Electronic device and method for predicting blockage of coronary arteries

Abstract

The present invention provides an electronic device and method for predicting the blockage of coronary arteries. The method comprises: obtaining a plurality of electrocardiogram data corresponding to a set of coronary arteries; generating a plurality of first probabilities corresponding to the plurality of electrocardiogram data respectively according to the plurality of electrocardiogram data and a first stage model, generating a first judgment result according to the plurality of first probabilities, and selecting a first data subset corresponding to the first probability subset from the plurality of electrocardiogram data in response to each of the first probability subsets of the plurality of first probabilities being greater than a first threshold; generating a plurality of second probabilities corresponding to the first data subset according to the first data subset and a second stage model, and generating a second judgment result according to the plurality of second probabilities.

Description

Technical Field

[0001] The present invention relates to an electronic device and method for predicting blockage of a coronary artery. Background Art

[0002] The mortality rate of cardiovascular diseases has been rising year by year. Common cardiovascular diseases include coronary artery disease (CAD) or atrial fibrillation. There are many ways to detect cardiovascular diseases in the medical field, such as blood tests, electrocardiograms, or cardiac catheterization. The cause of coronary artery disease is that the coronary arteries surrounding the surface of the heart are too narrow or blocked, resulting in the inability of the arterial blood flow that supplies oxygen and nutrients to the heart to function normally. Among the current detection methods, cardiac catheterization is the most accurate method to detect coronary artery disease.

[0003] However, cardiac catheterization is an invasive procedure that carries risks and may cause complications. If only necessary patients are examined and operated on, not only can medical costs be avoided, but patients can also be kept at risk. Therefore, if the coronary artery is blocked, it can help doctors more accurately determine whether a cardiac catheterization procedure is needed. In addition, if the location of the coronary artery blockage can be predicted, doctors can also have a better understanding of the patient's condition. Summary of the invention

[0004] The present invention provides an electronic device and method for predicting coronary artery blockage, which can predict whether a coronary artery is blocked, the type of blocked coronary artery or the location of blocked coronary artery.

[0005] An electronic device for predicting the blockage of a coronary artery of the present invention comprises a processor, a storage medium and a transceiver. The transceiver receives a plurality of electrocardiogram data corresponding to a coronary artery set. The storage medium stores a plurality of modules. The processor couples the storage medium and the transceiver, and accesses and executes the plurality of modules, wherein the plurality of modules comprises a first prediction module, a second prediction module and an output module. The first prediction module generates at least one first probability vector corresponding to at least one first stage model according to a plurality of electrocardiogram data and at least one first stage model, generates a first judgment result according to at least one first probability vector, and selects a first data subset corresponding to the first subset from a plurality of electrocardiogram data in response to each of the first subsets of at least one first probability vector being greater than a first threshold, wherein the first judgment result indicates the probability of blockage of the coronary artery set. The second prediction module generates at least one second probability vector corresponding to at least one second stage model according to the first data subset and at least one second stage model, and generates a second judgment result according to at least one second probability vector, wherein the second judgment result indicates the probability of blockage of the first coronary artery in the coronary artery set. The output module outputs the first judgment result and the second judgment result through the transceiver.

[0006] In one embodiment of the present invention, the second prediction module selects a second data subset corresponding to the second subset from the first data subset in response to each of the second subsets of the at least one second probability vector being greater than a second threshold, wherein the plurality of modules further comprises a third prediction module. The third prediction module generates at least one third probability vector corresponding to the at least one third stage model according to the second data subset and the at least one third stage model, and generates a third judgment result according to the at least one third probability vector, wherein the third judgment result indicates the probability of blockage occurring at the first position of the first coronary artery, wherein the output module outputs the third judgment result through the transceiver.

[0007] In one embodiment of the present invention, the above-mentioned transceiver receives multiple second electrocardiogram data corresponding to a coronary artery set, wherein the multiple electrocardiogram data correspond to a first lead, wherein the multiple second electrocardiogram data correspond to a second lead, wherein the first prediction module generates at least one fourth probability vector corresponding to at least one first stage model based on the multiple second electrocardiogram data and at least one first stage model, and generates a first judgment result based on the at least one first probability vector and the at least one fourth probability vector.

[0008] In one embodiment of the present invention, the above-mentioned first prediction module selects a fourth data subset corresponding to the fourth subset from multiple second electrocardiogram data in response to each of the fourth subset of at least one fourth probability vector being greater than the first threshold, wherein the second prediction module generates at least one fifth probability vector corresponding to at least one second stage model based on the fourth data subset and at least one second stage model, and generates a second judgment result based on the at least one second probability vector and the at least one fifth probability vector.

