Device for non-invasively calculating cardiac output and operating method therefor

Abstract

The present invention relates to a device for non-invasively calculating a cardiac output and an operation method therefor. The method comprises the steps of: receiving first bio-signal data including an electrocardiogram and a photoplethysmogram of a user; and determining a stroke volume of the user on the basis of the first bio-signal data of the user by using a deep learning model configured to output the stroke volume, with the electrocardiogram and the photoplethysmogram used as inputs.

Description

Device for calculating cardiac output noninvasively and method of operation thereof

[0001] The present invention relates to a device for noninvasively calculating cardiac output and an operating method thereof.

[0002] Cardiac output (CO) monitoring is used for hemodynamic assessment in high-risk patients. Cardiac output is the product of heart rate (HR) and stroke volume (SV). It represents the amount of blood pumped by the heart's two ventricles per unit of time (usually minutes). It is an indicator of not only cardiac function but also the overall circulatory system.

[0003] Predicting cardiac output is important because it is used to monitor and optimize systemic oxygen and drug delivery in high-risk surgical patients.

[0004] The current gold-standard (test for diagnosing the progression of the disease) methodology, the thermodilution-based method, involves inserting a pulmonary artery catheter to calculate cardiac output, which is invasive and carries the risk of complications such as infection.

[0005] The minimally invasive cardiac output calculation method (arterial pressure-based method) proposed after the thermodilution method is a method that calculates cardiac output using minimally invasive arterial blood pressure waveforms. However, this method has limitations in its usability because it is not a technology used in all patients in intensive care units and operating rooms in hospitals.

[0006] The background technology of the invention has been prepared to facilitate a better understanding of the present invention. It should not be construed as an admission that the matters described in the background technology of the invention constitute prior art.

[0007] In response to the problems of the above-described invasive or minimally invasive cardiac output calculation methods, a new method is required that can analyze and calculate cardiac output using only electrocardiogram and photoplethysmography, which are completely non-invasive biosignals.

[0008] As a result, the inventors of the present invention sought to develop a method for analyzing noninvasive bio-signal data collected by being mounted on a bio-signal measuring device such as a patient monitor by analyzing the patient's bio-signal and utilizing secondary indicators, and a device for performing the same.

[0009] The tasks of the present invention are not limited to the tasks mentioned above, and other tasks not mentioned will be clearly understood by those skilled in the art from the description below.

[0010] In order to solve the above-described problem, an operating method of a device for noninvasively calculating cardiac output according to an embodiment of the present invention is provided. The method is configured to include the steps of receiving first bio-signal data including an electrocardiogram and a photoplethysmogram of a user, and determining a stroke volume of the user based on the first bio-signal data of the user using a deep learning model configured to output a stroke volume by taking the electrocardiogram and the photoplethysmogram as inputs.

[0011] According to a feature of the present invention, the deep learning model is trained by training data for a plurality of patients, and the training data for the plurality of patients may include a VPG (first derivative PPG) waveform, which is a first differential waveform of the photoplethysmogram, and an APG (second derivative PPG) waveform, which is a second differential waveform of the photoplethysmogram.

[0012] According to another feature of the present invention, the learning data for the plurality of patients may include a PR interval, QRS duration, QRS amplitude, and ST segment extracted from the electrocardiogram of each of the plurality of patients.

[0013] According to another feature of the present invention, the method may further include a step of receiving demographic information data of the user. The deep learning model is configured to determine stroke volume of the user by using the first bio-signal data, the second bio-signal data, and the demographic information data of the user as inputs, wherein the demographic information data includes at least one of the user's age, sex, height, and weight, and the second bio-signal data may further include the VPG waveform, the APG waveform, the PR interval, the QRS duration, the QRS amplitude, the ST segment, and the PTT.

[0014] In order to solve the aforementioned problem, a device for noninvasively calculating cardiac output according to one embodiment of the present invention is provided.

