A method and apparatus for determining the probability of coronary heart disease

Abstract

A method and device for determining a probability of suffering from coronary heart disease, the method comprising: obtaining an ECG signal of one or more leads of a user, determining a reconstructed signal corresponding to the ECG signal of the one or more leads respectively according to the ECG signal of the one or more leads, and determining the probability of suffering from coronary heart disease of the user according to the ECG signal of the one or more leads and the reconstructed signal respectively. The method can estimate the probability of suffering from coronary heart disease of the user.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of terminal, and in particular, to a method and device for determining the probability of suffering from coronary heart disease. BACKGROUND

[0002] Coronary heart disease refers to coronary atherosclerotic heart disease, which is caused by coronary artery stenosis and obstruction due to atherosclerosis, resulting in myocardial ischemia and hypoxia. According to the pathology of coronary heart disease, the most effective diagnostic evidence of coronary heart disease is to find physical changes in blood vessels. Therefore, the gold standard for detecting coronary heart disease is an invasive detection technology, such as coronary angiography. Among them, coronary angiography refers to a radiographic examination of the anatomical morphology of the coronary artery by injecting contrast medium into a specially designed coronary angiography catheter. However, coronary angiography is invasive, expensive, and complex to operate.

[0003] In addition, coronary heart disease can also be diagnosed by the following methods. For example, the biomarkers troponin and myoglobin in blood tests can also be used for the detection of acute coronary heart disease. For another example, the blood supply and oxygen supply capacity of the coronary artery to the heart can also be indirectly evaluated according to the supply and demand level of the patient's heart oxygen consumption, so as to indirectly detect coronary heart disease, such as indirectly detecting coronary heart disease through electrocardiogram signal analysis, load test, etc. However, these methods have certain limitations and can only assist in the diagnosis of coronary heart disease, and are less convenient.

[0004] Therefore, there is an urgent need for a method and device that can estimate the probability of suffering from coronary heart disease more conveniently, and thus can be used to remind users to seek medical treatment or remind doctors to prescribe coronary angiography for users for further examination. SUMMARY

[0005] Embodiments of the present application provide a method and device for determining the probability of suffering from coronary heart disease to estimate the probability of suffering from coronary heart disease.

[0006] In a first aspect, the embodiments of the present application provide a method for determining the probability of suffering from coronary heart disease, which comprises: obtaining an ECG signal of one or more leads of a user, determining a reconstructed signal corresponding to the ECG signal of each of the one or more leads according to the ECG signal of the one or more leads, and determining the probability of suffering from coronary heart disease of the user according to the ECG signal of the one or more leads and the reconstructed signal corresponding to each of the one or more leads. Wherein the ECG signal of the one or more leads includes an ECG signal of a first lead, and the reconstructed signal corresponding to the ECG signal of the first lead is obtained by signal reconstruction based on the features of the ECG signal of the first lead. The above method can estimate the probability of suffering from coronary heart disease.

[0007] In a possible design, determining the coronary heart disease suffering probability of the user according to the ECG signals of the one or more leads and the respective corresponding reconstructed signals can be performed according to the following method: determining an abnormal signal in the ECG signals of the one or more leads according to the ECG signals of the one or more leads and the respective corresponding reconstructed signals, and determining the coronary heart disease suffering probability of the user according to the abnormal signal and the corresponding reconstructed signal.

[0008] The above design can first determine an abnormal signal, and then determine the coronary heart disease suffering probability of the user according to the abnormal signal and the corresponding reconstructed signal.

[0009] In a possible design, when the reconstructed signal corresponding to the ECG signal of each lead of the one or more leads is determined according to the ECG signals of the one or more leads, a first model is used to extract a feature of the ECG signal of each lead from the ECG signals of the one or more leads, and a second model is used to reconstruct a signal according to the feature of the ECG signal of each lead to obtain the corresponding reconstructed signal; wherein the first model and the second model are obtained by training according to the ECG signals of the leads of a healthy population.

[0010] The above design can determine the reconstructed signal corresponding to the ECG signal of each lead in the ECG signals of the one or more leads by using the first model and the second model. In addition, the first model and the second model can also be combined into one model, which is used to obtain the reconstructed signal corresponding to the ECG signal of each lead from the ECG signals of the one or more leads.

[0011] In a possible design, in determining the abnormal signal in the ECG signal of the one or more leads according to the ECG signal of the one or more leads and the corresponding reconstructed signal respectively, if the ECG signal of the first lead and the corresponding reconstructed signal satisfy one or more of the following abnormal signal conditions, it is determined that the ECG signal of the first lead is an abnormal signal: in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, there is a difference whose absolute value is greater than a first threshold; or, in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, there is a ratio of the absolute value of the difference to the maximum value of the de-encoding, which is greater than a second threshold; wherein the maximum value of the de-encoding refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or, the sum of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a third threshold; or, the sum of the squares of the ratio of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-encoding is greater than a fourth threshold; wherein the maximum value of the de-encoding refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or, the standard deviation of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a fifth threshold; or, the standard deviation of the ratio of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-encoding is greater than a sixth threshold; wherein the maximum value of the de-encoding refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or, the X1th percentile of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a seventh threshold, X1 being a preset value; or, the X2th percentile of the ratio of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-encoding is greater than an eighth threshold, X2 being a preset value; wherein the maximum value of the de-encoding refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead.

[0012] The above design can be used to determine the abnormal signal in the ECG signal of the one or more leads.

[0013] In a possible design, when determining the probability of the user suffering from coronary heart disease according to the abnormal signal and the corresponding reconstructed signal, a target signal is determined according to the abnormal signal and the corresponding reconstructed signal, where the target signal reflects a difference between the abnormal signal and the corresponding reconstructed signal. The probability of the user suffering from coronary heart disease is determined according to the target signal by using a third model, where the third model is obtained by training according to a first signal set and a second signal set, the first signal set including target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient being determined according to the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, and the second signal set including target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population being determined according to the ECG signals of each lead of the healthy population and corresponding reconstructed signals. Exemplarily, the probability of the user suffering from coronary heart disease and the probability of the user being the healthy population are equal to 1.

[0014] In a possible design, when determining the probability of the user suffering from coronary heart disease according to the abnormal signal and the corresponding reconstructed signal, a target signal is determined according to the abnormal signal and the corresponding reconstructed signal, where the target signal reflects a difference between the abnormal signal and the corresponding reconstructed signal. The probability of the user suffering from coronary heart disease is determined according to the target signal by using a third model. The third model is obtained by training according to a first signal set, a second signal set, and a third signal set, the first signal set including target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient being determined according to the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, the second signal set including target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population being determined according to the ECG signals of each lead of the healthy population and corresponding reconstructed signals, and the third signal set including target signals corresponding to ECG signals of each lead of a population suffering from other diseases, the target signals corresponding to the ECG signals of each lead of the population suffering from other diseases being determined according to the ECG signals of each lead of the population suffering from other diseases and corresponding reconstructed signals. Exemplarily, the probability of the user suffering from coronary heart disease, the probability of the user suffering from other diseases, and the probability of the user being the healthy population are equal to 1.

[0015] The above design can be used to determine the probability of the user suffering from coronary heart disease according to the abnormal signal and the corresponding reconstructed signal.

[0016] In a possible design, the method further includes: determining a first feature according to the abnormal signal and the corresponding reconstructed signal; determining, according to the first feature, a position information of the first feature in a coronary heart disease lesion space by using a third model; and determining a coronary heart disease lesion position with a maximum probability of the user according to the position information of the first feature in the coronary heart disease lesion space.

[0017] The design can be used to determine the coronary heart disease lesion position with the maximum probability of the user.

[0018] In a possible design, the coronary heart disease lesion space is obtained by training according to the first signal set and diagnostic labels of ECG signals of each lead of a coronary heart disease patient.

[0019] In a possible design, the first feature is a fully connected layer feature.

[0020] In a second aspect, an embodiment of the present application provides a device for determining a coronary heart disease probability, the device comprising: a transceiver unit configured to obtain electrocardiogram (ECG) signals of one or more leads of a user; and a processing unit configured to determine reconstructed signals corresponding to the ECG signals of the one or more leads according to the ECG signals of the one or more leads, wherein the ECG signals of the one or more leads include ECG signals of a first lead, and the reconstructed signals corresponding to the ECG signals of the first lead are obtained by signal reconstruction based on features of the ECG signals of the first lead; and determine the coronary heart disease probability of the user according to the ECG signals of the one or more leads and the corresponding reconstructed signals.

[0021] In a possible design, when determining the coronary heart disease probability of the user according to the ECG signals of the one or more leads and the corresponding reconstructed signals, the processing unit is configured to determine an abnormal signal in the ECG signals of the one or more leads according to the ECG signals of the one or more leads and the corresponding reconstructed signals; and determine the coronary heart disease probability of the user according to the abnormal signal and the corresponding reconstructed signal.

[0022] In a possible design, when determining the reconstructed signals corresponding to the ECG signals of the one or more leads according to the ECG signals of the one or more leads, the processing unit is configured to extract features of the ECG signals of each lead in the ECG signals of the one or more leads by using a first model; and obtain the reconstructed signals corresponding to the ECG signals of each lead by signal reconstruction according to the features of the ECG signals of each lead by using a second model; wherein the first model and the second model are obtained by training according to ECG signals of each lead of a healthy population.

[0023] In a possible design, the processing unit is configured to, when determining the abnormal signal in the ECG signal of the one or more leads according to the ECG signal of the one or more leads and the corresponding reconstructed signal respectively, determine that the ECG signal of the first lead is an abnormal signal if the ECG signal of the first lead and the corresponding reconstructed signal satisfy one or more of the following abnormal signal conditions: there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the absolute value of the difference value is greater than a first threshold value; or there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the ratio of the absolute value of the difference value to a maximum value of the de-sign is greater than a second threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the sum of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a third threshold value; or the sum of the squares of the ratio of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than a fourth threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the standard deviation of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a fifth threshold value; or the standard deviation of the ratio of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than a sixth threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the X1th percentile of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a seventh threshold value, where X1 is a preset value; or the X2th percentile of the ratio of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than an eighth threshold value, where X2 is a preset value, and the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead.