[0009] In one embodiment of the present invention, the above-mentioned second prediction module selects a fifth data subset corresponding to the fifth subset from the fourth data subset in response to each of the fifth subsets of at least one fifth probability vector being greater than the second threshold, wherein the third prediction module generates at least one sixth probability vector corresponding to at least one third stage model based on the fifth data subset and at least one third stage model, and generates a third judgment result based on the at least one third probability vector and the at least one sixth probability vector.

[0010] In one embodiment of the present invention, the first prediction module performs baseline drift removal, denoising and wavelet transformation on a plurality of training data to generate a plurality of first calibrated training data, and trains at least one first stage model according to the plurality of first calibrated training data.

[0011] In one embodiment of the present invention, the first prediction module trains at least one first stage model according to at least one feature, wherein the at least one feature is associated with at least one of the following: P wave, Q wave, R wave, S wave and T wave.

[0012] In one embodiment of the present invention, the at least one first stage model, the at least one second stage model and the at least one third stage model correspond to machine learning algorithms.

[0013] In one embodiment of the present invention, the first coronary artery includes one of the following: the left main coronary artery, the anterior descending branch of the left coronary artery, the circumflex branch of the left coronary artery, and the right coronary artery.

[0014] A method for predicting coronary artery blockage of the present invention comprises: obtaining a plurality of electrocardiogram data corresponding to a set of coronary arteries; generating a plurality of first probabilities corresponding to the plurality of electrocardiogram data respectively according to the plurality of electrocardiogram data and at least one first stage model, generating a first judgment result according to the plurality of first probabilities, and selecting a first data subset corresponding to the first probability subset from the plurality of electrocardiogram data in response to each of the first probability subsets of the plurality of first probabilities being greater than a first threshold, wherein the first judgment result indicates the probability of blockage occurring in the set of coronary arteries; generating a plurality of second probabilities corresponding to the first data subset according to the first data subset and at least one second stage model, and generating a second judgment result according to the plurality of second probabilities, wherein the second judgment result indicates the probability of blockage occurring in a first coronary artery in the set of coronary arteries; and outputting the first judgment result and the second judgment result.

[0015] Based on the above, the present invention can use machine learning models to predict the type of blocked coronary artery or the location of blocked coronary artery. Medical personnel can use the judgment results generated by the present invention to more accurately assess whether the subject suffers from coronary artery disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of an electronic device is shown according to an embodiment of the present invention;

[0017] Figure 2 A flow chart of a method for predicting coronary artery obstruction is shown according to an embodiment of the present invention;

[0018] Figure 3 According to an embodiment of the present invention, a flow chart of a method for generating a first judgment result based on an electrocardiogram of multiple leads is shown;

[0019] Figure 4 A flowchart showing a method for generating a second judgment result from an electrocardiogram of multiple leads according to an embodiment of the present invention;

[0020] Figure 5 A flowchart showing a method for generating a third judgment result from an electrocardiogram of multiple leads according to an embodiment of the present invention;

[0021] Figure 6 A flow chart of a method for predicting coronary artery obstruction is shown according to another embodiment of the present invention.

[0022] Description of Reference Numerals

[0023] 100: Electronic devices

[0024] 110: Processor

[0025] 120: Storage media

[0026] 121: First prediction module

[0027] 122: Second prediction module

[0028] 123: The third prediction module

[0029] 124: Output module

[0030] 130: Transceiver

[0031] S201, S202, S203, S204, S205, S206, S207, S208, S209, S210, S211, S212, S301, S302, S303, S304, S401, S402, S403, S404, S501, S502, S503, S504, S601, S602, S603, S604: Steps DETAILED DESCRIPTION

[0032] Reference will now be made in detail to exemplary embodiments of the present invention, examples of which are illustrated in the accompanying drawings. Whenever possible, the same reference numerals are used in the drawings and the description to refer to the same or like parts.

[0033] Figure 1 A schematic diagram of an electronic device 100 is shown according to an embodiment of the present invention. The electronic device 100 is suitable for predicting blockage of a coronary artery. The electronic device 100 may include a processor 110 , a storage medium 120 , and a transceiver 130 .

[0034] The processor 110 is, for example, a central processing unit (CPU), or other programmable general-purpose or special-purpose micro control unit (MCU), microprocessor, digital signal processor (DSP), programmable controller, application specific integrated circuit (ASIC), graphics processing unit (GPU), image signal processor (ISP), image processing unit (IPU), arithmetic logic unit (ALU), complex programmable logic device (CPLD), field programmable gate array (FPGA), or other similar components or combinations thereof. The processor 110 may be coupled to the storage medium 120 and the transceiver 130, and access and execute multiple modules and various applications stored in the storage medium 120.

