Method, program, and device for generating long-term heart rate variability on basis of cardiac signal measured in short term

Abstract

One embodiment of the present disclosure discloses a method, a program, and a device for generating long-term heart rate variability based on a short-term measured cardiac signal performed by an arithmetic device. The method may include the steps of: acquiring cardiac data including cardiac signals measured within a first reference time; and generating heart rate variability at a second reference time greater than the first reference time on the basis of the heart data by using a pre-learned deep learning model.

Description

Technical Field

[0001] The present disclosure discloses an artificial intelligence technology in the medical field, specifically, a method that can use artificial intelligence to generate heart rate variability from short-term measured heart signals as if extracted from long-term measured heart signals and extract features from the generated heart rate variability. Background Art

[0002] The heart does not beat in a constant manner. The intervals between heartbeats can lengthen or shorten. When a person is healthy, the intervals between heartbeats are usually large and complex. On the other hand, when a person is sick or under stress, the complexity of the intervals between heartbeats is usually significantly reduced. This is because large changes in heartbeats mean that the person has excellent adaptability to changes in the body or outside the body, while small changes in heartbeats mean that the person has poor adaptability to the body or outside environment.

[0003] As mentioned above, the state of the body can be known from the changing pattern of the heartbeat, and the concept derived from this is heart rate variability (HRV). Heart rate variability refers to the pattern of changes in the intervals between heartbeats measured in sequence and expressed in numbers. In other words, heart rate variability is a collection of multiple values ​​that measure the intervals between heartbeats, such as the interval between the first beat and the second beat, and the interval between the second beat and the third beat, and express their patterns in numbers.

[0004] Heart rate variability usually uses 24-hour analysis and 5-minute analysis as the gold standard. Several papers have studied the suitability of using ultra-short-term (less than 5 minutes) heart rate variability, and the results show that the shorter the time, the greater the difference from the gold standard. In particular, there are large differences in the characteristics of heart rate variability related to frequency. Therefore, in order to analyze heart rate variability, it is usually recommended to measure the heart signal for more than 5 minutes.

[0005] However, it is not easy for users to maintain a state of concentration in order to measure heart signals for 5 minutes. Therefore, noise caused by their movements is easily added during the 5-minute measurement of heart signals, and in particular, a large amount of noise is inevitably added during the 24-hour measurement. This noise reduces the reliability of the results in the heart rate variability analysis. Therefore, if performance comparable to the best standard can be obtained through short-term measurement, not only can the convenience of users be greatly improved, but also the reliability of heart rate variability analysis can be improved by limiting noise. Summary of the invention

[0006] Technical issues

[0007] The purpose of the present disclosure is to provide a method that can use artificial intelligence to generate heart rate variability from a short-term measured heart signal as if it were extracted from a long-term measured heart signal and extract features from the generated heart rate variability.

[0008] However, the technical problems to be solved by the present disclosure are not limited to the aforementioned problems, and other problems not mentioned above can be clearly understood from the following records.

[0009] Technical Solution

[0010] According to an embodiment of the present disclosure for achieving the above-mentioned technical problem, a method for generating long-term heart rate variability based on a short-term measured heart signal, which is executed by a computing device, is provided. The method may include the following steps: acquiring heart data including a heart signal measured within a first reference time; and generating heart rate variability of a second reference time greater than the first reference time based on the heart data using a pre-learned deep learning model.

[0011] Alternatively, the cardiac data may further include at least one of biological information, disease information, or physical activity information of a person on whom the cardiac signal is measured.

[0012] Alternatively, using a pre-learned deep learning model, the step of generating a heart rate variability of a second reference time greater than the first reference time based on the heart data may include the following steps: inputting the heart data into a pre-learned first model to calculate the mean and standard deviation of the heart rate variability of the second reference time; predicting the probability distribution of the heart rate variability of the second reference time based on the mean and standard deviation; and sampling based on the predicted probability distribution to calculate multiple values ​​included in the heart rate variability of the second reference time.

[0013] Alternatively, the step of calculating multiple values ​​included in the heart rate variability of the second reference time by sampling based on the predicted probability distribution may include the following steps: when the predicted probability distribution corresponds to a normal distribution, performing random sampling on the predicted probability distribution to calculate multiple values ​​included in the heart rate variability of the second reference time.