[0024] In a possible design, the processing unit is configured to, when determining the probability of coronary heart disease of the user according to the abnormal signal and the corresponding reconstructed signal, determine a target signal according to the abnormal signal and the corresponding reconstructed signal, where the target signal reflects the difference between the abnormal signal and the corresponding reconstructed signal, and determine the probability of coronary heart disease of the user according to the target signal and a third model.

[0025] The third model is obtained by training according to the first signal set and the second signal set, the first signal set includes target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, and the second signal set includes target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and corresponding reconstructed signals.

[0026] In a possible design, the processing unit is configured to, when determining the coronary heart disease suffering probability of the user according to the abnormal signal and the corresponding reconstructed signal, determine a target signal according to the abnormal signal and the corresponding reconstructed signal, the target signal reflecting a difference between the abnormal signal and the corresponding reconstructed signal; and determine the coronary heart disease suffering probability of the user according to the third model and the target signal.

[0027] The third model is obtained by training according to the first signal set, the second signal set and the third signal set, the first signal set includes target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, the second signal set includes target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and corresponding reconstructed signals, and the third signal set includes target signals corresponding to ECG signals of each lead of a population with other diseases, the target signals corresponding to the ECG signals of each lead of the population with other diseases are determined by the ECG signals of each lead of the population with other diseases and corresponding reconstructed signals.

[0028] In a possible design, the processing unit is configured to, when determining the coronary heart disease suffering probability of the user according to the abnormal signal and the corresponding reconstructed signal, determine a target signal according to the abnormal signal and the corresponding reconstructed signal, the target signal reflecting a difference between the abnormal signal and the corresponding reconstructed signal; and determine the coronary heart disease suffering probability of the user according to the third model and the target signal.

[0029] In a possible design, the coronary heart disease lesion space is obtained by training according to the first signal set and diagnosis labels of the ECG signals of each lead of the coronary heart disease patient.

[0030] In a possible design, the first feature is a fully connected layer feature.

[0031] In a third aspect, an embodiment of the present application provides a communication device, including a processor and an interface circuit, the interface circuit being configured to receive a signal from another communication device outside the communication device and transmit the signal to the processor or send a signal from the processor to the other communication device outside the communication device, and the processor being configured to implement the first aspect and any possible design of the first aspect by means of a logic circuit or executing code instructions.

[0032] In a fourth aspect, an embodiment of the present application provides a communication device, including a processor and a memory, the memory storing computer-executable code instructions, and the processor being configured to implement the first aspect and any possible design of the first aspect by invoking the computer-executable code instructions.

[0033] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, the storage medium storing a computer program or instructions, and when the computer program or instructions are executed by a communication device, the first aspect and any possible design of the first aspect are implemented.

[0034] In a sixth aspect, an embodiment of the present application provides an ECG wearable device, including a processing module and an acquisition module, the acquisition module being configured to acquire an ECG signal of one or more leads of a user, and the processing module being configured to implement the first aspect and any possible design of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 An illustration of various possible features of an ST segment in an embodiment of the present application;

[0036] Figure 2 An illustration of various possible features of a T wave in an embodiment of the present application;

[0037] Figure 3 An illustration of feature extraction and reconstruction of ECG signals of each lead of a healthy population in an embodiment of the present application;

[0038] Figure 4 An illustration of ECG signals of each lead of a healthy population and reconstructed signals corresponding to the ECG signals of each lead of the healthy population in an embodiment of the present application;

[0039] Figure 5 An illustration of ECG signals of each lead of a coronary heart disease patient and reconstructed signals corresponding to the ECG signals of each lead of the coronary heart disease patient in an embodiment of the present application;

[0040] Figure 6 An illustration of a coronary heart disease lesion space in an embodiment of the present application;

[0041] Figure 7 An overview flowchart for determining the probability of coronary heart disease in an embodiment of the present application;

[0042] Figure 8 A diagram showing the full connection layer features corresponding to the coronary heart disease lesion space in an embodiment of the present application;

[0043] Figure 9A A diagram showing the posture and method of collecting ECG signals by an ECG wearable watch in an embodiment of the present application;

[0044] Figure 9B One of the diagrams of 6-lead ECG signals in an embodiment of the present application;

[0045] Figure 9C One of the diagrams of reconstructed signals corresponding to the 6-lead ECG signals in an embodiment of the present application;

[0046] Figure 9D One of the diagrams showing abnormal signal leads in an embodiment of the present application;

[0047] Figure 9E A diagram showing the coronary heart disease lesion space of user A in an embodiment of the present application;

[0048] Figure 9F A diagram showing the probability of coronary heart disease of user A in an embodiment of the present application;

[0049] Figure 10A A diagram showing the contact position of collecting ECG signals by a heart rate chest strap in an embodiment of the present application;

[0050] Figure 10B The second diagram of 6-lead ECG signals in an embodiment of the present application;

[0051] Figure 10C The second diagram of reconstructed signals corresponding to the 6-lead ECG signals in an embodiment of the present application;

[0052] Figure 10D The second diagram showing abnormal signal leads in an embodiment of the present application;

[0053] Figure 10E A diagram showing the coronary heart disease lesion space of user B in an embodiment of the present application;

[0054] Figure 10F A diagram showing the probability of coronary heart disease of user B in an embodiment of the present application;

[0055] Figure 11A A diagram of 12-lead ECG signals in an embodiment of the present application;

[0056] Figure 11BFig. 1 is a schematic diagram of a reconstructed signal corresponding to 12-lead ECG signals in an embodiment of the present application;

[0057] Figure 11C Fig. 2 is a schematic diagram of displaying abnormal signal leads in 12-lead ECG signals in an embodiment of the present application;

[0058] Figure 11D Fig. 3 is a schematic diagram of displaying coronary heart disease lesion space of user C in an embodiment of the present application;

[0059] Figure 11E Fig. 4 is a schematic diagram of displaying coronary heart disease risk of user C in an embodiment of the present application;

[0060] Figure 12 Fig. 5 is a schematic diagram of a communication device in an embodiment of the present application;

[0061] Figure 13 Fig. 6 is another schematic diagram of a communication device in an embodiment of the present application. DETAILED DESCRIPTION

[0062] The general coronary heart disease includes coronary atherosclerosis, coronary inflammation, spasm, embolism, trauma and congenital malformation, but the non-coronary atherosclerosis causes are very rare. The coronary heart disease can be classified into five types according to symptoms, i.e. concealed type, angina pectoris type, myocardial infarction type, heart failure type and sudden death type.

[0063] The cardiac depolarization and repolarization process generates an electrocardiogram vector, which is transmitted to the whole body through volume conduction and generates a potential difference. The two electrodes are placed at any two points of the human body and connected to an electrocardiograph, and the electrocardiogram can be recorded. The line connecting the electrodes and the electrocardiograph is called an electrocardiogram lead. The commonly used lead is the standard lead, also known as the bipolar limb lead, which reflects the potential difference between two limbs. The international standard lead system has twelve leads, namely I, II, III, avL, avF, avR, v1, v2, v3, v4, v5 and v6.

[0064] For example, lead I refers to connecting the left upper limb electrode to the positive terminal of the electrocardiograph and the right upper limb electrode to the negative terminal, reflecting the potential difference between the left upper limb (I) and the right upper limb (r). When the potential of I is higher than that of r, an upward waveform is recorded; when the potential of r is higher than that of I, a downward waveform is recorded. Lead II refers to connecting the left lower limb electrode to the positive terminal of the electrocardiograph and the right upper limb electrode to the negative terminal, reflecting the potential difference between the left lower limb (F) and the right upper limb (r). When the potential of F is higher than that of r, an upward wave is recorded; otherwise, a downward wave is recorded. Lead III refers to connecting the left lower limb to the positive terminal of the electrocardiograph and the left upper limb electrode to the negative terminal, reflecting the potential difference between the left lower limb (F) and the left upper limb (I). When the potential of F is higher than that of I, an upward wave is recorded; otherwise, a downward wave is recorded.

[0065] Generally speaking, there are mainly four cases of typical electrocardiogram manifestations of coronary heart disease: (1) T wave is bidirectional, low or inverted; (2) ST segment is lowered or elevated; (3) R wave is decreased or disappeared; (4) the change of electrocardiogram is dynamic. If there are typical clinical symptoms of coronary heart disease and characteristic changes in electrocardiogram, most of them can be directly diagnosed as coronary heart disease by electrocardiogram, except for a small part of the population such as menopausal women. In addition, if there are typical clinical symptoms of coronary heart disease, but there is no obvious change in ordinary electrocardiogram, it may be because the arterial circulation of the heart has a large blood supply capacity, and it is not easy to check the abnormality at rest, and the real change of electrocardiogram can be found only by exercise treadmill electrocardiogram test, and further coronary angiography examination is necessary. Therefore, electrocardiogram is an essential and preferred examination method for the diagnosis of coronary heart disease, which provides important clinical reference value for the diagnosis of coronary heart disease. Among them, the main advantages of electrocardiogram in diagnosing myocardial infarction include: (1) Electrocardiogram changes appear early, which is beneficial to early diagnosis and intervention of myocardial infarction; (2) not only qualitative, but also positioning; (3) not only diagnosis, but also staging; (4) prognostic value; (5) simple equipment, popularization, non-invasive, convenient, repeatable, economical, etc.

[0066] Because the fundamental pathology of coronary heart disease is poor, insufficient, ischemic, etc. of coronary artery blood supply, ST-T changes will be shown on electrocardiogram, and the typical manifestations are ST segment horizontal downshift, T wave disappearance, inversion, etc. The following describes various abnormal characteristics that electrocardiogram may include.

[0067] I. ST segment downshift: normal ST segment is usually smooth and connected with T wave, often leading to difficulty in determining the end of ST segment and the beginning of T wave. Typical myocardial ischemia causes ST segment downshift. The earliest and most subtle change of ST segment is ST segment flattening, causing more obvious angle between ST segment and T wave.