[0035] The storage medium 120 is, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), flash memory, hard disk drive (HDD), solid state drive (SSD) or similar components or a combination of the above components, and is used to store multiple modules or various applications that can be executed by the processor 110. In this embodiment, the storage medium 120 can store multiple modules including a first prediction module 121, a second prediction module 122, a third prediction module 123 and an output module 124, and their functions will be described later.

[0036] The transceiver 130 transmits and receives signals in a wireless or wired manner. The transceiver 130 may also perform operations such as low noise amplification, impedance matching, mixing, up or down frequency conversion, filtering, amplification, and the like.

[0037] Figure 2 According to an embodiment of the present invention, a flow chart of a method for predicting coronary artery obstruction is shown, wherein the method may be performed as follows: Figure 1 The electronic device 100 is shown as an implementation.

[0038] In step S201, the electronic device 100 may receive X electrocardiogram (ECG) data corresponding to a coronary artery set through the transceiver 130, where X may be any positive integer. The coronary artery set may include a plurality of coronary arteries, such as the left main coronary artery (LM), the left anterior descending branch of the left coronary artery (LAD), the left circumflex artery (LCX), and / or the right coronary artery (RCA).

[0039] In steps S202 and S203, the first prediction module 121 may generate at least one first probability vector corresponding to at least one first stage model according to X pieces of electrocardiogram data and at least one first stage model, wherein the at least one first probability vector may include X first probabilities respectively corresponding to the X pieces of electrocardiogram data.

[0040] Specifically, in step S202, the first prediction module 121 may input X pieces of electrocardiogram data into N first-stage models, where N may be any positive integer. Then, in step S203, the N first-stage models may generate N first probability vectors according to the X pieces of electrocardiogram data. Each of the N probability vectors may include X first probabilities corresponding to the X pieces of electrocardiogram data.

[0041] The first stage model can be used to determine the probability of blockage of any one of the coronary artery set according to the electrocardiogram data. After a piece of electrocardiogram data is input into the first stage model, the first stage model can output a first probability corresponding to the electrocardiogram data, wherein the first probability can represent the probability of blockage of any one of the coronary artery set.

[0042] The first stage model is, for example, a machine learning algorithm, wherein the machine learning algorithm corresponds to, for example, a random forest algorithm (RF), a support vector machine (SVM) algorithm, an extreme gradient boosting (XGBoost) algorithm, or a deep learning algorithm, but the present invention is not limited thereto. The deep learning algorithm corresponds to, for example, a convolutional neural network (CNN) or a long short-term memory (LSTM) algorithm, but the present invention is not limited thereto.

[0043] In one embodiment, the first prediction module 121 may train the first stage model according to a plurality of training data. Each piece of training data in the plurality of training data may include electrocardiogram data and a label indicating whether any one of the coronary artery set of the subject is blocked. Specifically, the first prediction module 121 may perform baseline wandering removal, denoising, and wavelet transform on the plurality of training data to generate a plurality of first corrected training data. Then, the first prediction module 121 may train the first stage model according to the plurality of first corrected training data.

[0044] In one embodiment, the first prediction module 121 may remove the baseline drift of the plurality of training data according to a cubic splines smoothing algorithm.

[0045] In one embodiment, the first prediction module 121 may perform denoising on the plurality of training data according to a Butterworth filter.

[0046] In one embodiment, the first prediction module 121 may train the first stage model by using one or more features in the training data, wherein the one or more features are, for example, associated with the P wave, Q wave, R wave, S wave, or T wave. For example, the above features may include a standard deviation of the RR interval calculated based on the R wave or a slope of the ST wave calculated based on the S wave and the T wave, but the present invention is not limited thereto.

[0047] In step S204, the first prediction module 121 may generate a first judgment result according to the N first probability vectors. The output module 124 may output the first judgment result through the transceiver 130. The first judgment result may include a probability S1 that any one of the coronary artery set of the subject is blocked. For example, the first judgment result may be a first probability that the coronary artery set is blocked. After the doctor receives the first judgment result from the electronic device 100 through the terminal device, the doctor may decide whether to conduct further examination of the coronary artery of the subject according to the first judgment result.