[0015] The device comprises a memory, a transceiver for receiving at least one of first bio-signal data including an electrocardiogram and a photoplethysmogram, and demographic information data, and a processor including a deep learning model, wherein the processor is configured to preprocess the first bio-signal data, extract second bio-signal data from the preprocessed bio-signal data, and infer stroke volume using a deep learning model configured to output stroke volume using the first bio-signal data and the second bio-signal data as inputs.

[0016] According to a feature of the present invention, the processor may be further configured to slice the first biosignal data into a predetermined length, extract sampling data from the sliced ​​first biosignal data, and remove noise data of the first biosignal data.

[0017] According to another feature of the present invention, the second biosignal data may include at least one of a VPG (first derivative PPG) waveform which is a first derivative waveform of the photoplethysmogram, an APG (second derivative PPG) waveform which is a second derivative waveform of the photoplethysmogram, a PR interval, a QRS duration, a QRS amplitude, and an ST segment extracted from the electrocardiogram, and a PTT (Pulse Transit Time) extracted from the electrocardiogram and the photoplethysmogram.

[0018] According to another feature of the present invention, the processor is further configured to perform minimum-maximum normalization of the photoplethysmogram and extract the VPG waveform and the APG waveform of the photoplethysmogram, and the VPG waveform and the APG waveform can be input to the deep learning model in the form of time series data having the predetermined length.

[0019] According to another feature of the present invention, the PR interval, the QRS duration, the QRS amplitude, the ST segment, and the PTT can be input to the deep learning model as an average of respective values ​​measured multiple times during a time corresponding to the predetermined length.

[0020] According to another feature of the present invention, the learning data includes the first biosignal data, the second biosignal data, stroke volume data, and the demographic information data of a plurality of patients, and the demographic information data may include at least one of age, gender, height, and weight.

[0021] According to another feature of the present invention, the memory may be configured to store a set of the learning data matched for each of the plurality of patients.

[0022] According to another feature of the present invention, the deep learning model may have a CNN (convolution neural network) structure including at least one SE (Squeeze and Excitation) block and at least one Residual block.

[0023] According to another feature of the present invention, the transceiver may further be configured to receive the first bio-signal data and the second bio-signal data of the user corresponding to a period from t (where t is a positive integer) seconds to (t+n) seconds, and the processor may further be configured to calculate the stroke volume of the user at (t+n) seconds using the deep learning model.

[0024] Specific details of other embodiments are included in the description and drawings of the invention.

[0025] The present invention can calculate a patient's cardiac output using biosignal data collected by a non-invasive method.

[0026] Specifically, the present invention can calculate a patient's cardiac output using only electrocardiogram and photoplethysmogram, which are completely non-invasive biosignals in which the limitations of invasive or minimally invasive measurement methods are eliminated.

[0027] In addition, the present invention can quickly predict and output a result value for new input data, i.e. cardiac output, by analyzing high-resolution input data and learning the same using deep learning technology of a deep neural network structure.

[0028] The effects according to the present invention are not limited to those exemplified above, and more diverse effects are included within the present invention.

[0029] FIG. 1 is a schematic diagram of a device for noninvasively calculating cardiac output according to an embodiment of the present invention.

[0030] FIG. 2 is a block diagram showing the configuration of a cardiac output calculation device according to one embodiment of the present invention.

[0031] Figure 3 is a flowchart schematically illustrating a method for calculating cardiac output according to one embodiment of the present invention.

[0032] FIG. 4 is a graph showing input data and output data of a cardiac output calculation device according to one embodiment of the present invention.

[0033] Figure 5 is a flowchart illustrating a method for calculating cardiac output according to one embodiment of the present invention.

[0034] Figure 6 is a graph showing an example of noise in an electrocardiogram waveform and a photoplethysmogram waveform.

[0035] Figure 7 is a graph showing a photoplethysmographic waveform, a VPG waveform, and a PPG waveform.