[0068] II. ST segment elevation: Transient ST segment elevation in patients with chest pain is characteristic of myocardial ischemia and is usually seen in patients with coronary vasospastic angina. A subset of these patients will develop severe proximal coronary stenosis. Deep T wave inversion can follow the ST segment elevation and resolution, even in the absence of evidence of myocardial injury enzyme abnormalities.

[0069] Figure 1 Figures A through I in the above table show the schematic representation of various possible features of ST segment. In the above table, A is the schematic representation of normal ST segment, B is the schematic representation of ST segment with upsloping arching upward, C is the schematic representation of ST segment with downsloping arching upward, D is the schematic representation of ST segment with rapid upsloping downward, E is the schematic representation of ST segment with slow upsloping downward, F is the schematic representation of ST segment with horizontal downward, G is the schematic representation of ST segment with rapid downsloping downward, H is the schematic representation of ST segment with slow downsloping downward, and I is the schematic representation of ST segment with fishhook downward. Figure 1

[0070] III. T wave changes: Myocardial ischemia can cause a variety of T wave changes, such as tall, biphasic, inverted, or flat T waves.

[0071] In the above table, deep and symmetrical T wave inversion is strongly suggestive of myocardial ischemia. However, T wave inversion can also be a normal phenomenon, such as T wave inversion with QRS wave pointing downward in leads III, aVR, and VI.

[0072] In the above table, biphasic T wave changes are seen in some patients with non-transmural myocardial ischemia. Biphasic T wave, especially in precordial leads, is characteristic of acute myocardial ischemia. Biphasic T wave usually evolves into symmetrical T wave. These changes in patients with unstable angina or variant angina are strongly suggestive of myocardial ischemia.

[0073] Figure 2 Figures A through I in the above table show the schematic representation of various possible features of T wave. In the above table, A is the schematic representation of normal T wave, B is the schematic representation of inverted T wave, C is the schematic representation of tall upright T wave, D is the schematic representation of low flat T wave, E is the schematic representation of coronaric T wave, F is the schematic representation of tall peaked T wave, G is the schematic representation of double peaked T wave, H is the schematic representation of positive-negative biphasic T wave, and I is the schematic representation of negative-positive biphasic T wave.

[0074] ​As can be seen from the above, the ECG changes of myocardial ischemia patients are mainly ST-T changes. However, ST-T changes are not necessarily caused by myocardial ischemia. ST-T changes are related to various factors, and ST-T changes cannot be equated with myocardial ischemia or myocardial infarction. For example, the influencing factors of ST-T changes can include physiological factors, pharmaceutical factors, non-cardiac diseases, cardiac diseases, etc. Among them, the physiological factors can include body position, body temperature, excessive ventilation, anxiety, tachycardia, neurogenic effects, physical exercise, age, etc. The pharmaceutical factors can include digitalis, antiarrhythmic drugs and antipsychotic drugs. The non-cardiac diseases can include electrolyte disorders, cerebrovascular accidents, shock, anemia, allergic reactions, infections, endocrine disorders, acute abdominal pain, pulmonary embolism, etc. The cardiac diseases can include ischemic heart disease, primary cardiomyopathy, secondary myocardial changes, pericardial disease, cardiac electrical abnormalities, etc.

[0075] The ST-T changes can also include primary ST-T changes and secondary ST-T changes. Among them, the primary ST-T changes refer to the ST-T changes when the ventricular depolarization is unchanged, which refers to the changes of ECG ST-T caused by abnormal myocardial conditions and abnormal ventricular repolarization. It is commonly seen in clinical practice that chronic coronary artery insufficiency, angina pectoris, myocardial infarction, ventricular hypertrophy, myocarditis, pericarditis, cardiomyopathy, drug effects and electrolyte disorders, etc. The secondary ST-T changes refer to the ST-T changes when the ventricular depolarization changes, which refers to the changes of ECG ST-T caused by abnormal ventricular repolarization due to abnormal ventricular depolarization. It is commonly seen in clinical practice that ventricular hypertrophy, bundle branch block, ventricular pre-excitation, ventricular excitation, paced cardiac rhythm, etc.

[0076] It should be noted that the embodiments of the present application can be applied to an electrocardiogram (ECG) signal acquisition device. Exemplarily, the ECG signal acquisition device can be a wearable device, such as a smart watch, a smart chest strap, a single-lead ECG watch, a multi-lead ECG watch, a chest ECG patch, a single-lead or multi-lead ECG patch, etc. Among them, the ECG signal acquisition device can perform single-lead or multi-lead ECG signal acquisition, and process the acquired ECG signal to obtain the coronary heart disease (CHD) risk of a user, and also obtain the most likely CHD lesion location of the user. The ECG signal acquisition device with a display screen can also display the CHD risk of the user and the most likely CHD lesion location of the user and other related information through the display screen. In addition, in some embodiments, at least one ECG signal acquisition device (hereinafter referred to as a first device) among the at least two ECG signal acquisition devices can acquire ECG signals and obtain ECG signals acquired by other ECG signal acquisition devices. The first device combines the ECG signals acquired by itself and the ECG signals acquired by other ECG signal acquisition devices to process together to obtain the CHD risk of the user and the most likely CHD lesion location of the user and other related information.

[0077] In some embodiments, the ECG signal acquisition device can also only acquire ECG signals and send the acquired ECG signals to an ECG signal processing device for processing. The ECG signal processing device can process the obtained ECG signals to obtain the CHD risk of a user, and also obtain the most likely CHD lesion location of the user. The ECG signal processing device with a display screen can also display the CHD risk of the user and the most likely CHD lesion location of the user and other related information through the display screen. Exemplarily, the ECG signal processing device here can refer to a mobile phone, a tablet, a computer, a television, etc. It can be understood that if the ECG signal processing device does not have a display screen, the CHD risk of the user and the most likely CHD lesion location of the user and other related information can also be sent to a display device for display.

[0078] In an example, the ECG acquisition device is an ECG wearable watch, and a user A wears the ECG wearable watch to acquire ECG signals of the user A by using the ECG wearable watch as shown in FIG. 1. The ECG wearable watch can acquire ECG signals of the user A through the ECG wearable watch, and send the acquired ECG signals to a server for processing. The server can process the ECG signals to obtain the CHD risk of the user A and the most likely CHD lesion location of the user A. Figure 9AThe posture (electrode contact foot wrist preset position or leg preset position or chest preset position in the watch) shown can enable the ECG wearable watch to collect a 6-lead ECG signal of a limb. The ECG wearable watch can process the collected 6-lead ECG signal of the limb by using the method provided in the embodiments of the present application to obtain the coronary heart disease suffering probability of the user. Alternatively, the ECG wearable watch can transmit the 6-lead ECG signal of the limb to an ECG signal processing device (for example, a mobile phone or an iPad or a computer, etc.). The ECG signal processing device processes the collected 6-lead ECG signal of the limb by using the method provided in the embodiments of the present application to obtain the coronary heart disease suffering probability of the user A.

[0079] In an example, the ECG collection device is a heart rate chest strap, and the user wears the heart rate chest strap. The heart rate chest strap contacts the user B as shown in the preset position (V1-V6) to collect a 6-lead ECG signal of a chest. The heart rate chest strap can transmit the collected 6-lead ECG signal of the chest to an ECG signal processing device (for example, a mobile phone or an iPad or a computer, etc.). The ECG signal processing device processes the collected 6-lead ECG signal of the chest by using the method provided in the embodiments of the present application to obtain the coronary heart disease suffering probability of the user B. Figure 10A

[0080] In yet another example, the ECG collection device is an ECG wearable watch and a heart rate chest strap, and the user C wears the ECG wearable watch and the heart rate chest strap. Based on the above two examples, the ECG wearable watch can collect a 6-lead ECG signal of a limb, and the heart rate chest strap can collect a 6-lead ECG signal of a chest. The heart rate chest strap can transmit the collected 6-lead ECG signal of the chest to the ECG wearable watch. The ECG wearable watch can process the 12-lead ECG signal by using the method provided in the embodiments of the present application to obtain the coronary heart disease suffering probability of the user C. Alternatively, the heart rate chest strap transmits the collected 6-lead ECG signal of the chest to an ECG signal processing device, and the ECG wearable watch transmits the 6-lead ECG signal of the limb to the ECG signal processing device. The ECG signal processing device processes the 12-lead ECG signal by using the method provided in the embodiments of the present application to obtain the coronary heart disease suffering probability of the user C.

[0081] The embodiments of the present application provide a method and a device for determining a coronary heart disease suffering probability.

[0082] First, a first model M1, a second model M2 and a third model M3 need to be established.

[0083] ​The first model M1 is used to extract full-segment features of the original ECG signal. The second model M2 is used to reconstruct a reconstructed signal corresponding to the original ECG signal according to the full-segment features extracted by the first model M1. The full-segment features refer to features of the original ECG signal in at least one complete cycle. For example, the features obtained from the first P-wave starting position to the second P-wave starting position in the adjacent two heartbeats in the ECG graph are the full-segment features of one cycle. Or the features obtained from the starting position to the ending position in the ECG graph obtained by one detection are the full-segment features of one detection.

[0084] For example, the first model M1 and the second model M2 can be established by, but are not limited to, the following methods. The ECG signals of each lead of a healthy population are selected as original training signals, the full-segment features of each original training signal are obtained, and the first model M1 is trained according to the full-segment features of each original training signal. For example, the method for training the first model M1 can include, but is not limited to, a neural network, a random forest, and the like. The full-segment features herein include, but are not limited to, one-dimensional features, two-dimensional features, or multi-dimensional features. Further, the second model M2 is trained according to the full-segment features of each original training signal and the reconstructed signal. For example, the method for training the second model M2 includes, but is not limited to, a neural network, a random forest, and the like. The healthy population can refer to a population without coronary heart disease and other cardiovascular diseases.