[0048] In one embodiment, the first prediction module 121 may calculate the probability S1 that any one of the coronary artery sets is blocked according to formula (1) and formula (2) as shown below, wherein p1(I,J) is the Ith first probability in the Jth first probability vector among the N first probability vectors, and w1(J) is the weight corresponding to the Jth first stage model among the N first stage models. The first prediction module 121 may determine the weight corresponding to the first stage model based on the performance index of the first stage model. For example, among the N first stage models, if the performance index of the first first stage model is greater than the performance index of the Nth first stage model, the first prediction module 121 may determine that the weight w1(1) is greater than the weight w1(N). The performance index may correspond to a parameter in a confusion matrix, such as accuracy (ACC), precision, recall rate, false positive (FP), or F1 score.

[0049]

[0050] In step S205, the first prediction module 121 may select a first data subset corresponding to the first subset from the X-stroke electrocardiogram data in response to each of the first subset of the first probability vector (composed of one or more first probabilities included in the first probability vector) being greater than the first threshold. When there are multiple first probability vectors, the first prediction module 121 may calculate an average first probability vector based on the multiple first probability vectors, and may select a first data subset corresponding to the first subset from the X-stroke electrocardiogram data in response to each of the first subset of the average first probability vector (composed of one or more average first probabilities included in the average first probability vector) being greater than the first threshold.

[0051] For example, assuming that the first threshold is 30%, and Y first probabilities among the X first probabilities of the first probability vector are greater than 30%, where Y may be a positive integer less than or equal to X. Accordingly, the first prediction module 121 may select Y pieces of electrocardiogram data corresponding to the Y first probabilities from the X pieces of electrocardiogram data as the first data subset.

[0052] In steps S206 and S207, the second prediction module 122 can generate at least one second probability vector corresponding to at least one second stage model based on the first data subset (i.e., Y pen electrocardiogram data) and at least one second stage model, wherein the at least one second probability vector may include Y second probabilities corresponding to the first data subset.

[0053] Specifically, in step S206, the second prediction module 122 may input the first data subset (i.e., Y pen electrocardiogram data) into M second stage models, where M may be any positive integer. Then, in step S207, the M second stage models may generate M second probability vectors based on the first data subset. Each of the M second probability vectors may include Y second probabilities corresponding to the Y pen electrocardiogram data.

[0054] The second stage model can be used to determine the probability of blockage of a specific coronary artery in the coronary artery set based on the electrocardiogram data. For example, the second stage model can determine the probability of blockage of a specific coronary artery among the left main coronary artery, the anterior descending branch of the left coronary artery, the circumflex branch of the left coronary artery, and the right coronary artery based on the electrocardiogram data. After a piece of electrocardiogram data is input into the second stage model, the second stage model can output a second probability corresponding to the piece of electrocardiogram data, wherein the second probability can represent the probability of blockage of the specific coronary artery.

[0055] The second stage model is, for example, a machine learning algorithm, wherein the machine learning algorithm corresponds to, for example, a random forest algorithm, a support vector machine algorithm, an extreme gradient boosting algorithm, or a deep learning algorithm, but the present invention is not limited thereto. The deep learning algorithm corresponds to, for example, a convolutional neural network or a long short-term memory algorithm, but the present invention is not limited thereto.

[0056] In one embodiment, the second prediction module 122 may train the second stage model according to a plurality of training data. Each piece of training data in the plurality of training data may include electrocardiogram data and a label indicating whether a specific coronary artery in a set of coronary arteries of the subject is blocked. Specifically, the second prediction module 122 may perform baseline drift removal, denoising, and wavelet transformation on the plurality of training data to generate a plurality of second corrected training data. Then, the second prediction module 122 may train the second stage model according to the plurality of second corrected training data.

[0057] In one embodiment, the second prediction module 122 may remove the baseline drift of the plurality of training data according to a smoothing cubic spline algorithm.

[0058] In one embodiment, the second prediction module 122 may perform denoising on the plurality of training data according to a Butterworth filter.

[0059] In one embodiment, the second prediction module 122 may train the second stage model by one or more features in the training data, wherein the one or more features are, for example, associated with the P wave, Q wave, R wave, S wave, or T wave. For example, the above-mentioned features may include the standard deviation of the RR interval calculated based on the R wave or the slope of the ST wave calculated based on the S wave and the T wave, but the present invention is not limited thereto.

[0060] In step S208, the second prediction module 122 may generate a second judgment result according to the M second probability vectors. The output module 124 may output the second judgment result through the transceiver 130. The first judgment result may include the probability S2 that a specific coronary artery in the coronary artery set of the subject is blocked.

