be quantified as the area between the end-systolic PV relationship (ESPVR) and the end-diastolic PV relationship (EDPVR) and the leftmost side of the PV loop.

[0087] The beat-to-beat ventricular efficiency can be described as the ratio between useful ventricular mechanical work (SW) and the estimated O2 consumption by PVA:

[0088] The mean external power (MEP) delivered by the left ventricle can be expressed as:

[0089] The energy per ejected volume (EEV) is calculated as follows:

[0090] The EDPVR is connected to the end-systolic PV point or the point of maximum elastance. The slope of the ESPVR represents the maximum ventricular elastance (Ees), which provides an indication of the myocardial contractility. Therefore, when the ventricular contractility changes, the Ees changes proportionally. The ESPVR and EDPVR define the boundaries of the PV loop.

[0091] In addition, Sunagawa et al. described the arterial system characteristics as effective arterial elastance (Ea), which represents all extracardiac forces opposing ventricular ejection or arterial load. Ea can be approximated as the slope of the arterial end-systolic pressure-stroke volume relationship on the PV loop, which can be calculated as: Therefore, the interaction between cardiac contractility and the arterial system can be easily analyzed as the ratio of Ees to Ea, known as ventriculo-arterial coupling (VA coupling).

[0092] Step S5: Use the classification algorithm to build a classification prediction model.

[0093] Step S6: inputting the cardiac fluid dynamics parameters of the subject into the trained classification prediction model to determine the prediction classification result.

[0094] In actual application, according to echocardiography, the pressure-volume loop is used to obtain 15 cardiac fluid dynamic parameters of the subjects, including SBP, DBP, EDV, ESV, SW, EF, Mechanical PE, Efficiency, MEP, SV, CO, EEV, Ees, Ea, and VA coupling.

[0095] As a specific implementation, S6 specifically includes:

[0096] Step S61: inputting the cardiac fluid dynamics parameters of the subject into the trained classification prediction model to obtain the probability of the category output by the trained classification prediction model.

[0097] Step S62: Determine the category with the highest probability as the predicted classification result.

[0098] Specifically, the output result is the probability of belonging to each type of heart failure, which ranges from 0 to 1, and the heart failure disease type corresponding to the maximum probability is selected as the final classification prediction result.

[0099] In addition, the training process of the classification prediction model includes:

[0100] Step S01: obtaining a cardiogram of at least one cardiac cycle of a subject; the cardiogram is a standardized transthoracic echocardiogram.

[0101] Step S02: According to the set categories, the cardiogram of at least one cardiac cycle of the subject is labeled with categories to obtain standardized transthoracic echocardiograms of various categories.

[0102] In practical applications, the cardiac ultrasound image data are manually annotated in combination with clinical research, and the data are divided into four types: no heart failure, heart failure with preserved ejection fraction, heart failure with intermediate-reduced ejection fraction, and heart failure with reduced ejection fraction.

[0103] Step S03: extracting the endocardial border of the ventricle from the standardized transthoracic echocardiograms of each category to obtain the left ventricular volume curve of each category.

[0104] Step S04: According to the left ventricular volume curves of each category, using a time-varying elastic model, obtain the left ventricular pressure-volume loops of each category.

[0105] Step S05: Acquire cardiac fluid dynamic parameters of each category from the left ventricular pressure-volume loops of each category.

[0106] Step S06: taking the cardiac fluid dynamics parameters of each category as input and the probability of the corresponding category as output, training the classification prediction model to obtain a trained classification prediction model.

[0107] S06 specifically includes:

[0108] Step S061: Divide the cardiac fluid dynamics parameters of each category into a training set and a validation set.

[0109] Step S062: construct multiple initial classification prediction models based on multiple classification algorithms.

[0110] In practical applications, eight classification algorithms (Logistic regression, Decision Tree, Random Forests, SVC, k-NN, GBM, XGBoost, and LightGBM) were selected to establish corresponding initial classification prediction models.

[0111] Step S063: using the training set as input, training the multiple initial classification prediction models to obtain multiple trained initial classification prediction models.

[0112] In practical applications, heart failure classification training is performed, and the training set is used to train the initial classification prediction model.

[0113] Step S064: inputting the verification set into the multiple trained initial classification prediction models to obtain the category probabilities output by the multiple trained initial classification prediction models.

[0114] Step S065: Obtaining the predicted classification results of the multiple trained initial classification prediction models according to the category probabilities output by the multiple trained initial classification prediction models.