[0085] In some embodiments, the ECG signals of each lead of a healthy population are selected as original training signals, and each original training signal is subjected to data preprocessing. For example, the original training signal includes an ECG signal of a first lead. The ECG signal of the first lead can be represented by an N-dimensional signal [x1, x2, x3, … xN], where N is a positive integer. For example, xi in [x1, x2, x3, … xN] can refer to the signal amplitude at the i-th moment, or the value of the sampling point with the sampling order i, where i is a positive integer. For example, the sampling frequency is 100 Hz, and the sampling duration is 30 s, so N = 3000, that is, 100 samples per second, and 3000 sampling points are obtained in 30 s. The data preprocessing of the ECG signal of the first lead can include upsampling, downsampling, baseline removal, noise removal, filtering, and the like, where the specific operations of the data preprocessing are not limited in the embodiments of the present application.

[0086] Then, based on the pre-processed original training signal, a machine learning method (including but not limited to neural network, random forest, etc.) is used to train the first model M1. For example, based on the pre-processed first lead ECG signal, the full segment feature of the first lead ECG signal is obtained, which can be represented by an M-dimensional signal [y1, y2, y3, … yM], M < N, M is a positive integer. Wherein, the full segment feature herein can include but is not limited to one-dimensional feature, two-dimensional feature or multi-dimensional feature. Through the training process of extracting the full segment feature of a large number of original training signals, the parameters of the first model M1 are constantly improved.

[0087] Further, according to the full segment feature of each original training signal, signal reconstruction is performed, and a corresponding machine learning method (including but not limited to neural network, random forest, etc.) is used to train the second model M2. For example, according to the full segment feature [y1, y2, y3, … yM] of the first lead ECG signal, an N-dimensional signal [z1, z2, z3, … zN] is reconstructed. Through the training process of signal reconstruction on the full segment feature of a large number of original training signals, the parameters of the second model M2 are constantly improved.

[0088] Exemplarily, the first model M1 and the second model M2 can also not be distinguished, and can be combined into one model to obtain the reconstructed signal corresponding to each lead ECG signal from one or more lead ECG signals.

[0089] As shown in FIG. 1, Figure 3 Figure 3 The leftmost signal in FIG. 1 is the ECG signal of each lead of the healthy population, and each lead ECG signal is processed by the first model M1, which specifically includes a plurality of convolution pooling layers and fully connected layers and other processing to extract the full segment feature of the original signal. Further, according to the extracted full segment feature, the second model M2 is used for signal reconstruction, which specifically includes a plurality of deconvolution pooling layers and other processing to obtain the reconstructed signal, i.e. Figure 3 The rightmost signal in FIG. 1 is the reconstructed signal of each lead ECG signal.

[0090] Exemplarily, the N-dimensional signal [x1, x2, x3, … xN] can be processed by a plurality of convolution pooling layers and fully connected layers to obtain an M-dimensional signal [y1, y2, y3, … yM], and the M-dimensional signal [y1, y2, y3, … yM] is processed by a plurality of deconvolution pooling layers to reconstruct an N-dimensional signal [z1, z2, z3, … zN]. Wherein, the same sequence number in the N-dimensional signal [x1, x2, x3, … xN] and the N-dimensional signal [z1, z2, z3, … zN] is the signal amplitude at the same time, for example, x1 and z1 correspond to the amplitude of the original training signal at time 1 and the amplitude of the reconstructed signal corresponding to the original training signal at time 1, respectively. ​

[0091] It should be noted that the first model M1 and the second model M2 will process the entire ECG signal, not just the signal of the ST-T segment in the ECG signal. For example, all ECG signals within 30s (an average of 30 cycles for a person) are taken as input signals, rather than a certain part of a certain ECG signal as input signals.

[0092] Figure 4 A schematic diagram of the reconstructed signals corresponding to the ECG signals of each lead of the healthy population in the embodiments of the present application and the ECG signals of each lead of the healthy population. Figure 5 A schematic diagram of the reconstructed signals corresponding to the ECG signals of each lead of the coronary heart disease patients in the embodiments of the present application and the ECG signals of each lead of the coronary heart disease patients.

[0093] From the above first model M1 and second model M2, Figure 4 It can be seen that the reconstructed signals corresponding to the ECG signals of each lead of the healthy population and the ECG signals of each lead of the healthy population are basically consistent based on the above first model M1 and second model M2. As shown in Figure 5 The reconstructed signals corresponding to the ECG signals of each lead of the coronary heart disease patients and the ECG signals of each lead of the coronary heart disease patients differ greatly based on the above first model M1 and second model M2.

[0094] It can be understood that the ECG signals of any two healthy users have a high similarity, while the ECG signals of any two coronary heart disease patients generally have a low similarity, that is, the ECG signals of coronary heart disease patients are diverse and it is not easy to obtain all possible ECG signals of coronary heart disease patients. Specifically, the lesion positions and lesion degrees of different coronary heart disease patients are not the same, so the ECG signals of different coronary heart disease patients differ greatly, and some coronary heart disease patients may also have other diseases, which may also have some impact on the ECG signals. For example, a period of ECG signal of a certain lead is divided into multiple segments, for healthy users, each segment of ECG signal is within the corresponding interval range, while for coronary heart disease patients, there are several segments of ECG signals that are not within the corresponding interval range.

[0095] Exemplarily, the full segment features of the ECG signals of the healthy users are extracted according to the first model M1, because the ECG signals of different healthy users have a high similarity, so even the full segment features extracted by different healthy users also satisfy a certain variation rule or preset condition, further, the reconstructed signals similar to the ECG signals of the healthy users can be obtained based on the extracted features according to the second model M2. This is also the training purpose of the first model M1 and the second model M2, that is, the ECG signals of the healthy users can be basically restored by the first model M1 and the second model M2.

[0096] However, according to the first model M1, the full segment features of the ECG signals of the coronary heart disease patients are extracted, and the extracted full segment features do not satisfy the change rule or the preset condition. At this time, according to the M2, the signal is reconstructed based on the extracted features, and the reconstructed signal is quite different from the original ECG signal of the coronary heart disease patient, that is, it cannot be restored to the original ECG signal of the coronary heart disease patient.

[0097] The third model M3 can be established by the following methods, but is not limited to the following methods. The method for training the third model M3 can include, but is not limited to, neural network, random forest, etc.

[0098] Example 1: First, the ECG signals of each lead of the coronary heart disease population, the ECG signals of each lead of the other disease population, and the ECG signals of each lead of the healthy population are selected as the original training signals. According to each original training signal and the first model M1 and the second model M2, the reconstructed signal corresponding to each original training signal is obtained. The third model M3 is trained according to the original training signal and the reconstructed signal corresponding to the original training signal. Exemplarily, the other disease population can refer to the population without coronary heart disease but with other cardiovascular diseases.

[0099] In some embodiments, first, the ECG signals of each lead of the coronary heart disease population, the ECG signals of each lead of the other disease population, and the ECG signals of each lead of the healthy population are selected as the original training signals. According to each original training signal and the first model M1 and the second model M2, the reconstructed signal corresponding to each original training signal is obtained. Then, according to each original training signal and the corresponding reconstructed signal, the target signal corresponding to each original training signal is determined, wherein the target signal corresponding to each original training signal reflects the difference between the original training signal and the reconstructed signal corresponding to the original training signal.

[0100] Further, based on the target signal corresponding to each original training signal, a machine learning method (including but not limited to neural network, random forest, etc.) is used to train a third model M3 according to the label of each original training signal. The type label of each lead ECG signal of the coronary heart disease population, the type label of each lead ECG signal of the other disease population, and the type label of each lead ECG signal of the healthy population are different from each other. For example, the type label of each lead ECG signal of the coronary heart disease population is coronary heart disease, the type label of each lead ECG signal of the other disease population is other disease, and the type label of each lead ECG signal of the healthy population is healthy. Alternatively, the type label of each lead ECG signal of the coronary heart disease population is label 1, the type label of each lead ECG signal of the other disease population is label 2, and the type label of each lead ECG signal of the healthy population is label 3. It can be understood that the specific form of the above type label is only an example and is not limited by the embodiments of the present application.

[0101] For example, the original training signal includes the ECG signal of the first lead. The ECG signal of the first lead can be represented by an N-dimensional signal [x1, x2, x3, … xN], and the reconstructed signal corresponding to the ECG signal of the first lead can be represented by an N-dimensional signal [z1, z2, z3, … zN], where N is a positive integer. The difference between the ECG signal of the first lead [x1, x2, x3, … xN] and the corresponding reconstructed signal [z1, z2, z3, … zN] is an N-dimensional signal [a1, a2, a3, … aN], which is denoted as the target signal corresponding to the ECG signal of the first lead. Similarly, the corresponding target signal determined by the difference between the other original training signal and the corresponding reconstructed signal can be obtained. Further, based on the target signal corresponding to each original training signal obtained above, a machine learning method (including but not limited to neural network, random forest, etc.) is used to train a third model M3 according to the label of each original training signal (e.g., coronary heart disease, other disease, and healthy). Through the training process based on the type label of the original training signal for a large number of original training signals corresponding to the target signal, the parameters of the third model M3 are constantly improved.

[0102] Based on the third model M3 obtained in Example 1, first, the target signal corresponding to the ECG signal of the user is determined according to the ECG signal of the user and the reconstructed signal corresponding to the ECG signal. Inputting the target signal corresponding to the ECG signal into the third model M3 can obtain the probability that the ECG signal is the ECG signal of the coronary heart disease population, the probability that the ECG signal is the ECG signal of the other disease population, and the probability that the ECG signal is the ECG signal of the healthy population, which can be used as the coronary heart disease probability of the user, the other disease probability of the user, and the probability that the user is a healthy population. The sum of the coronary heart disease probability of the user, the other disease probability of the user, and the probability that the user is a healthy population is 1.