[0014] Alternatively, the step of calculating the multiple values ​​included in the heart rate variability of the second reference time by sampling based on the predicted probability distribution may include the following steps: when the predicted probability distribution does not correspond to the normal distribution, performing modeling of the predicted probability distribution; and performing random sampling of the probability distribution generated by the modeling to calculate the multiple values ​​included in the heart rate variability of the second reference time.

[0015] Alternatively, the step of calculating multiple values ​​included in the heart rate variability of the second reference time by sampling based on the predicted probability distribution may include the following steps: when the predicted probability distribution corresponds to a normal distribution, performing random sampling on the predicted probability distribution to calculate multiple values ​​included in the heart rate variability of the second reference time.

[0016] Alternatively, the step of calculating the multiple values ​​included in the heart rate variability of the second reference time by sampling based on the predicted probability distribution may include the following steps: when the predicted probability distribution does not correspond to the normal distribution, performing modeling of the predicted probability distribution; and performing random sampling of the probability distribution generated by the modeling to calculate the multiple values ​​included in the heart rate variability of the second reference time.

[0017] Alternatively, modeling the predicted probability distribution may include a Gaussian mixture model or a normalizing flow.

[0018] Alternatively, the step of calculating the multiple values ​​included in the heart rate variability of the second reference time by sampling based on the predicted probability distribution may include the following steps: determining whether the predicted probability distribution corresponds to a normal distribution; based on a result of the determination, selecting one of multiple sampling methods for extracting the heart rate variability of the second reference time; and sampling the probability distribution using the selected sampling method to calculate the multiple values ​​included in the heart rate variability of the second reference time.

[0019] Alternatively, the step of generating a second reference time heart rate variability greater than the first reference time based on the cardiac data using a pre-learned deep learning model may also include the following steps: using a pre-learned second model based on a self-attention mechanism to rearrange the order of multiple values ​​included in the calculated second reference time heart rate variability.

[0020] Alternatively, using a pre-learned second model based on a self-attention mechanism, the step of rearranging the order of multiple values ​​included in the calculated heart rate variability of the second reference time may include the following steps: inputting the heart rate variability of the second reference time into the second model to generate an attention map showing the correlation between the multiple values ​​included in the heart rate variability; and rearranging the order of the multiple values ​​included in the heart rate variability based on the attention map.

[0021] Alternatively, using a pre-learned deep learning model, the step of generating a heart rate variability of a second reference time greater than the first reference time based on the cardiac data may include the following steps: inputting the cardiac data into a pre-learned third model as a generative model, and calculating multiple values ​​included in the heart rate variability of the second reference time.

[0022] Alternatively, using a pre-learned deep learning model, the step of generating a heart rate variability of a second reference time greater than the first reference time based on the cardiac data may include the following steps: inputting the cardiac data into a pre-learned fourth model as a sequence-to-sequence model, and calculating multiple values ​​included in the heart rate variability of the second reference time.

[0023] According to an embodiment of the present disclosure for achieving the above-mentioned technical problem, a computer program stored in a computer-readable storage medium is provided. When the computer program is executed on one or more processors, multiple actions are performed for generating long-term heart rate variability based on short-term measured heart signals. At this time, the multiple actions may include the following actions: obtaining heart data including heart signals measured within a first reference time; and using a pre-learned deep learning model, generating heart rate variability of a second reference time greater than the first reference time based on the heart data.

[0024] According to an embodiment of the present disclosure for achieving the above-mentioned technical problem, a computing device for generating long-term heart rate variability based on a short-term measured heart signal is provided. The device may include: a processor including at least one core; a memory including a plurality of program codes executable by the processor; and a network unit for acquiring heart data including a heart signal measured within a first reference time. At this time, the processor can use a pre-learned deep learning model to generate a heart rate variability of a second reference time greater than the first reference time based on the heart data.

[0025] Effects of the Invention

[0026] According to the present disclosure, in a situation where long-term measurement of cardiac signals is required, the performance of long-term measurement can be exerted by only a short-term measurement, thereby improving the convenience of the user.

[0027] Moreover, heart rate variability can be generated by measuring only a short time, which greatly reduces the probability of noise in the heart signal measurement process.

[0028] At the same time, heart rate variability as extracted from a long-term measured heart signal can be generated from a short-term measured heart signal, thereby realizing a system having versatility that can be directly applied to various diagnostic algorithms using heart rate variability. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a block diagram of a computing device according to an embodiment of the present disclosure.

[0030] Figure 2 2 is a conceptual diagram showing the heart rate variability of a heart signal measured within a first reference time and a second reference time according to an embodiment of the present disclosure.