[0061] In one embodiment, the second prediction module 122 may calculate the probability S2 that a specific coronary artery in the coronary artery set is blocked according to formula (3) and formula (4) as shown below, where p2(I,J) is the Ith second probability in the Jth second probability vector in the M second probability vectors, and w2(J) is the weight corresponding to the Jth second stage model in the M second stage models. The second prediction module 122 may determine the weight corresponding to the second stage model based on the performance indicator of the second stage model. For example, among the M second stage models, if the performance indicator of the first second stage model is greater than the performance indicator of the Mth second stage model, the second prediction module 122 may determine that the weight w2(1) is greater than the weight w2(M). The performance indicator may correspond to a parameter in a confusion matrix, such as accuracy, precision, recall rate, false positive or F1 score.

[0062]

[0063] In step S209, the second prediction module 122 may select a second data subset corresponding to the second subset from the Y-pen electrocardiogram data in response to each of the second subset of the second probability vector (composed of one or more second probabilities included in the second probability vector) being greater than the second threshold. When there are multiple second probability vectors, the second prediction module 122 may calculate an average second probability vector based on the multiple second probability vectors, and may select a second data subset corresponding to the second subset from the Y-pen electrocardiogram data in response to each of the second subset of the average second probability vector (composed of one or more average second probabilities included in the average second probability vector) being greater than the second threshold.

[0064] For example, assuming that the second threshold is 20%, and among the Y second probabilities of the second probability vector, Z second probabilities are greater than 20%, where Z may be a positive integer less than or equal to Y. Accordingly, the second prediction module 122 may select Z pen electrocardiogram data corresponding to the Z second probabilities from the Y pen electrocardiogram data as the second data subset.

[0065] In steps S210 and S211, the third prediction module 123 can generate at least one third probability vector corresponding to at least one third stage model based on the second data subset (i.e., Z pen electrocardiogram data) and at least one third stage model, wherein the at least one third probability vector may include Z third probabilities corresponding to the second data subset.

[0066] Specifically, in step S210, the third prediction module 123 may input the second data subset (i.e., Z-pen electrocardiogram data) into K third-stage models, where K may be any positive integer. Then, in step S211, the K third-stage models may generate K third probability vectors based on the second data subset. Each of the K third probability vectors may include Z third probabilities corresponding to the Z-pen electrocardiogram data.

[0067] The third stage model can be used to determine the probability of blockage at a specific location in a specific coronary artery based on the electrocardiogram data. For example, the third stage model can determine the probability of blockage at the anterior, middle or posterior segment of the left main coronary artery based on the electrocardiogram data. It is worth noting that the above-mentioned specific location is not limited to the anterior, middle or posterior segment of the left main coronary artery. After a piece of electrocardiogram data is input into the third stage model, the third stage model can output a third probability corresponding to the electrocardiogram data, wherein the third probability can represent the probability of blockage at a specific location in the specific coronary artery.

[0068] The third stage model is, for example, a machine learning algorithm, wherein the machine learning algorithm corresponds to, for example, a random forest algorithm, a support vector machine algorithm, an extreme gradient boosting algorithm, or a deep learning algorithm, but the present invention is not limited thereto. The deep learning algorithm corresponds to, for example, a convolutional neural network or a long short-term memory algorithm, but the present invention is not limited thereto.

[0069] In one embodiment, the third prediction module 123 may train the third stage model according to a plurality of training data. Each piece of training data in the plurality of training data may include electrocardiogram data and a label indicating whether a specific position in a specific coronary artery of the subject is blocked. Specifically, the third prediction module 123 may perform baseline drift removal, denoising, and wavelet transformation on the plurality of training data to generate a plurality of third corrected training data. Then, the third prediction module 123 may train the third stage model according to the plurality of third corrected training data.

[0070] In one embodiment, the third prediction module 123 may remove the baseline drift of the plurality of training data according to a smoothing cubic spline algorithm.

[0071] In one embodiment, the third prediction module 123 may perform denoising on the plurality of training data according to a Butterworth filter.

[0072] In one embodiment, the third prediction module 123 may train the third stage model by one or more features in the training data, wherein the one or more features are, for example, associated with the P wave, Q wave, R wave, S wave, or T wave. For example, the above-mentioned features may include the standard deviation of the RR interval calculated based on the R wave or the slope of the ST wave calculated based on the S wave and the T wave, but the present invention is not limited thereto.

[0073] In step S212, the third prediction module 123 may generate a third judgment result according to the K third probability vectors. The output module 124 may output the third judgment result through the transceiver 130. The third judgment result may include the probability S3 that a blockage occurs at a specific location in a specific coronary artery of the subject.