[0115] Step S066: From the multiple trained initial classification prediction models, select the trained initial classification prediction model with the highest category probability for outputting the predicted classification result as the trained classification prediction model.

[0116] In practical applications, the weight of the result with the highest accuracy in the validation set is selected as the final hyperparameter, and the model is saved as the trained classification prediction model. It has been verified that the accuracy of the four-category heart failure classification of the XGBoost algorithm on the test data set reaches a maximum of 88.89%.

[0117] In order to perform detailed heart failure classification, 15 feature parameters were input into the trained classification prediction model. In order to compare the impact of the classification algorithm on the classification results, the results of XGBoost and other classification algorithms were compared, as shown in Table 2.

[0118] Table 2 Average test results of different algorithms for 100 times

[0119]

[0120]

[0121] The classification method based on the non-invasive pressure-volume loop provided by the present invention has the advantages of safe, reliable and rapid diagnosis, improves the diagnostic efficiency and reduces the workload of doctors. The method can automatically perform classification diagnosis of heart failure and can provide doctors with reliable and objective auxiliary diagnosis results.

[0122] Embodiment 2

[0123] In order to execute the method corresponding to the above-mentioned embodiment 1 to achieve the corresponding functions and technical effects, a classification system based on a non-invasive pressure-volume ring is provided below.

[0124] The present invention provides a classification system based on non-invasive pressure-volume rings, such as Figure 3 As shown, the system comprises:

[0125] The echocardiogram acquisition module 1 is used to acquire a cardiogram of at least one cardiac cycle of the subject; the cardiogram is a standardized transthoracic echocardiogram.

[0126] The left ventricular volume curve extraction module 2 is used to extract the endocardial boundary of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, so as to obtain the left ventricular volume curve of the subject to be tested.

[0127] The left ventricular pressure-volume loop determining module 3 is used to determine the left ventricular pressure-volume loop of the subject to be tested based on the left ventricular volume curve of the subject to be tested and using a time-varying elastic model.

[0128] The cardiac fluid dynamics parameter acquisition module 4 is used to acquire the cardiac fluid dynamics parameters of the subject from the left ventricular pressure-volume loop of the subject.

[0129] The model building module 5 is used to build a classification prediction model using a classification algorithm.

[0130] The prediction and classification result determination module 6 is used to input the cardiac fluid dynamics parameters of the subject to be tested into the trained classification prediction model to determine the prediction and classification result.

[0131] Embodiment 3

[0132] An embodiment of the present invention provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the non-invasive pressure-volume ring-based classification method of embodiment 1.

[0133] Optionally, the above-mentioned electronic device may be a server.

[0134] In addition, an embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the classification method based on the non-invasive pressure-volume ring of the first embodiment.

[0135] The beneficial results of the classification method, system, electronic device and medium based on non-invasive pressure-volume ring provided by the present invention are:

[0136] The present invention provides a fully automatic diagnosis of heart failure. By inputting the echocardiogram, brachial artery blood pressure and blood pressure, a pressure-volume loop can be generated and a diagnosis of heart failure can be obtained without manual intervention, which reduces the workload of doctors and provides a diagnostic reference for doctors. In addition, the present invention has a fast calculation speed and high diagnostic efficiency, which helps doctors improve the quality of diagnosis.

[0137] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0138] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A classification method based on non-invasive pressure-volume ring, characterized in that: The method comprises: Acquiring a cardiogram of at least one cardiac cycle of the subject to be tested; the cardiogram is a standardized transthoracic echocardiogram; Extracting the endocardial border of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, to obtain a left ventricular volume curve of the subject to be tested; According to the left ventricular volume curve of the subject to be tested, using a time-varying elastic model, determining the left ventricular pressure-volume loop of the subject to be tested; Acquiring cardiac fluid dynamics parameters of the subject from the left ventricular pressure-volume loop of the subject; Use classification algorithms to build classification prediction models; The training process of the classification prediction model includes: obtaining a cardiogram of at least one cardiac cycle of the subject; the cardiogram is a standardized transthoracic echocardiogram; according to the set category, the cardiogram of at least one cardiac cycle of the subject is labeled in category to obtain standardized transthoracic echocardiograms of various categories; the endocardial boundary of the ventricle is extracted from the standardized transthoracic echocardiograms of various categories to obtain left ventricular volume curves of various categories; according to the left ventricular volume curves of various categories, a time-varying elastic model is used to obtain left ventricular pressure-volume rings of various categories; cardiac fluid dynamics parameters of various categories are obtained from the left ventricular pressure-volume rings of various categories; the classification prediction model is trained with the cardiac fluid dynamics parameters of various categories as input and the probability of the corresponding category as output to obtain a trained classification prediction model; The cardiac fluid dynamics parameters of the subject to be tested are input into a trained classification prediction model to determine a prediction classification result.