[0031] Figure 3 4 is a block diagram showing a process of generating heart rate variability at a second reference time based on heart data measured within a first reference time according to an embodiment of the present disclosure.

[0032] Figure 4 is a block diagram illustrating a process of generating heart rate variability at a second reference time using a deep learning model according to an embodiment of the present disclosure.

[0033] Figure 5 is a block diagram illustrating a process of extracting heart rate variability features of a second reference time using a deep learning model according to an embodiment of the present disclosure.

[0034] Figure 6 Detailed description is a flow chart showing a method for generating long-term heart rate variability based on short-term measured heart signals according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] The following is a detailed description of the present disclosure in conjunction with the accompanying drawings in order to enable a person of ordinary skill in the art to which the present disclosure belongs (hereinafter referred to as "a person of ordinary skill in the art") to easily implement the embodiments of the present disclosure. The embodiments disclosed in the present disclosure are used to enable a person of ordinary skill in the art to utilize or implement the contents of the present disclosure. Therefore, various variations of the embodiments of the present disclosure are obvious to a person of ordinary skill in the art. That is, the present disclosure can be implemented in various different forms and is not limited to the following embodiments.

[0036] The same or similar graphic symbols in the entire specification of the present disclosure represent the same or similar elements. In addition, the graphic symbols of the parts not related to the description of the present disclosure may be omitted in the drawings for the purpose of clearly explaining the present disclosure.

[0037] The term "or" used in the present disclosure does not mean an exclusive "or" but an inclusive "or". That is, if the present disclosure does not specify otherwise or its meaning is unclear in the context of the sentence, "X utilizes A or B" should be understood to mean one of the natural inclusive permutations. For example, if the present disclosure does not specify otherwise or its meaning is unclear in the context of the sentence, "X utilizes A or B" can be interpreted as one of the cases where X utilizes A or the case where X utilizes B or the case where X utilizes A and B.

[0038] The term "and / or" used in the present disclosure should be understood to refer to or include all possible combinations of more than one of the listed related concepts.

[0039] The terms "comprising" and / or "having" used in this disclosure should be understood to mean the presence of specific features and / or elements. However, the terms "comprising" and / or "having" should be understood to not exclude the presence or addition of one or more other features, other elements and / or combinations thereof.

[0040] If the disclosure does not specify or it is not clear from the context of the sentence that it indicates a singular form, the singular should generally be interpreted as including "one or more".

[0041] The term "Nth (N is a natural number)" used in the present disclosure can be understood as a way of expressing the elements of the present disclosure to be distinguished from each other according to a preset benchmark such as a functional viewpoint, a structural viewpoint, or for convenience of explanation. For example, elements that perform different functions in the present disclosure can be distinguished as first elements or second elements. However, elements that are essentially the same in the technical spirit of the present disclosure but need to be distinguished for the convenience of explanation can also be distinguished as first elements or second elements.

[0042] The term "acquire" used in the present disclosure may be understood to mean not only receiving data through a wired or wireless communication network with an external device or system, but also generating data in an on-device form.

[0043] On the other hand, the term "module" or "unit" used in the present disclosure can be understood as a term referring to an independent functional unit for processing computing resources such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, a combination of software and hardware, etc. In this case, a "module" or "unit" can be a unit composed of a single element, or a unit represented by a combination or set of multiple elements. For example, a "module" or "unit" as a protocol concept can refer to a hardware element or a set thereof of a computing device, an application program that executes a specific function of software, a process (procedure) implemented by executing software, or a set of instructions for executing a program, etc. Moreover, a "module" or "unit" as a broad concept can refer to the computing device itself that constitutes the system or an application program executed in the computing device, etc. However, the aforementioned concept is only an example, and the concept of "module" or "unit" can be variously defined within the scope that can be understood by ordinary technicians in the field based on the content of the present disclosure.

[0044] The term "model" used in this disclosure can be understood as a system implemented using mathematical concepts and language to solve a specific problem, a collection of software units used to solve a specific problem, or an abstract model of a process used to solve a specific problem. For example, a neural network "model" can refer to the entire system implemented by a neural network that has the ability to solve problems through learning. In this case, the neural network can optimize the parameters of the connecting nodes (nodes) or neurons (neurons) through learning and thus have the ability to solve problems. A neural network "model" may include a single neural network or a neural network collection composed of multiple neural networks.