[0036] Figure 8 is a graph showing an electrocardiogram waveform and a secondary index extracted from the electrocardiogram waveform.

[0037] Figure 9 is a graph showing secondary indices extracted from electrocardiogram waveforms and photoplethysmogram waveforms.

[0038] FIG. 10 is a block diagram showing the architecture of a cardiac output calculation model according to one embodiment of the present invention.

[0039] Figure 11 is a graph comparing predicted values ​​and actual values ​​of a cardiac output calculation model according to an embodiment of the present invention.

[0040] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described in detail below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms. These embodiments are provided solely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims. In connection with the description of the drawings, similar reference numerals may be used for similar components.

[0041] In this document, the expressions "has," "may have," "includes," or "may include" indicate the presence of a feature (e.g., a number, function, operation, or component such as a part), but do not exclude the presence of additional features.

[0042] In this document, the expressions "A or B," "at least one of A and / or B," or "one or more of A or / and B" can include all possible combinations of the listed items. For example, "A or B," "at least one of A and B," or "at least one of A or B" can all refer to cases where (1) at least one A is included, (2) at least one B is included, or (3) at least one A and at least one B are included.

[0043] The terms "first," "second," "first," or "second," as used herein, may describe various components, regardless of order and / or importance, and are only used to distinguish one component from another, without limiting the components. For example, a first user device and a second user device may represent different user devices, regardless of order or importance. For example, without departing from the scope of the rights set forth in this document, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

[0044] When it is said that a component (e.g., a first component) is "(operatively or communicatively) coupled with / to" or "connected to" another component (e.g., a second component), it should be understood that the component is directly coupled to the other component, or can be connected via another component (e.g., a third component). Conversely, when it is said that a component (e.g., a first component) is "directly coupled to" or "directly connected to" another component (e.g., a second component), it should be understood that no other component (e.g., a third component) exists between the first component and the other component.

[0045] The expression "configured to" as used herein can be used interchangeably with, for example, "suitable for," "having the capacity to," "designed to," "adapted to," "made to," or "capable of." The term "configured to" does not necessarily mean something is "specifically designed to" in hardware. Instead, in some contexts, the expression "a device configured to" can mean that the device, together with other devices or components, is "capable of." For example, the phrase "a processor configured (or set) to perform A, B, and C" may mean a dedicated processor (e.g., an embedded processor) for performing those operations, or a general-purpose processor (e.g., a CPU or application processor) that can perform those operations by executing one or more software programs stored in a memory device.

[0046] The terms used in this document are used only to describe specific embodiments and may not be intended to limit the scope of other embodiments. The singular expression may include the plural expression unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as commonly understood by those of ordinary skill in the art described in this document. Terms defined in general dictionaries among the terms used in this document may be interpreted as having the same or similar meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined in this document. In some cases, even if a term is defined in this document, it cannot be interpreted to exclude the embodiments of this document.

[0047] The individual features of the various embodiments of the present invention can be partially or wholly combined or combined with each other, and as can be fully understood by those skilled in the art, various technical connections and operations are possible, and each embodiment can be implemented independently of each other or can be implemented together in a related relationship.

[0048] For clarity in the interpretation of this specification, the terms used in this specification are defined below.

[0049] Figure 1 is a schematic diagram of a cardiac output calculation device according to an embodiment of the present invention.

[0050] A cardiac output calculation device (100) according to an embodiment of the present invention may include a transceiver (10), a memory (12), and a processor (14). The cardiac output calculation device (100) according to an embodiment of the present invention may calculate cardiac output using an electrocardiogram and / or photoplethysmogram received through the transceiver (10). The memory (12) may store data regarding the received electrocardiogram and / or photoplethysmogram.