[0103] Exemplarily, the target signal is determined according to the ECG signal of the I lead of the user and the corresponding reconstructed signal, it is assumed that 1000 features can be extracted from the target signal by using the third model, and it is determined that 700 features of the target signal correspond to the ECG signal of the I lead of the coronary heart disease patient, 200 features of the target signal correspond to the ECG signal of the I lead of the healthy user, and 100 features of the target signal correspond to the ECG signal of the I lead of the other disease. Therefore, the output result of the third model is that the coronary heart disease probability of the user is 70%, the probability of the user belonging to the healthy population is 20%, and the probability of the user being other diseases is 10%.

[0104] It should be noted that even if the target signal is determined according to the ECG signal of the coronary heart disease patient in a certain original training signal and the corresponding reconstructed signal, and then the target signal is input into the third model, the coronary heart disease probability of the user cannot be obtained.

[0105] Example 2: First, the ECG signals of each lead of the coronary heart disease population and the ECG signals of each lead of the healthy population are selected as the original training signals, the reconstructed signal corresponding to each original training signal is obtained according to each original training signal and the first model M1 and the second model M2, and the third model M3 obtained by training according to the original signal and the reconstructed signal corresponding to the original signal is obtained.

[0106] In some embodiments, first, the ECG signals of each lead of the coronary heart disease population and the ECG signals of each lead of the healthy population are selected as the original training signals, and the reconstructed signal corresponding to each original training signal is obtained according to each original training signal and the first model M1 and the second model M2. Then, the target signal corresponding to each original training signal is determined according to each original training signal and the corresponding reconstructed signal, wherein the target signal corresponding to each original training signal reflects the difference between the original training signal and the reconstructed signal corresponding to the original training signal.

[0107] Further, based on the target signal corresponding to each original training signal, a machine learning method (including but not limited to neural network, random forest, etc.) is used to train a third model M3 according to the label of each original training signal. The type label of each lead ECG signal of the coronary heart disease population and the type label of each lead ECG signal of the healthy population are different from each other. For example, the type label of each lead ECG signal of the coronary heart disease population is coronary heart disease, and the type label of each lead ECG signal of the healthy population is healthy. Alternatively, the type label of each lead ECG signal of the coronary heart disease population is label 1, and the type label of each lead ECG signal of the healthy population is label 2. It can be understood that the specific form of the above type label is only an example and is not limited by the embodiments of the present application.

[0108] For example, the original training signal includes the ECG signal of the first lead. The ECG signal of the first lead can be represented by an N-dimensional signal [x1, x2, x3, … xN], and the reconstructed signal corresponding to the ECG signal of the first lead can be represented by an N-dimensional signal [z1, z2, z3, … zN], N is a positive integer. The difference between the ECG signal [x1, x2, x3, … xN] of the first lead and the corresponding reconstructed signal [z1, z2, z3, … zN] is an N-dimensional signal [a1, a2, a3, … aN], which is denoted as the target signal corresponding to the ECG signal of the first lead. Similarly, the corresponding target signal determined by the difference between the other original training signal and the corresponding reconstructed signal can be obtained. Further, based on the target signal corresponding to each original training signal obtained above, a machine learning method (including but not limited to neural network, random forest, etc.) is used to train a third model M3 according to the type label (such as coronary heart disease, other diseases, and healthy) of each original training signal. Through the training process based on the type label of the original training signal for a large number of original training signal corresponding target signals, the parameters of the third model M3 are constantly improved.

[0109] Based on the third model M3 obtained in Example 2, first, the target signal corresponding to the ECG signal of the user is determined according to the ECG signal of the user and the reconstructed signal corresponding to the ECG signal. Inputting the target signal corresponding to the ECG signal into the third model M3 can obtain the probability that the ECG signal is the ECG signal of the coronary heart disease population and the probability that the ECG signal is the ECG signal of the healthy population, which can be used as the coronary heart disease probability of the user and the probability that the user is a healthy population. The sum of the coronary heart disease probability of the user and the probability that the user is a healthy population is 1.

[0110] Wherein, on the basis of the third model, if the corresponding coronary heart disease of each coronary heart disease patient has been determined through diagnosis, for example, posterior wall, lateral wall, inferior wall, anterior wall, or left coronary artery trunk, anterior descending branch, circumflex branch, right coronary artery, etc. coronary heart disease, the diagnostic labels of the ECG signals of each lead of the coronary heart disease patient (for example, posterior wall, lateral wall, inferior wall, anterior wall, or left coronary artery trunk, anterior descending branch, circumflex branch, right coronary artery) can be combined to train the obtained coronary heart disease lesion space. Specifically, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are processed by the third model, the first features (such as full connection layer features) are extracted, and the first features are processed by dimension reduction, including but not limited to 3 dimensions, combined with the diagnostic labels of the ECG signals of each lead of the coronary heart disease patient, the coronary heart disease lesion space can be trained. As shown in Figure 6 The schematic diagram of the 3-dimensional coronary heart disease lesion space is shown. Wherein, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and the reconstructed signals corresponding to the ECG signals of each lead of the coronary heart disease patient. Wherein, Figure 6 The X-axis, Y-axis and Z-axis in the above can be understood as the following meanings but are not limited to the following meanings: the X-axis, Y-axis and Z-axis can be used to identify the simulated three-dimensional space of the heart, or the X-axis, Y-axis and Z-axis can be used to identify the limb lead lesion degree, the chest lead lesion degree, the limb and chest lead management lesion degree, the X-axis, Y-axis and Z-axis can be used to identify the ST lesion, T lesion and P wave lesion. Therefore, the third model M3 can determine the first features (such as full connection layer features) according to the ECG signals of the user and the reconstructed signals corresponding to the ECG signals, and process the first features by dimension reduction (such as reducing to 3 dimensions), and determine the maximum feasible lesion position of the user according to the position information of the coronary heart disease lesion space as shown in Figure 6

[0111] As shown in Figure 7 The following takes the first device as an example to illustrate the specific process of determining the coronary heart disease probability of the user. Wherein, the first device can be an ECG signal acquisition device or an ECG signal processing device, or a chip in the ECG signal acquisition device or the ECG signal processing device.

[0112] Step 1: The first device acquires the ECG signals of one or more leads of the user.

[0113] When the first device is an ECG signal acquisition device, the first device can acquire the ECG signals according to the collected ECG signals, or the first device can also collect the ECG signals collected by other ECG signal acquisition devices at the same time.

[0114] When the first device is an ECG signal processing device, the first device collects the ECG signals collected by other ECG signal acquisition devices.​

[0115] Step 2: The first device determines the reconstructed signal corresponding to each of the one or more lead ECG signals according to the one or more lead ECG signals. The one or more lead ECG signals include the first lead ECG signal, and the reconstructed signal corresponding to the first lead ECG signal is obtained by signal reconstruction based on the feature of the first lead ECG signal.

[0116] The first device extracts the feature of each of the one or more lead ECG signals, for example, the full segment feature, and reconstructs the feature of each of the one or more lead ECG signals to obtain the reconstructed signal corresponding to each of the one or more lead ECG signals.

[0117] For example, the full segment feature of each of the one or more lead ECG signals is extracted based on the first model M1, and the reconstructed signal corresponding to each of the one or more lead ECG signals is obtained by reconstructing the extracted full segment feature of each of the one or more lead ECG signals based on the second model M2.

[0118] In some embodiments, after determining the reconstructed signal corresponding to each of the one or more lead ECG signals, the first device can further determine the abnormal signal in the one or more lead ECG signals according to the one or more lead ECG signals and the reconstructed signal corresponding to each of the one or more lead ECG signals.

[0119] For example, for any lead ECG signal and the reconstructed signal corresponding to the lead ECG signal, if one or more of the following abnormal signal conditions is met, the lead ECG signal can be determined as an abnormal signal:

[0120] Condition 1: In the difference between the two signals, there is a difference whose absolute value is greater than a threshold T1.

[0121] For example, the maximum value of the absolute value of the difference between the I lead ECG signal and the reconstructed signal of the I lead ECG signal is 0.3 mV, and the threshold T1 corresponding to the I lead is 0.2 mV. Therefore, the I lead ECG signal is an abnormal signal.

[0122] Condition 2: In the difference between the two signals, there is a difference whose normalized value (absolute value of the difference / denormalization maximum value) is greater than a threshold T2, wherein the denormalization maximum value refers to the maximum value of the absolute value of the amplitude of the original signal in the two signals.

[0123] Condition 3: The square of the difference between the two signals is greater than a threshold T3.

[0124] Condition 4: The square sum of the normalized value (difference / denormalization maximum value) of the difference between the two signals is greater than a threshold T4.

[0125] Condition 5: the standard deviation of the difference of the two segments of the signal is greater than a threshold T5;

[0126] Condition 6: the standard deviation of the normalized value (difference / demodulation maximum value) of the difference of the two segments of the signal is greater than a threshold T6;

[0127] Condition 7: the X1 percentile of the absolute value of the difference of the two segments of the signal is greater than a threshold T7;

[0128] Condition 8: the X2 percentile of the normalized value (absolute value of the difference / demodulation maximum value) of the absolute value of the difference of the two segments of the signal is greater than a threshold T8.

[0129] It can be understood that the above conditions are only examples, and other methods can be used to determine abnormal signals.

[0130] The thresholds T1 to T8 and X1 and X2 can be determined according to the statistical results of the differences between the original lead ECG signals of the coronary heart disease patients and the reconstructed signals corresponding to the original lead ECG signals, such as the 99th percentile.

[0131] For example, if the reconstructed signal corresponding to the ECG signal of the first lead and the ECG signal of the first lead satisfy one or more of the above conditions 1 to 8, it is determined that the ECG signal of the first lead is an abnormal signal, also known as a pathological lead signal.

[0132] For example, as shown in FIG. 1, I, aVR and aVL respectively correspond to ECG signals that satisfy one or more of the above conditions 1 to 8, which are abnormal signals. Figure 5

[0133] It can be understood that the first device can determine whether the obtained ECG signal of each lead is an abnormal signal. If it is determined that there is no abnormal signal in the obtained ECG signal of one or more leads, it can be determined that the user belongs to a healthy population, or the coronary heart disease probability of the user is 0, for example, by displaying the user belongs to a healthy population or the coronary heart disease probability of the user is 0 on the display screen. In addition, the first device can not determine whether the ECG signal of each lead is an abnormal signal, and directly execute step 3.