[0045] The explanation of the above terms is to help understand the present disclosure. Therefore, unless the above terms are clearly described as matters that limit the content of the present disclosure, they are not used in the sense of limiting the technical concept of the present disclosure.

[0046] Figure 1 is a block diagram of a computing device according to an embodiment of the present disclosure.

[0047] The computing device 100 of one embodiment of the present disclosure may be a hardware device or a part of a hardware device that performs comprehensive data processing and computing, or may be a computing environment based on software connected by a communication network. For example, the computing device 100 may be a server that is a main body that performs intensive data processing functions and shares resources, or may be a client that shares resources through interaction with a server. Moreover, the computing device 100 may also be a cloud system that performs comprehensive data processing by means of interaction between multiple servers and multiple clients. The foregoing content is merely an example related to the types of computing devices 100, and the types of computing devices 100 may be constructed in various ways within the scope that can be understood by ordinary technicians in the field based on the content of the present disclosure.

[0048] See also Figure 1 , the computing device 100 of an embodiment of the present disclosure may include a processor 110, a memory 120 and a network unit 130. However, Figure 1 This is just an example, and the computing device 100 may include other elements for implementing a computing environment. Moreover, the computing device 100 may also include only a portion of the disclosed multiple elements.

[0049] The processor 110 of one embodiment of the present disclosure may be understood as a component unit including hardware and / or software for performing operations. For example, the processor 110 can read a computer program and then perform data processing for machine learning. The processor 110 can process operations such as processing input data for machine learning, extracting features for machine learning, and calculating errors based on backpropagation. The processor 110 for performing data processing as described above may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASICc), or a field programmable gate array (FPGA). The aforementioned types of processors 110 are merely examples, and the types of processors 110 can be configured in various ways within the scope that can be understood by ordinary technicians in the field based on the content of the present disclosure.

[0050] The processor 110 can use a pre-learned deep learning model to generate a heart rate variability of a second reference time greater than the first reference time based on the heart data including the heart signal measured within the first reference time. At this time, the heart signal can be understood as a signal showing the activity or state of the heart. Moreover, the reference time can be a value pre-set by the user after considering various factors such as the measurement environment and purpose of use of the heart signal. For example, the heart signal may include a 12-lead electrocardiogram, a dynamic (holter) electrocardiogram, or a photoplethysmography (PPG). The first reference time can be 10 seconds to 30 seconds, and the second reference time can be set to 5 minutes to 24 hours based on the best standard (gold-standard) for analyzing heart rate variability.

[0051] The processor 110 can calculate the characteristics of the heart rate variability from the heart rate variability of the second reference time generated by the deep learning model. For example, the processor 110 can calculate the characteristics of the heart rate variability such as the standard deviation of RR interval (SDRR), the mean heart rate (mean RR), the standard deviation of all NN intervals (SDNN), the root mean square of the successive differences (RMSSD), the low frequency power (LF) or the high frequency power (HF) by mathematical calculation of a plurality of values ​​included in the heart rate variability of the second reference time. In addition, the processor 110 can combine the calculated characteristics of the heart rate variability to derive values ​​required for diagnosis such as anxiety index, anxiety resistance, autonomic balance, autonomic activity, fatigue, depression or anxiety disorder.

[0052] The memory 120 of an embodiment of the present disclosure may be understood as a hardware and / or software component unit that stores and manages data processed by the computing device 100. That is, the memory 120 can store any form of data generated or determined by the processor 110 and any form of data received by the network unit 130. For example, the memory 120 may include at least one storage medium in the form of a flash memory type, a hard disk type, a multimedia card micro type, a card memory, a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. Moreover, the memory 120 may also include a database system that controls and manages data with a preset system. The aforementioned type of memory 120 is merely an example, and the type of memory 120 can be variously defined within a scope that can be understood by a person of ordinary skill in the art based on the contents of this disclosure.

[0053] The memory 120 can structure and organize the data and data combinations required by the processor 110 when performing operations, and multiple program codes that can be executed by the processor 110. For example, the memory 120 can store the heart data received through the network unit 130 described later. The memory 120 can store program codes that allow the neural network model to learn after receiving the input of heart data, program codes that allow the neural network model to infer according to the purpose of use of the computing device 100 after receiving the input of heart data, and processed data generated by executing the program codes.