2. The classification method based on non-invasive pressure-volume ring according to claim 1, characterized in that: The method of training the classification prediction model by taking the cardiac fluid dynamics parameters of each category as input and the probability of the corresponding category as output to obtain the trained classification prediction model specifically includes: The cardiac fluid dynamics parameters of each category are divided into a training set and a validation set; Construct multiple initial classification prediction models based on multiple classification algorithms; Taking the training set as input, training the multiple initial classification prediction models to obtain multiple trained initial classification prediction models; Inputting the validation set into the multiple trained initial classification prediction models to obtain the category probabilities output by the multiple trained initial classification prediction models; Obtaining prediction classification results of the multiple trained initial classification prediction models according to the category probabilities output by the multiple trained initial classification prediction models; From the multiple trained initial classification prediction models, a trained initial classification prediction model that outputs the predicted classification result with the highest category probability is selected as the trained classification prediction model.

3. The classification method based on non-invasive pressure-volume ring according to claim 1, characterized in that: The time-varying elastic model is: Among them, E(t) is the change of myocardial stiffness at time t represented by time-varying elasticity, P(t) is the ventricular pressure at time t, V(t) is the ventricular volume at time t, and V0 is the ventricular volume at zero pressure.

4. The classification method based on non-invasive pressure-volume ring according to claim 1, characterized in that: The determining of the left ventricular pressure-volume loop of the subject according to the left ventricular volume curve of the subject using a time-varying elastic model specifically includes: converting the time-varying elastic model into a double Hill equation; The double Hill equation is solved according to the left ventricular volume curve of the subject to be tested to obtain the left ventricular pressure-volume loop of the subject to be tested.

5. The classification method based on non-invasive pressure-volume ring according to claim 1, characterized in that: The step of inputting the cardiac fluid dynamics parameters of the subject into a trained classification prediction model to determine the prediction classification result specifically includes: Inputting the cardiac fluid dynamics parameters of the subject to be tested into a trained classification prediction model to obtain the probability of the category output by the trained classification prediction model; The category with the highest probability is determined as the predicted classification result.

6. A classification system based on non-invasive pressure-volume loops, characterized in that: The system comprises: An echocardiogram acquisition module, used for acquiring a cardiogram of at least one cardiac cycle of the subject to be tested; the cardiogram is a standardized transthoracic echocardiogram; A left ventricular volume curve extraction module, used for extracting the endocardial boundary of the ventricle from the cardiogram of at least one cardiac cycle of the subject to be tested, so as to obtain the left ventricular volume curve of the subject to be tested; A left ventricular pressure-volume loop determination module, used to determine the left ventricular pressure-volume loop of the subject to be tested based on the left ventricular volume curve of the subject to be tested and using a time-varying elastic model; A cardiac fluid dynamics parameter acquisition module, used for acquiring cardiac fluid dynamics parameters of the subject to be tested from the left ventricular pressure-volume loop of the subject to be tested; A model building module is used to build a classification prediction model using a classification algorithm; the training process of the classification prediction model includes: obtaining a cardiogram of at least one cardiac cycle of the subject; the cardiogram is a standardized transthoracic echocardiogram; according to the set category, the cardiogram of at least one cardiac cycle of the subject is categorized to obtain standardized transthoracic echocardiograms of various categories; the endocardial boundary of the ventricle is extracted from the standardized transthoracic echocardiograms of various categories to obtain left ventricular volume curves of various categories; according to the left ventricular volume curves of various categories, a time-varying elastic model is used to obtain left ventricular pressure-volume loops of various categories; cardiac fluid dynamics parameters of various categories are obtained from the left ventricular pressure-volume loops of various categories; the classification prediction model is trained with the cardiac fluid dynamics parameters of various categories as input and the probability of the corresponding category as output to obtain a trained classification prediction model; The prediction and classification result determination module is used to input the cardiac fluid dynamics parameters of the subject to be tested into the trained classification prediction model to determine the prediction and classification result.

7. An electronic device, characterized in that: The device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the classification method based on a non-invasive pressure-volume ring according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The device stores a computer program, which, when executed by a processor, implements the classification method based on the non-invasive pressure-volume ring as claimed in any one of claims 1 to 5.

Images