[0074] In one embodiment, the third prediction module 123 may calculate the probability S3 of blockage occurring at a specific location in a specific coronary artery according to formula (5) and formula (6) as shown below, wherein p3(I,J) is the Ith third probability in the Jth third probability vector among the K third probability vectors, and w3(J) is the weight corresponding to the Jth third stage model among the K third stage models. The third prediction module 123 may determine the weight corresponding to the third stage model based on the performance indicator of the third stage model. For example, among the K third stage models, if the performance indicator of the first third stage model is greater than the performance indicator of the Kth third stage model, the third prediction module 123 may determine that the weight w3(1) is greater than the weight w3(K). The performance indicator may correspond to a parameter in a confusion matrix, such as accuracy, precision, recall rate, false positive or F1 score.

[0075]

[0076] The electrocardiogram may correspond to a variety of different leads. For example, the electrocardiogram may correspond to lead I, II, III, aVL, aVF, aVR, V1, V2, V3, V4, V5, or V6. In some embodiments, the electronic device 100 may generate a first judgment result, a second judgment result, or a third judgment result based on the electrocardiograms of the various different leads.

[0077] Figure 3 According to an embodiment of the present invention, a flow chart of a method for generating a first judgment result based on an electrocardiogram with multiple leads is shown, wherein the method can be performed as follows: Figure 1 The electronic device 100 is shown as an implementation.

[0078] In step S301, the electronic device 100 may receive L electrocardiogram data sets corresponding to L leads respectively through the transceiver 130, wherein each of the L electrocardiogram data sets may include X pieces of electrocardiogram data, wherein L and X may be any positive integers. For example, the electronic device 100 may receive an electrocardiogram data set corresponding to a first lead through the transceiver 130, and may receive an electrocardiogram data set corresponding to a second lead through the transceiver 130.

[0079] In step S302, the first prediction module 121 may input the L electrocardiogram data sets into at least one first stage model respectively. For example, the first prediction module 121 may input the electrocardiogram data set corresponding to the first lead into N first stage models, and may input the electrocardiogram data set corresponding to the Lth lead into N first stage models, where N may be any positive integer.

[0080] In step S303, each of the N first-stage models may generate L probability vectors corresponding to the L leads, wherein each of the L probability vectors may include X probabilities corresponding to X pieces of electrocardiogram data. For example, the first first-stage model may generate L first probability vectors, and each of the L first probability vectors may include X probabilities corresponding to X pieces of electrocardiogram data. The Nth first-stage model may generate L Nth probability vectors, and each of the L Nth probability vectors may include X probabilities corresponding to X pieces of electrocardiogram data.

[0081] In step S304, the first prediction module 121 may generate a first judgment result according to N probability vector sets, wherein each of the N probability vector sets may include L probability vectors corresponding to the L leads. The N probability vector sets may be outputs of the N first stage models in step S303.

[0082] Figure 4 According to an embodiment of the present invention, a flow chart of a method for generating a second judgment result from an electrocardiogram with multiple leads is shown, wherein the method can be performed as follows: Figure 1 The electronic device 100 is shown as an implementation.

[0083] In step S401 , the electronic device 100 may receive L electrocardiogram data sets corresponding to L leads respectively through the transceiver 130 , wherein each of the L electrocardiogram data sets may include Y pen electrocardiogram data, where L and Y may be any positive integers.

[0084] In one embodiment, each of the L electrocardiogram data sets may include X pieces of electrocardiogram data, as described in step S301. The first prediction module 121 may select Y pieces of electrocardiogram data from the X pieces of electrocardiogram data included in each of the L electrocardiogram data sets according to the method shown in step S205, and delete the unselected electrocardiogram data, where Y may be a positive integer less than or equal to X.

[0085] In step S402, the second prediction module 122 may input the L electrocardiogram data sets into at least one second stage model respectively. For example, the second prediction module 122 may input the electrocardiogram data set corresponding to the first lead into M second stage models, and may input the electrocardiogram data set corresponding to the Lth lead into M second stage models, where M may be any positive integer.

[0086] In step S403, each of the M second-stage models may generate L probability vectors corresponding to the L leads, wherein each of the L probability vectors may include Y probabilities corresponding to the Y pen electrocardiogram data. For example, the first second-stage model may generate L first probability vectors, and each of the L first probability vectors may include Y probabilities corresponding to the Y pen electrocardiogram data. The Mth second-stage model may generate L Mth probability vectors, and each of the L Mth probability vectors may include Y probabilities corresponding to the Y pen electrocardiogram data.

[0087] In step S404, the second prediction module 122 may generate a second judgment result according to M probability vector sets, wherein each of the M probability vector sets may include L probability vectors corresponding to the L leads. The M probability vector sets may be outputs of the M second stage models in step S403.