[0054] The network unit 130 of an embodiment of the present disclosure may be understood as a component unit for transmitting and receiving data through any form of a known wired wireless communication system. For example, the network unit 130 may use a wired wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), fifth generation mobile communication (5G), ultra wide-band, ZigBee, radio frequency (RF), wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth to transmit and receive data. The aforementioned multiple communication systems are merely examples, and various wired wireless communication systems for transmitting and receiving data by the network unit 130 may be applied in addition to the aforementioned examples.

[0055] The network unit 130 can receive data required by the processor 110 for calculation through wired and wireless communication with any system or any client. In addition, the network unit 130 can send data generated by calculation by the processor 110 through wired and wireless communication with any system or any client. For example, the network unit 130 can receive cardiac data through communication with a database in a hospital environment, a cloud server that performs tasks such as medical data standardization, a client such as a smart watch, or a medical calculation device. The network unit 130 can send output data of the neural network model and intermediate data and processed data derived during the calculation of the processor 110 through communication with the aforementioned database, server, client, or calculation device.

[0056] Figure 2 2 is a conceptual diagram showing the heart rate variability of a heart signal measured within a first reference time and a second reference time according to an embodiment of the present disclosure.

[0057] The multiple numerical values ​​showing the intervals between peaks in the heart signal 10 measured within the first reference time represent multiple values ​​included in the heart rate variability within the first reference time. Figure 2As shown, the value 822 in the first peak interval 822ms represents the first value of the heart rate variability, and the value 857 in the second peak interval 857ms represents the second value of the heart rate variability. As described above, if the heart signal 10 lists a plurality of values ​​indicating the peak intervals during the measurement period within the first reference time, the heart rate variability within the first reference time can be calculated.

[0058] See also Figure 2 The heart rate variability 20 at the second reference time generated by the computing device 100 of one embodiment of the present disclosure from the heart signal 10 measured within the first reference time may include multiple values ​​included in the heart rate variability within the first reference time and multiple newly generated values ​​25. The multiple newly generated values ​​25 can be understood as values ​​calculated based on the prediction results, and the prediction is a prediction of what kind of characteristics and patterns the heart signal 10 measured within the first reference time will show if it is measured until the second reference time. That is, the computing device 100 can input the heart signal 10 measured within the first reference time into the deep learning model and generate multiple predicted heart rate variability values ​​above the first reference time and within the second reference time. Moreover, the computing device 100 can connect the multiple heart rate variability values ​​that can be calculated from the heart signal 10 measured within the first reference time and the multiple newly generated values ​​to generate the heart rate variability at the second reference time.

[0059] Figure 3 4 is a block diagram showing a process of generating heart rate variability at a second reference time based on heart data measured within a first reference time according to an embodiment of the present disclosure.

[0060] See also Figure 3 The cardiac data 30 of an embodiment of the present disclosure may include cardiac signals such as electrocardiograms measured within the first reference time. The cardiac data 30 may also include at least one of biological information, disease information, or physical activity information of the person whose cardiac signal is measured. The biological information may include the age, gender, height, and weight of the person whose cardiac signal is measured. The disease information may include whether the person is ill or not and the details of the disease at the time of measurement. The physical activity information may include the physical condition and activity status at the time of measurement.

[0061] The computing device 100 of an embodiment of the present disclosure can calculate the long-term heart rate variability 35 corresponding to the heart rate variability of the second reference time after inputting the heart data 30 into the deep learning model 200. The deep learning model 200 may include the following models: a first model 210, which predicts the probability distribution of heart rate variability and collects heart rate variability in the predicted distribution; a second model 215, which refines the heart rate variability collected by the first model 210; a third model 220, which corresponds to a generative model; and a fourth model 230, which corresponds to a sequence-to-sequence model. The computing device 100 can calculate the long-term heart rate variability 35 by selectively or compositely using the first model 210 to the fourth model 230 according to factors such as the type of heart signal contained in the heart data 30 or the user's intention.

[0062] For reference, the generation process of the long-term heart rate variability 35 using the first model 210 and the second model 215 will be described later in conjunction with Figure 4 and Figure 5 Specific instructions.

[0063] The third model 220 can learn the distribution of heart rate variability within the first reference time from the heart data 30 and generate the long-term heart rate variability 35 equivalent to the heart rate variability of the second reference time from the learned distribution. That is, the third model 220 can grasp the distribution of heart rate variability calculated from the heart data 30 including the signal actually measured within the first reference time and newly generate the following long-term heart rate variability 35, which is similar to the heart rate variability calculated from the signal actually measured until the second reference time. As an example, the third model 220 may include a generative neural network such as a generative adversarial neural network (GAN) or a variational auto encoder (VAE). However, the type of the neural network of the third model 220 is only an example, and the neural network of the third model 220 can also be applied to various generative neural networks in addition to the example.