[0134] Step 3: The first device determines the coronary heart disease probability of the user according to the ECG signal of one or more leads and the reconstructed signal corresponding to the ECG signal of one or more leads.

[0135] For example, when the coronary heart disease probability is high, the first device can further suggest that the user go to the hospital for coronary heart disease treatment. When the coronary heart disease probability is low, the user can be prompted to perform ECG measurement again, or the user can be scheduled for the next ECG measurement to monitor the user's health status.​

[0136] It should be noted that the form of the coronary heart disease attack probability can adopt, but is not limited to, the following one or a combination of multiple ways:

[0137] 1. The coronary heart disease attack probability can be a specific numerical value.

[0138] 2. The coronary heart disease attack probability can be indicated by high, medium or low.

[0139] When the coronary heart disease attack probability can be indicated by high, medium or low, if the coronary heart disease attack probability is indicated as low, it can represent that the user has a lower risk of coronary heart disease, or the user belongs to a healthy population, depending on the classification rule. For example, when the coronary heart disease attack probability is 70% or more, the coronary heart disease attack probability is indicated as high, when the coronary heart disease attack probability is 50%-70%, the coronary heart disease attack probability is indicated as medium. When the coronary heart disease attack probability is 30%-50%, the coronary heart disease attack probability is indicated as low, and when the coronary heart disease attack probability is 30% or less, it is determined that the user belongs to a healthy population. Further, when the coronary heart disease attack probability is indicated as high or medium, the user can be further advised to go to the hospital for examination. When the coronary heart disease attack probability is indicated as low, the user can be further prompted to perform ECG measurement again, or the user can be scheduled for the next ECG measurement to monitor the user's health status. For another example, when the coronary heart disease attack probability is 65% or more, the coronary heart disease attack probability is indicated as high, when the coronary heart disease attack probability is 35%-65%, the coronary heart disease attack probability is indicated as medium. When the coronary heart disease attack probability is 35% or less, the coronary heart disease attack probability is indicated as low, and at this time, the coronary heart disease attack probability indicated as low represents that the user belongs to a healthy population.

[0140] It can be understood that the specific numerical value is only an example and is not limited to the embodiments of the present application.

[0141] In addition, the coronary heart disease attack probability can be indicated by more categories, which is not limited by the embodiments of the present application. For example, the coronary heart disease attack probability is indicated from high to low as A+, A, A-, B+, B, B-, and C.

[0142] 3. The coronary heart disease attack probability can also be indicated by a bar chart or a pie chart.

[0143] In some embodiments, the first device determines the coronary heart disease attack probability of the user by using a third model according to the ECG signals of one or more leads and the reconstructed signals corresponding to the ECG signals of one or more leads. At this time, the first device does not screen the signals input into the third model.

[0144] In some embodiments, the first device determines the probability of the user suffering from coronary heart disease according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal by using a third model. At this time, the first device first screens out the abnormal signal and the reconstructed signal corresponding to the abnormal signal, and then processes by using the third model.

[0145] It can be understood that the first device does not screen the signal input into the third model, which may result in a large amount of calculation of the third model, but saves the time and workload consumed in the early stage of screening the signal input into the third model. The first device first screens out the abnormal signal and the reconstructed signal corresponding to the abnormal signal, and then processes by using the third model, which can reduce the calculation amount of the third model. When no abnormal signal is found through screening, it can be directly concluded that the user is not a coronary heart disease user, without the need for signal processing by the third model.

[0146] In an example, the first device determines a target signal corresponding to the abnormal signal according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal, and the target signal corresponding to the abnormal signal reflects the difference between the abnormal signal and the reconstructed signal corresponding to the abnormal signal. For example, assuming that the ECG signal of the first lead is an abnormal signal, the ECG signal of the first lead can be represented by an N-dimensional signal [x1, x2, x3, … xN], and the reconstructed signal corresponding to the ECG signal of the first lead can be represented by an N-dimensional signal [z1, z2, z3, … zN], N being a positive integer. The difference between the ECG signal [x1, x2, x3, … xN] of the first lead and the corresponding reconstructed signal [z1, z2, z3, … zN] is an N-dimensional signal [a1, a2, a3, … aN], and the N-dimensional signal [a1, a2, a3, … aN] is recorded as the target signal corresponding to the ECG signal of the first lead.

[0147] The first device determines the probability of the user suffering from coronary heart disease, the probability of the user suffering from other diseases, and the probability of the user being a healthy population according to the target signal corresponding to the abnormal signal by using the third model. Exemplarily, the sum of the probability of the user suffering from coronary heart disease, the probability of the user suffering from other diseases, and the probability of the user being a healthy population is 1.

[0148] At this time, the third model is obtained by using the scheme of example 1 in the above third model related paragraph, and the repeated parts will not be described again.

[0149] For example, the first device determines, according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal, that the probability of the user suffering from coronary heart disease is 70%, the probability of the user suffering from other diseases is 15%, and the probability of the user being a healthy population is 15%. At this time, the first device can display only that the probability of the user suffering from coronary heart disease is 70%, or display that the probability of the user suffering from coronary heart disease is 70%, the probability of the user suffering from other diseases is 15%, and the probability of the user being a healthy population is 15%, or display that the probability of the user suffering from coronary heart disease is high, or display a column chart of the probability of the user suffering from coronary heart disease, or display a sector diagram corresponding to the probability of the user suffering from coronary heart disease, the probability of the user suffering from other diseases, and the probability of the user being a healthy population, and the like, which are not limited in the embodiments of the present application.

[0150] For example, the first device can only focus on the probability of the user suffering from coronary heart disease and does not prompt that the probability of the user suffering from other diseases is high. Alternatively, if the probability of the user suffering from other diseases is high, the first device can prompt the user to go to the hospital for further examination.

[0151] It can be understood that the user corresponding to the abnormal signal is not necessarily a coronary heart disease patient, but can be a patient of other diseases or a healthy population.

[0152] In an example, the first device determines, according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal, a target signal corresponding to the abnormal signal, and the target signal corresponding to the abnormal signal reflects the difference between the abnormal signal and the reconstructed signal corresponding to the abnormal signal. For example, assuming that the ECG signal of the first lead is an abnormal signal, the ECG signal of the first lead can be represented by an N-dimensional signal [x1, x2, x3, … xN], and the reconstructed signal corresponding to the ECG signal of the first lead can be represented by an N-dimensional signal [z1, z2, z3, … zN], where N is a positive integer. The difference between the ECG signal [x1, x2, x3, … xN] of the first lead and the corresponding reconstructed signal [z1, z2, z3, … zN] is an N-dimensional signal [a1, a2, a3, … aN], and the N-dimensional signal [a1, a2, a3, … aN] is referred to as the target signal corresponding to the ECG signal of the first lead.

[0153] The first device determines, according to the target signal corresponding to the abnormal signal, the probability of the user suffering from coronary heart disease and the probability of the user being a healthy population by using the third model. For example, the sum of the probability of the user suffering from coronary heart disease and the probability of the user being a healthy population is 1.

[0154] At this time, the third model is obtained by using the scheme of Example 2 in the above-mentioned third model related paragraphs, and the repeated parts will not be described again.

[0155] Further, in some embodiments, the first device determines a first feature according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal by using a third model, determines position information of the first feature in a coronary heart disease lesion space according to the first feature, and determines the maximum likelihood coronary heart disease lesion position of the user according to the position information of the first feature in the coronary heart disease lesion space. For example, the first feature can be a fully connected layer feature. The determination of the coronary heart disease lesion space can refer to the above related paragraphs, and the repeated parts will not be described herein.

[0156] As shown in Figure 8 , the first device inputs the abnormal signal and the reconstructed signal corresponding to the abnormal signal into the third model M3, for example, processes through a convolutional pooling layer, a fully connected layer, etc., to determine the position information of the fully connected layer feature in the coronary heart disease lesion space, and determine the maximum likelihood coronary heart disease lesion position of the user according to the position information of the fully connected layer feature in the coronary heart disease lesion space.

[0157] For example, the first device inputs the abnormal signal and the reconstructed signal corresponding to the abnormal signal into the third model M3, determines the feature 1 in Figure 8 , and determines that the feature 1 falls into the posterior wall lesion space according to the coordinates of the feature 1, and then can know that the maximum likelihood coronary heart disease lesion position of the user is the posterior wall, i.e., the user is likely to be a posterior wall lesion coronary heart disease patient. For another example, the first device inputs the abnormal signal and the reconstructed signal corresponding to the abnormal signal into the third model M3, determines the feature 2 in Figure 8 , and determines that the feature 2 does not fall into any coronary heart disease lesion space, so it can be known that the user is likely to be a patient with other diseases, but not a coronary heart disease patient.

[0158] Similarly, the above-mentioned third model is also applicable to the first device directly determining a first feature according to the ECG signal of one or more leads and the reconstructed signal corresponding to the ECG signal of one or more leads, determining position information of the first feature in a coronary heart disease lesion space according to the first feature, and determining the maximum likelihood coronary heart disease lesion position of the user according to the position information of the first feature in the coronary heart disease lesion space. The repeated parts will not be described herein.

[0159] The embodiments shown in Figure 7 will be described below in conjunction with specific embodiments.