[0088] Figure 5 According to an embodiment of the present invention, a flow chart of a method for generating a third judgment result from an electrocardiogram with multiple leads is shown, wherein the method can be performed as follows: Figure 1 The electronic device 100 is shown as an implementation.

[0089] In step S501 , the electronic device 100 may receive L electrocardiogram data sets corresponding to L leads respectively through the transceiver 130 , wherein each of the L electrocardiogram data sets may include Z pen electrocardiogram data, where L and Z may be any positive integers.

[0090] In one embodiment, each of the L electrocardiogram data sets may include Y pen electrocardiogram data, as described in step S401. The third prediction module 123 may select Z pen electrocardiogram data from the Y pen electrocardiogram data included in each of the L electrocardiogram data sets according to the method shown in step S207, and delete the unselected electrocardiogram data, where Z may be a positive integer less than or equal to Y.

[0091] In step S502, the third prediction module 123 may input the L electrocardiogram data sets into at least one third stage model respectively. For example, the third prediction module 123 may input the electrocardiogram data set corresponding to the first lead into K third stage models, and may input the electrocardiogram data set corresponding to the Lth lead into K third stage models, where K may be any positive integer.

[0092] In step S503, each of the K third-stage models may generate L probability vectors corresponding to the L leads, wherein each of the L probability vectors may include Z probabilities corresponding to the Z pen electrocardiogram data. For example, the first third-stage model may generate L first probability vectors, and each of the L first probability vectors may include Z probabilities corresponding to the Z pen electrocardiogram data. The Kth third-stage model may generate L Kth probability vectors, and each of the L Kth probability vectors may include Z probabilities corresponding to the Z pen electrocardiogram data.

[0093] In step S504, the third prediction module 123 may generate a third judgment result according to K probability vector sets, wherein each of the K probability vector sets may include L probability vectors corresponding to the L leads respectively. The K probability vector sets may be outputs of the K third stage models in step S503 respectively.

[0094] Figure 6 According to another embodiment of the present invention, a flow chart of a method for predicting blockage of a coronary artery is shown, wherein the method may be performed as follows: Figure 1 The electronic device 100 is shown as an implementation.

[0095] In step S601, a plurality of electrocardiogram data corresponding to a set of coronary arteries are obtained. In step S602, a plurality of first probabilities corresponding to the plurality of electrocardiogram data are generated according to the plurality of electrocardiogram data and at least one first stage model, a first judgment result is generated according to the plurality of first probabilities, and a first data subset corresponding to the first probability subset is selected from the plurality of electrocardiogram data in response to each of the first probability subsets of the plurality of first probabilities being greater than a first threshold, wherein the first judgment result indicates the probability of blockage of the set of coronary arteries. In step S603, a plurality of second probabilities corresponding to the first data subset are generated according to the first data subset and at least one second stage model, and a second judgment result is generated according to the plurality of second probabilities, wherein the second judgment result indicates the probability of blockage of a first coronary artery in the set of coronary arteries. In step S604, the first judgment result and the second judgment result are output.

[0096] In summary, the present invention can use one or more machine learning algorithms to determine whether a coronary artery set including the left main coronary artery, the anterior descending branch of the left coronary artery, the circumflex branch of the left coronary artery, and the right coronary artery is at risk of blockage. If the coronary artery set is at risk of blockage, the present invention can further use one or more machine learning algorithms to predict which of the coronary arteries in the coronary artery set is the blocked coronary artery. After confirming the type of blocked coronary artery, the present invention can further use one or more machine learning algorithms to predict the exact location of the blockage in the coronary artery.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An electronic device for predicting coronary artery blockage, characterized in that: include: a transceiver that receives a plurality of electrocardiogram data corresponding to a set of coronary arteries; A storage medium for storing multiple modules; as well as A processor is coupled to the storage medium and the transceiver, and accesses and executes the plurality of modules, wherein the plurality of modules include: A first prediction module generates at least one first probability vector corresponding to the at least one first stage model according to the plurality of electrocardiogram data and at least one first stage model, generates a first judgment result according to the at least one first probability vector, and selects a first data subset corresponding to the first subset from the plurality of electrocardiogram data in response to each of a first subset of the at least one first probability vector being greater than a first threshold, wherein the first judgment result indicates the probability of blockage of the coronary artery set, wherein the first judgment result is a probability calculated according to formula (1) and formula (2), wherein the formula (1) and the formula (2) are as follows: wherein p1(I,J) is the I-th first probability in the J-th first probability vector in the N first probability vectors, w1(J) is the weight corresponding to the J-th first stage model in the N first stage models, N is the number of the at least one first stage model, and X is the number of first probabilities in the first probability vector, wherein the first prediction module determines the weight corresponding to the first stage model based on the performance indicator of the first stage model; a second prediction module, generating at least one second probability vector corresponding to the at least one second stage model according to the first data subset and the at least one second stage model, and generating a second judgment result according to the at least one second probability vector, wherein the second judgment result indicates a probability of blockage of a first coronary artery in the coronary artery set; and An output module outputs the first judgment result and the second judgment result through the transceiver.