[0064] The fourth model 230 can learn the characteristics of the heart data 30 that change over time and generate the long-term heart rate variability 35 equivalent to the heart rate variability of the second reference time based on the learned results. That is, the fourth model 230 can predict what kind of changes the heart data 30 will show after the first reference time and generate the long-term heart rate variability 35 of the second reference time. As an example, the fourth model 230 may include a sequence-to-sequence model such as a recurrent neural network (RNN) and a long short-term memory (LSTM). However, the type of neural network of the fourth model 230 is only an example, and the neural network of the fourth model 230 can also be applied to various generative neural networks in addition to the example.

[0065] The computing device 100 can calculate the heart rate variability feature 39 from the long-term heart rate variability 35 corresponding to the heart rate variability at the second reference time. At this time, the heart rate variability feature 39 may include SDRR, mean RR, SDNN, RMSSD, LF, or HF, etc. The computing device 100 does not directly extract the heart rate variability feature 39 from the heart data 30 that does not meet the optimal standard, but generates the long-term heart rate variability 35 that meets the optimal standard and then extracts the heart rate variability feature 39 therefrom, so various features can be easily extracted.

[0066] Figure 4 is a block diagram illustrating a process of generating heart rate variability at a second reference time using a deep learning model according to an embodiment of the present disclosure. Figure 5 is a block diagram illustrating a process of extracting heart rate variability features of a second reference time using a deep learning model according to an embodiment of the present disclosure.

[0067] See also Figure 4 , the computing device 100 can input the cardiac data 40 including the cardiac signal measured within the first reference time into the first model 210, and then calculate the average 43 and the standard deviation 45 of the long-term heart rate variability corresponding to the second reference time. At this time, the first model 210 can be a model learned based on supervised learning. In addition to learning based on supervised learning, the first model 210 can also be learned based on the following learning method, which is a learning method that can be applied based on the content disclosed in this disclosure.

[0068] The computing device 100 can predict the probability distribution 47 of the long-term heart rate variability using the mean 43 and the standard deviation 45 of the long-term heart rate variability. Furthermore, the computing device 100 can collect a plurality of long-term heart rate variability values ​​49 from the predicted probability distribution 47. At this time, collection can be understood as a process of extracting required values ​​from the probability distribution. For example, when the predicted probability distribution 47 corresponds to a normal distribution, the computing device 100 can perform random sampling on the predicted probability distribution 47 to calculate a plurality of long-term heart rate variability values ​​49. That is, when the predicted probability distribution 47 corresponds to a normal distribution, the computing device 100 can generate a plurality of long-term heart rate variability values ​​49 by randomly extracting from a plurality of values ​​existing on the predicted probability distribution 47 until the second reference time is satisfied. On the contrary, when the predicted probability distribution 47 does not correspond to a normal distribution, the computing device 100 can perform modeling for the predicted probability distribution 47. At this time, the modeling of the probability distribution 47 is an operation of deforming the probability distribution 47, and may include modeling such as a Gaussian mixture model or a normalizing flow. That is, even if the predicted probability distribution 47 does not follow the normal distribution, the operation device 100 can perform the processing operation on the probability distribution 47 so that the required value can be easily extracted from the probability distribution 47. In addition, the operation device 100 can perform random sampling on the modeled probability distribution to calculate a plurality of long-term heart rate variability values ​​49.

[0069] On the other hand, in addition to the random sampling, the computing device 100 can also use multiple sampling methods in other ways. For example, first, the computing device 100 can determine whether the predicted probability distribution 47 corresponds to the normal distribution. When the predicted probability distribution 47 corresponds to the normal distribution, the computing device 100 can perform random sampling to extract multiple samples existing on the normal distribution and generate multiple long-term heart rate variability values ​​49. When the predicted probability distribution 47 does not correspond to the normal distribution, the probability of the error in the result of random sampling will increase. Therefore, the computing device 100 can select other sampling methods to perform sampling on the probability distribution 47 predicted and extract multiple long-term heart rate variability values ​​49. The other sampling methods include sampling methods such as clustering methods based on the distance of multiple samples. That is, the computing device 100 can also select one of the multiple sampling methods for sampling according to whether the predicted probability distribution 47 corresponds to the normal distribution.