[0160] Example 1: A user A collects the ECG signal of a certain lead or certain leads in the limb lead ECG signal by using an ECG wearable watch. For example, the user A wears the ECG wearable watch in a posture (the electrode in the watch contacts the preset position of the wrist or the preset position of the leg or the preset position of the chest) as shown in Figure 9A , so that the ECG wearable watch can collect the limb 6-lead ECG signal, and the specific process is as shown in Figure 9BThe limb 6-lead ECG signal is shown. The ECG wearable watch extracts the full segment features of the acquired 6-lead ECG signal through the first model M1 respectively, and reconstructs the full segment features of the ECG signal of each lead extracted through the second model M2, to obtain the reconstructed signal corresponding to the ECG signal of each lead, as shown in Figure 9C . The ECG wearable watch determines at least one abnormal signal according to the acquired ECG signal of each lead and the reconstructed signal corresponding to the ECG signal of each lead, for example, aVR, aVL respectively corresponding to the ECG signal as the abnormal signal, as shown in Figure 9D . Further, the ECG wearable watch determines the coronary heart disease incidence probability of user A as 76% according to the abnormal signal and the reconstructed signal corresponding to the abnormal signal using the third model M3 (as shown in Figure 9F . In addition, the fully connected layer features and the position information of the fully connected layer features in the coronary heart disease lesion space can be determined to further obtain the most likely coronary heart disease lesion position of user A. For example, the fully connected layer features fall into the lateral wall coronary heart disease lesion space, as shown in Figure 9E . Further, the ECG wearable device can also suggest the user to go to the hospital for coronary heart disease treatment.

[0161] Example 2: User B uses a heart rate chest strap to collect a certain lead or certain leads of the chest lead ECG signal. For example, the user wears a heart rate chest strap. The heart rate chest strap contacts the user B in a preset position (V1-V6) as shown in Figure 10A , and can collect 6-lead ECG signals of the chest, Figure 10B is a schematic diagram of the 6-lead ECG signal of the chest. The heart rate chest strap can transmit the collected 6-lead ECG signals of the chest to an ECG signal processing device, such as the mobile phone of user B. The mobile phone of user B extracts the full segment features of the acquired 6-lead ECG signals through the first model M1 respectively, and reconstructs the full segment features of the ECG signal of each lead extracted through the second model M2, to obtain the reconstructed signal corresponding to the ECG signal of each lead, as shown in Figure 10C . The ECG wearable device determines at least one abnormal signal according to the acquired ECG signal of each lead and the reconstructed signal corresponding to the ECG signal of each lead, for example, V1, V2, V6 respectively corresponding to the ECG signal as the abnormal signal, as shown in Figure 10D . Further, the mobile phone of user B inputs the abnormal signal and the reconstructed signal corresponding to the abnormal signal into the third model M3 to determine the coronary heart disease incidence probability of user B as 28% and the probability of other diseases as 52% by the ECG wearable device inputting the abnormal signal and the reconstructed signal corresponding to the abnormal signal into the third model M3, as shown in Figure 10FAs shown, in addition, the fully connected layer features and their corresponding location information in the coronary lesion space can also be determined. For example, the fully connected layer features do not fall into any coronary lesion space, such as... Figure 10E As shown. Furthermore, User B's phone can also prompt User B that although the risk of coronary heart disease is low, there is a risk of developing other diseases, and it is recommended that the user go to the hospital for further examination.

[0162] Example 3: User C uses an ECG wearable watch to collect one or more leads from the limbs, and a heart rate chest strap to collect one or more leads from the chest. For example, User C wears both an ECG wearable watch and a heart rate chest strap. Based on the above two examples, the ECG wearable watch can collect 6-lead ECG signals from the limbs, and the heart rate chest strap can collect 6-lead ECG signals from the chest. The heart rate chest strap sends the 6-lead ECG signals from the chest to the ECG wearable watch. Therefore, the ECG wearable watch acquires 12-lead ECG signals, including 6-lead ECG signals from the limbs and 6-lead ECG signals from the chest. Figure 11A This is a schematic diagram of a 12-lead ECG signal. The ECG wearable watch extracts full-segment features from the acquired 12-lead ECG signal using a first model M1, and reconstructs the extracted full-segment features from each ECG signal using a second model M2, obtaining the reconstructed signal corresponding to each ECG signal, such as... Figure 11B As shown. The ECG wearable watch determines at least one abnormal signal based on each acquired ECG signal and the corresponding reconstructed signal. For example, signals II, III, and aVF correspond to abnormal ECG signals, such as... Figure 11C As shown. Furthermore, the ECG wearable watch inputs abnormal signals and abnormal signals into the third model M3 to determine that user C's probability of developing coronary heart disease is 81%, as... Figure 11E As shown, in addition, the fully connected layer features and their corresponding positional information in the coronary lesion space can be determined to obtain the most likely location of the coronary lesion for user C. For example, the fully connected layer features fall into the inferior wall coronary lesion space, such as... Figure 11D As shown. Furthermore, ECG wearable devices can also advise users to visit a hospital for treatment of coronary heart disease.

[0163] It is understood that, in order to achieve the functions described in the above embodiments, those skilled in the art should readily recognize that, in conjunction with the units and method steps of the various examples disclosed in this application, this application can be implemented in hardware or a combination of hardware and computer software. Whether a certain function is executed in a hardware or computer software-driven hardware manner depends on the specific application scenario and design constraints of the technical solution.

[0164] Figure 12 andFigure 13 A possible structure of a communication apparatus is provided for embodiments of the present application. The communication apparatus can be used to implement the functions of the first device in the above-mentioned method embodiments, and thus can also achieve the beneficial effects possessed by the above-mentioned method embodiments. In embodiments of the present application, the communication apparatus can be an ECG signal acquisition device or an ECG signal processing device, and can also be a module (such as a chip) applied to an ECG signal acquisition device or an ECG signal processing device.

[0165] As shown in Figure 12 , the communication apparatus 1200 includes a processing unit 1210 and a transceiver unit 1220. The communication apparatus 1200 is used to implement the functions of the first device in the method embodiments shown in the above-mentioned Figure 7 .

[0166] When the communication apparatus 1200 is used to implement the functions of the first device in the method embodiments shown in the above-mentioned Figure 12 , the transceiver unit 1220 is used to acquire ECG signals of one or more leads of a user; the processing unit 1210 is used to determine reconstructed signals respectively corresponding to the ECG signals of the one or more leads according to the ECG signals of the one or more leads; wherein the ECG signals of the one or more leads include ECG signals of a first lead, and the reconstructed signal corresponding to the ECG signals of the first lead is obtained by signal reconstruction based on a feature of the ECG signals of the first lead; and the probability of coronary heart disease of the user is determined according to the ECG signals of the one or more leads and the reconstructed signals respectively corresponding thereto.

[0167] Exemplarily, the first device can be an ECG signal acquisition device, which can be a wearable device, such as a smart watch, a smart chest strap, a single-lead ECG watch, a multi-lead ECG watch, a chest ECG patch, a single-lead or multi-lead ECG patch, etc. The ECG signal acquisition device can perform single-lead or multi-lead ECG signal acquisition, and process the acquired ECG signals to obtain the probability of coronary heart disease of the user and other related information. The first device with a display screen can also display the obtained probability of coronary heart disease of the user and other related information through the display screen.

[0168] In addition, in some embodiments, the first device can acquire ECG signals, and acquire ECG signals acquired by other ECG signal acquisition devices. The first device processes the ECG signals acquired by itself and the ECG signals acquired by other ECG signal acquisition devices together to obtain the probability of coronary heart disease of the user and other related information.

[0169] In some embodiments, the first device can be an ECG signal processing device, which can process the acquired ECG signals to obtain information such as the probability of coronary heart disease of the user. The first device with a display screen can also display the obtained information such as the probability of coronary heart disease of the user through the display screen. Exemplarily, the first device herein can refer to a tablet, a computer, a television, etc. It can be understood that if the first device does not have a display screen, the information such as the probability of coronary heart disease of the user can also be sent to a display device for display.

[0170] In a possible design, the processing unit is configured to, when determining the probability of coronary heart disease of the user according to the ECG signals of the one or more leads and the respective reconstructed signals, determine an abnormal signal in the ECG signals of the one or more leads according to the ECG signals of the one or more leads and the respective reconstructed signals; and determine the probability of coronary heart disease of the user according to the abnormal signal and the corresponding reconstructed signal.

[0171] In a possible design, the processing unit is configured to, when determining the reconstructed signal corresponding to each of the ECG signals of the one or more leads according to the ECG signals of the one or more leads, extract a feature of the ECG signal of each of the one or more leads by using a first model; and reconstruct the signal according to the feature of the ECG signal of each of the one or more leads by using a second model to obtain the reconstructed signal corresponding to the ECG signal of each of the one or more leads.

[0172] The first model and the second model are obtained by training according to the ECG signals of the leads of a healthy population.

[0173] In a possible design, the processing unit is configured to, when determining the abnormal signal in the ECG signal of the one or more leads according to the ECG signal of the one or more leads and the corresponding reconstructed signal respectively, determine that the ECG signal of the first lead is an abnormal signal if the ECG signal of the first lead and the corresponding reconstructed signal satisfy one or more of the following abnormal signal conditions: there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the absolute value of the difference value is greater than a first threshold value; or there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the ratio of the absolute value of the difference value to a maximum value of the de-sign is greater than a second threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the sum of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a third threshold value; or the sum of the squares of the ratio of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than a fourth threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the standard deviation of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a fifth threshold value; or the standard deviation of the ratio of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than a sixth threshold value, where the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or the X1th percentile of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a seventh threshold value, where X1 is a preset value; or the X2th percentile of the ratio of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal to the maximum value of the de-sign is greater than an eighth threshold value, where X2 is a preset value, and the maximum value of the de-sign is a maximum value of the absolute value of the amplitude of the ECG signal of the first lead.

[0174] In a possible design, the processing unit is configured to, when determining the probability of coronary heart disease of the user according to the abnormal signal and the corresponding reconstructed signal, determine a target signal according to the abnormal signal and the corresponding reconstructed signal, where the target signal reflects the difference between the abnormal signal and the corresponding reconstructed signal; and determine the probability of coronary heart disease of the user according to the target signal and a third model.

[0175] The third model is obtained by training according to the first signal set and the second signal set, the first signal set includes target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, and the second signal set includes target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and corresponding reconstructed signals.