2. The electronic device of claim 1 , wherein the second prediction module selects a second data subset corresponding to the second subset from the first data subset in response to each of the second subsets of the at least one second probability vector being greater than a second threshold, wherein the plurality of modules further comprises: a third prediction module, generating at least one third probability vector corresponding to the at least one third stage model according to the second data subset and the at least one third stage model, and generating a third judgment result according to the at least one third probability vector, wherein the third judgment result indicates the probability of blockage occurring at the first position of the first coronary artery, wherein The output module outputs the third determination result through the transceiver.

3. The electronic device according to claim 2, wherein The transceiver receives a plurality of second electrocardiogram data corresponding to the coronary artery set, wherein the plurality of second electrocardiogram data corresponds to a first lead, wherein the plurality of second electrocardiogram data corresponds to a second lead, wherein The first prediction module generates at least one fourth probability vector corresponding to the at least one first stage model according to the plurality of second electrocardiogram data and the at least one first stage model, and generates the first judgment result according to the at least one first probability vector and the at least one fourth probability vector.

4. The electronic device according to claim 3, wherein The first prediction module selects a fourth data subset corresponding to the fourth subset from the plurality of second electrocardiogram data in response to each of the fourth subsets of the at least one fourth probability vector being greater than the first threshold, wherein The second prediction module generates at least one fifth probability vector corresponding to the at least one second stage model according to the fourth data subset and the at least one second stage model, and generates the second judgment result according to the at least one second probability vector and the at least one fifth probability vector.

5. The electronic device according to claim 4, wherein The second prediction module selects a fifth data subset corresponding to the fifth subset from the fourth data subset in response to each of the fifth subsets of the at least one fifth probability vector being greater than the second threshold, wherein The third prediction module generates at least one sixth probability vector corresponding to the at least one third stage model according to the fifth data subset and the at least one third stage model, and generates the third judgment result according to the at least one third probability vector and the at least one sixth probability vector.

6. The electronic device according to claim 1, wherein The first prediction module performs baseline drift removal, denoising and wavelet transformation on a plurality of training data to generate a plurality of first corrected training data, and trains the at least one first stage model according to the plurality of first corrected training data.

7. The electronic device of claim 6, wherein the first prediction module trains the at least one first-stage model according to at least one feature, wherein the at least one feature is associated with at least one of the following: P wave, Q wave, R wave, S wave and T wave.

8. The electronic device according to claim 2, wherein the at least one first stage model, the at least one second stage model, and the at least one third stage model correspond to a machine learning algorithm.

9. The electronic device of claim 1, wherein the first coronary artery comprises one of the following: The left main coronary artery, the anterior descending branch of the left coronary artery, the circumflex branch of the left coronary artery, and the right coronary artery.

10. A method for predicting coronary artery obstruction, characterized in that: include: Acquiring a plurality of electrocardiogram data corresponding to a set of coronary arteries; A plurality of first probabilities corresponding to the plurality of electrocardiogram data are generated according to the plurality of electrocardiogram data and at least one first stage model, a first judgment result is generated according to the plurality of first probabilities, and in response to each of the first probability subsets of the plurality of first probabilities being greater than a first threshold, a first data subset corresponding to the first probability subset is selected from the plurality of electrocardiogram data, wherein the first judgment result indicates the probability of blockage of the coronary artery set, wherein the first judgment result is a probability calculated according to formula (1) and formula (2), wherein the formula (1) and the formula (2) are as follows: wherein p1(I,J) is the I-th first probability in the J-th first probability vector in the N first probability vectors, w1(J) is the weight corresponding to the J-th first stage model in the N first stage models, N is the number of the at least one first stage model, and X is the number of first probabilities in the first probability vector, wherein the first prediction module determines the weight corresponding to the first stage model based on the performance indicator of the first stage model; generating a plurality of second probabilities corresponding to the first data subset according to the first data subset and at least one second stage model, and generating a second determination result according to the plurality of second probabilities, wherein the second determination result indicates a probability of blockage of a first coronary artery in the coronary artery set; as well as The first judgment result and the second judgment result are output.

Images