[0070] See also Figure 5 , the computing device 100 can use the second model 220 based on the self-attention mechanism to rearrange the order of the plurality of long-term heart rate variability values ​​50. The plurality of long-term heart rate variability values ​​50 are as follows: Figure 4As shown, the heart data is generated after sampling the probability distribution predicted by the first model 210, and therefore does not have the order information of the heart data original input to the first model 210. This situation does not constitute a major problem when calculating multiple heart rate variability features (such as SDNN, etc.) whose order information is not important, but it becomes a problem when calculating multiple heart rate variability features (such as RMSSD, etc.) whose order information is important. Therefore, in order to add order information to the multiple long-term heart rate variability values ​​50 generated after sampling the probability distribution predicted by the first model 210, the computing device 100 can reconfigure the order of the multiple long-term heart rate variability values ​​50 after inputting the multiple long-term heart rate variability values ​​50 to the second model 220 based on the self-attention mechanism.

[0071] Specifically, the computing device 100 can generate an attention map 55 showing the correlation between the multiple long-term heart rate variability values ​​50 after inputting the multiple long-term heart rate variability values ​​50 into the second model. The attention map can express how highly the other multiple long-term heart rate variability values ​​are related to each other based on a specific long-term heart rate variability value. At this time, a higher correlation can mean that the order is closer. For example, assuming that the value most related to the long-term heart rate variability value 800 corresponding to the first value is 890, the area where 800 and 890 are matched in the attention map can be represented by the darkest color. The mark on the attention map can be explained as follows, that is, it is better to configure 890 after 800 in the long-term heart rate variability. That is, the computing device 100 can grasp the order of the multiple long-term heart rate variability values ​​50 through the attention map and re-arrange them, and can generate long-term heart rate variability values ​​59 with aligned order.

[0072] Figure 6 Detailed description is a flow chart showing a method for generating long-term heart rate variability based on short-term measured heart signals according to an embodiment of the present disclosure.

[0073] See also Figure 6 , the computing device 100 of an embodiment of the present disclosure can obtain cardiac data including cardiac signals measured within a first reference time (step S210). For example, if the computing device 100 is not separately equipped with a measuring unit, the computing device 100 can obtain cardiac data including cardiac signals measured within the first reference time through wired and wireless communication with a device that measures cardiac signals. If the computing device 100 is separately equipped with a measuring unit, the computing device 100 can measure cardiac signals within the first reference time and generate cardiac data through the measuring unit. On the other hand, in addition to the cardiac signals measured within the first reference time, the cardiac data can further include at least one of biological information, disease information, or physical activity information of a person stored when measuring the cardiac signal.

[0074] The computing device 100 can use the pre-learned deep learning model to generate the heart rate variability of the second reference time greater than the first reference time based on the heart data (step S220). Specifically, the computing device 100 can input the heart data into the pre-learned first model and calculate the average and standard deviation of the heart rate variability of the second reference time. The computing device 100 can predict the probability distribution of the heart rate variability of the second reference time based on the average and standard deviation. The computing device 100 can sample based on the predicted probability distribution and calculate the multiple values ​​included in the heart rate variability of the second reference time. Moreover, the computing device 100 can use the pre-learned second model based on the self-attention mechanism to rearrange the order of the multiple values ​​included in the calculated heart rate variability of the second reference time. Moreover, the computing device 100 can also input the heart data into the pre-learned third model as a generative model and calculate the multiple values ​​included in the heart rate variability of the second reference time. Furthermore, the calculation device 100 can also input the heart data into a fourth model that is a sequence-to-sequence model learned in advance and calculate a plurality of values ​​included in the heart rate variability at the second reference time.

[0075] Although Figure 6 Although not shown in the figure, the operation device 100 can extract the characteristics of the heart rate variability at the second reference time based on the multiple values ​​included in the heart rate variability at the second reference time calculated in step S210.

[0076] The various embodiments of the present disclosure described above can be combined with other different embodiments and can be changed within the scope of understanding of ordinary technicians in the field according to the detailed description of the above text. The embodiments of the present disclosure are illustrative in all aspects and should be interpreted as not restrictive. For example, each element described in a single overall form can be implemented in a dispersed form, and similarly, each element described in a dispersed form can also be implemented in a combined form. Therefore, the meaning and scope of the claims of the present disclosure and all modifications and deformations derived from their equivalent concepts should be interpreted as belonging to the scope of the present disclosure.