[0176] In a possible design, the processing unit is configured to: determine a target signal according to the abnormal signal and the corresponding reconstructed signal when determining the coronary heart disease suffering probability of the user according to the abnormal signal and the corresponding reconstructed signal, the target signal reflecting a difference between the abnormal signal and the corresponding reconstructed signal; and determine the coronary heart disease suffering probability of the user according to the target signal by using the third model.

[0177] The third model is obtained by training according to the first signal set, the second signal set and the third signal set, the first signal set includes target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and corresponding reconstructed signals, the second signal set includes target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and corresponding reconstructed signals, and the third signal set includes target signals corresponding to ECG signals of each lead of a population with other diseases, the target signals corresponding to the ECG signals of each lead of the population with other diseases are determined by the ECG signals of each lead of the population with other diseases and corresponding reconstructed signals.

[0178] In a possible design, the processing unit is configured to: determine a first feature according to the abnormal signal and the corresponding reconstructed signal by using the third model; determine position information of the first feature in a coronary heart disease lesion space according to the first feature; and determine a most likely coronary heart disease lesion position of the user according to the position information of the first feature in the coronary heart disease lesion space.

[0179] In a possible design, the coronary heart disease lesion space is obtained by training according to the first signal set and diagnosis labels of the ECG signals of each lead of the coronary heart disease patient.

[0180] In a possible design, the first feature is a fully connected layer feature.

[0181] More details of the processing unit 1210 and the transceiver unit 1220 can be referred to the description of the method embodiments shown above directly. Figure 7 More details of the processing unit 1210 and the transceiver unit 1220 can be referred to the description of the method embodiments shown above directly.

[0182] As shown in the method embodiments shown above, the communication device 1300 includes a processor 1310 and an interface circuit 1320. The processor 1310 and the interface circuit 1320 are coupled with each other. It can be understood that the interface circuit 1320 can be a transceiver or an input / output interface. Optionally, the communication device 1300 can further include a memory 1330 for storing instructions executed by the processor 1310 or storing input data required by the processor 1310 for executing instructions or storing data generated after the processor 1310 executes instructions. Figure 13 When the communication device 1300 is used to implement the method shown above, the processor 1310 is configured to implement the functions of the processing unit 1210, and the interface circuit 1320 is configured to implement the functions of the transceiver unit 1220.

[0183] Figure 12 It can be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0184] It can be understood that the processor in the embodiments of the present application can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.

[0185] ​The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in a random access memory (RAM), a flash memory, a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a register, a hard disk, a mobile hard disk, a CD-ROM, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor, so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC. In addition, the ASIC can be located in a network device or a terminal device. Of course, the processor and the storage medium can also exist as discrete components in the network device or the terminal device.

[0186] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer programs or instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are performed. The computer can be a general purpose computer, a special purpose computer, a computer network, a network device, a user equipment, or other programmable apparatus. The computer programs or instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium, for example, the computer programs or instructions can be transferred from one website site, computer, server, or data center to another website site, computer, server, or data center through a wired or wireless manner. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be a magnetic medium, such as a floppy disk, a hard disk, a magnetic tape; an optical medium, such as a digital video disc (DVD); or a semiconductor medium, such as a solid state drive (SSD).

[0187] In the various embodiments of the present application, the terms and / or descriptions between different embodiments are consistent and can be referred to each other if there is no special description and logical conflict, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0188] In the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the relationship between the associated objects, which means that there can be three kinds of relationships, for example, A and / or B can represent the following three cases: A exists alone, A and B exist together, and B exists alone, where A and B can be singular or plural. In the textual description of the present application, the character " / " generally represents that the associated objects before and after are in an "or" relationship; in the formula of the present application, the character " / " represents that the associated objects before and after are in a "division" relationship.

[0189] It can be understood that various numbers involved in the embodiments of the present application are only distinguished for convenience of description, and are not used to limit the scope of the embodiments of the present application. The size of the serial number of the above processes does not mean the order of execution, and the execution order of the processes should be determined according to their functions and inherent logic.

Claims

1. An apparatus for determining a probability of suffering from coronary heart disease, characterized in that, The device comprises: a transceiver unit configured to acquire ECG signals of one or more leads of a user; a processing unit configured to determine, according to the ECG signals of the one or more leads, reconstructed signals respectively corresponding to the ECG signals of the one or more leads; wherein the ECG signals of the one or more leads comprise an ECG signal of a first lead, and the reconstructed signal corresponding to the ECG signal of the first lead is obtained by signal reconstruction based on a feature of the ECG signal of the first lead; the processing unit is configured to, when determining the reconstructed signals respectively corresponding to the ECG signals of the one or more leads according to the ECG signals of the one or more leads, extract, by using a first model, the feature of the ECG signal of each lead in the ECG signals of the one or more leads; and perform signal reconstruction according to the feature of the ECG signal of each lead by using a second model to obtain the corresponding reconstructed signal; wherein the first model and the second model are obtained by training according to the ECG signals of each lead of a healthy population; wherein the reconstructed signals respectively corresponding to the ECG signals of each lead of the healthy population obtained by the first model and the second model restore the ECG signals of each lead of the healthy population, and the reconstructed signals respectively corresponding to the ECG signals of each lead of a coronary heart disease patient obtained by the first model and the second model cannot restore the ECG signals of each lead of the coronary heart disease patient; the ECG signals of each lead of the healthy population are substantially consistent with the reconstructed signals respectively corresponding to the ECG signals of each lead of the healthy population. The processing unit is configured to determine, according to the ECG signals of the one or more leads and the reconstructed signals respectively corresponding thereto, a coronary heart disease incidence probability of the user and a maximum likelihood coronary heart disease lesion position of the user.

2. The apparatus of claim 1, wherein, The processing unit is configured to, when determining the coronary heart disease incidence probability of the user according to the ECG signals of the one or more leads and the reconstructed signals respectively corresponding thereto, determine an abnormal signal in the ECG signals of the one or more leads according to the ECG signals of the one or more leads and the reconstructed signals respectively corresponding thereto; and determine the coronary heart disease incidence probability of the user according to the abnormal signal and the corresponding reconstructed signal.

3. The apparatus of claim 2, wherein, The processing unit is configured to, when determining the abnormal signal in the ECG signals of the one or more leads according to the ECG signals of the one or more leads and the reconstructed signals respectively corresponding thereto, determine that the ECG signal of the first lead is an abnormal signal if the ECG signal of the first lead and the corresponding reconstructed signal satisfy one or more abnormal signal conditions as follows: there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the absolute value of the difference value is greater than a first threshold value; or, there is a difference value in the difference between the ECG signal of the first lead and the corresponding reconstructed signal, and the ratio of the absolute value of the difference value to a de-sign maximum value is greater than a second threshold value; wherein the de-sign maximum value refers to a maximum value of the absolute value of the amplitude of the ECG signal of the first lead; or, the sum of the absolute values of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a third threshold value. Or, the square sum of the ratio of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal and the maximum value of the de-sign, is greater than a fourth threshold value; wherein the maximum value of the de-sign refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead; Or, the standard deviation of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a fifth threshold value; Or, the standard deviation of the ratio of the difference between the ECG signal of the first lead and the corresponding reconstructed signal and the maximum value of the de-sign is greater than a sixth threshold value; wherein the maximum value of the de-sign refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead; Or, the X1 percentile of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal is greater than a seventh threshold value, X1 being a preset value; Or, the X2 percentile of the ratio of the absolute value of the difference between the ECG signal of the first lead and the corresponding reconstructed signal and the maximum value of the de-sign is greater than an eighth threshold value, X2 being a preset value; wherein the maximum value of the de-sign refers to the maximum value of the absolute value of the amplitude of the ECG signal of the first lead.

4. The apparatus of claim 2, wherein, The processing unit is configured to determine a target signal according to the abnormal signal and the corresponding reconstructed signal when determining the coronary heart disease risk probability of the user according to the abnormal signal and the corresponding reconstructed signal, the target signal reflecting the difference between the abnormal signal and the corresponding reconstructed signal; and determine the coronary heart disease risk probability of the user according to the third model and the target signal. The third model is obtained by training a first signal set and a second signal set, wherein the first signal set includes target signals corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and the corresponding reconstructed signals, and the second signal set includes target signals corresponding to ECG signals of each lead of a healthy population, the target signals corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and the corresponding reconstructed signals.

5. The apparatus of claim 2, wherein, The processing unit is configured to determine a target signal according to the abnormal signal and the corresponding reconstructed signal when determining the coronary heart disease risk probability of the user according to the abnormal signal and the corresponding reconstructed signal, the target signal reflecting the difference between the abnormal signal and the corresponding reconstructed signal; and determine the coronary heart disease risk probability of the user according to the third model and the target signal. The third model is obtained by training according to a first signal set, a second signal set and a third signal set, wherein the first signal set comprises target signals respectively corresponding to ECG signals of each lead of a coronary heart disease patient, the target signals respectively corresponding to the ECG signals of each lead of the coronary heart disease patient are determined by the ECG signals of each lead of the coronary heart disease patient and corresponding reconstruction signals, the second signal set comprises target signals respectively corresponding to ECG signals of each lead of a healthy population, the target signals respectively corresponding to the ECG signals of each lead of the healthy population are determined by the ECG signals of each lead of the healthy population and corresponding reconstruction signals, and the third signal set comprises target signals respectively corresponding to ECG signals of each lead of a population with other diseases, the target signals respectively corresponding to the ECG signals of each lead of the population with other diseases are determined by the ECG signals of each lead of the population with other diseases and corresponding reconstruction signals.

6. The apparatus of claim 4 or 5, wherein, The processing unit is configured to determine a first feature by using the third model according to the abnormal signal and the corresponding reconstruction signal, determine position information of the first feature in a coronary heart disease lesion space according to the first feature, and determine a most likely coronary heart disease lesion position of the user according to the position information of the first feature in the coronary heart disease lesion space.

7. The apparatus of claim 6, wherein, The coronary heart disease lesion space is obtained by training according to the first signal set and diagnosis labels of the ECG signals of each lead of the coronary heart disease patient.

8. The apparatus of claim 6, wherein, The first feature is a fully connected layer feature.

Images