Claims

1. A method for generating long-term heart rate variability based on a short-term measured heart signal, performed by a computing device comprising at least one processor, characterized in that: The following steps are involved: Acquiring cardiac data including cardiac signals measured within a first reference time; and A heart rate variability of a second reference time greater than the first reference time is generated based on the heart data using a pre-learned deep learning model.

2. The method according to claim 1, characterized in that The cardiac data also includes at least one of biological information, disease information, or physical activity information of a person on whom the cardiac signal is measured.

3. The method according to claim 1, characterized in that The step of generating a heart rate variability of a second reference time greater than the first reference time based on the heart data by using a pre-learned deep learning model comprises the following steps: Input the heart data into a pre-learned first model to calculate the mean and standard deviation of the heart rate variability at the second reference time; Predicting a probability distribution of heart rate variability at the second reference time based on the mean and standard deviation; and A plurality of values ​​included in the heart rate variability at the second reference time are calculated by sampling based on the predicted probability distribution.

4. The method according to claim 3, characterized in that The step of calculating a plurality of values ​​included in the heart rate variability at the second reference time by sampling based on the predicted probability distribution comprises the following steps: When the predicted probability distribution corresponds to a normal distribution, random sampling is performed on the predicted probability distribution to calculate a plurality of values ​​included in the heart rate variability at the second reference time.

5. The method according to claim 3, characterized in that: The step of calculating a plurality of values ​​included in the heart rate variability at the second reference time by sampling based on the predicted probability distribution comprises the following steps: When the predicted probability distribution does not correspond to a normal distribution, performing modeling of the predicted probability distribution; as well as Random sampling is performed on the probability distribution generated by the modeling to calculate a plurality of values ​​included in the heart rate variability at the second reference time.

6. The method according to claim 5, characterized in that Modeling of the predicted probability distribution includes Gaussian mixture models or normalized flows.

7. The method according to claim 3, characterized in that The step of calculating a plurality of values ​​included in the heart rate variability at the second reference time by sampling based on the predicted probability distribution comprises the following steps: Determining whether the predicted probability distribution corresponds to a normal distribution; Based on the result of the determination, selecting one of a plurality of sampling methods for extracting the heart rate variability of the second reference time; as well as The probability distribution is sampled using the selected sampling method, and a plurality of values ​​included in the heart rate variability at the second reference time are calculated.

8. The method according to claim 3, characterized in that The step of generating a heart rate variability of a second reference time greater than the first reference time based on the heart data by using a pre-learned deep learning model further includes the following steps: The order of the plurality of values ​​included in the calculated heart rate variability of the second reference time is rearranged using the pre-learned second model based on the self-attention mechanism.

9. The method according to claim 8, characterized in that The step of rearranging the order of the plurality of values ​​included in the calculated heart rate variability of the second reference time by using the pre-learned second model based on the self-attention mechanism comprises the following steps: inputting the heart rate variability of the second reference time into the second model, and generating an attention map showing a correlation relationship between a plurality of values ​​included in the heart rate variability; as well as Based on the attention map, the order of the multiple values ​​included in the heart rate variability is rearranged.

10. The method according to claim 1, characterized in that The step of generating a heart rate variability of a second reference time greater than the first reference time based on the heart data by using a pre-learned deep learning model comprises the following steps: The heart data is input into a third model that is a generative model learned in advance, and a plurality of values ​​included in the heart rate variability at the second reference time are calculated.

11. The method according to claim 1, characterized in that: The step of generating a heart rate variability of a second reference time greater than the first reference time based on the heart data by using a pre-learned deep learning model comprises the following steps: The cardiac data is input into a fourth model that is a sequence-to-sequence model learned in advance, and a plurality of values ​​included in the heart rate variability at the second reference time are calculated.

12. A computer program stored in a computer-readable storage medium, characterized in that The computer program, when executed on one or more processors, performs a plurality of actions for generating long-term heart rate variability based on short-term measured heart signals, The multiple actions include the following actions: Acquiring cardiac data including cardiac signals measured within a first reference time; and A heart rate variability of a second reference time greater than the first reference time is generated based on the heart data using a pre-learned deep learning model.

13. A computing device for generating long-term heart rate variability based on a short-term measured heart signal, characterized in that: include: a processor, comprising at least one core; a memory including a plurality of program codes executable by the processor; and a network unit, configured to obtain cardiac data including a cardiac signal measured within a first reference time, The processor generates heart rate variability of a second reference time greater than the first reference time based on the heart data using a pre-learned deep learning model.

Images