[0051] Referring to FIG. 1, the cardiac output calculation device (100) can receive an electrocardiogram (ECG) of a user (102) obtained through electrocardiogram (ECG) monitoring. An electrocardiogram is an analysis of the electrical activity or electrical activity of the heart and recorded in the form of waves. ECG monitoring can be performed by attaching electrodes to the skin of a user (102) and monitoring changes in the ECG displayed in time series through an electrocardiogram measuring device (104).

[0052] In addition, referring to FIG. 1, the cardiac output calculation device (100) can receive the user's photoplethysmogram obtained through photoplethysmogram (PPG) monitoring. The measurement method using the photoplethysmogram is a method of measuring pulse waves using light, and by detecting changes in optical properties such as reflection, absorption, and transmittance of biological tissue that appear when the volume of blood vessels changes, an optical sensor can measure the speed of blood flow, oxygen saturation in the blood, etc. The photoplethysmogram monitoring can be performed by attaching a photoplethysmogram measuring device (106) to the user's finger and monitoring changes in the photoplethysmogram displayed in time series through the photoplethysmogram measuring device (106).

[0053] The cardiac output calculation device (100) according to an embodiment of the present invention can be implemented in a personal computer (PC), a data server, or a portable device.

[0054] The portable device may be implemented as a laptop computer, a mobile phone, a smart phone, a tablet PC, a mobile internet device (MID), a personal digital assistant (PDA), an enterprise digital assistant (EDA), a digital still camera, a digital video camera, a portable multimedia player (PMP), a personal navigation device (PND), a handheld game console, an e-book, or a smart device. The smart device may be implemented as a smart watch, a smart band, or a smart ring, which are merely examples and do not limit the implementation examples of the cardiac output calculation device (100).

[0055] The electrocardiogram measuring device (104) and / or the photoplethysmography measuring device (106) are electronic devices capable of measuring and outputting biosignals, and, like the cardiac output calculation device (100), can be implemented in a personal computer (PC), a data server, or a portable device.

[0056] The processor (14) of the cardiac output calculation device (100) may include a deep learning model. For example, the processor (14) of the cardiac output calculation device (100) may include a cardiac output calculation model (16).

[0057] The cardiac output calculation model (16) may include a neural network (or artificial neural network). Neural networks include statistical learning algorithms that mimic biological neurons in machine learning and cognitive science. A neural network may refer to a general model in which artificial neurons (nodes) formed by synaptic connections form a network and change the strength of synaptic connections through learning, thereby possessing problem-solving capabilities. In embodiments of the present invention, the cardiac output calculation model (16) may be expressed as a type of neural network and may include a cardiac output calculation algorithm.

[0058] Neurons in a neural network can contain a combination of weights or biases. A neural network can include one or more layers, each consisting of one or more neurons or nodes. A neural network can infer a desired outcome from any input by changing the weights of its neurons through learning.

[0059] Neural networks may include convolutional neural networks (CNNs). Neural networks include Deep Neural Network (DNN), Recurrent Neural Network (RNN), perceptron, multilayer perceptron, Feed Forward (FF), Radial Basis Network (RBF), Deep Feed Forward (DFF), Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Auto Encoder (AE), and Variational Auto (VAE). Encoder), DAE (Denoising Auto Encoder), SAE (Sparse Auto Encoder), MC (Markov Chain), HN (Hopfield Network), BM (Boltzmann Machine), RBM (Restricted Boltzmann Machine), DBN (Depp Belief Network), DCN (Deep Convolutional Network), DN (Deconvolutional Network), DCIGN (Deep Convolutional Inverse Graphics Network), Generative Adversarial Network (GAN), Liquid State Machine (LSM), Extreme Learning Machine (ELM), It will be understood by those skilled in the art that any neural network may be included, including but not limited to, an Echo State Network (ESN), a Deep Residual Network (DRN), a Differentiable Neural Computer (DNC), a Neural Turning Machine (NTM), a Capsule Network (CN), a Kohonen Network (KN), and / or an Attention Network (AN).

[0060] A cardiac output calculation model (16) according to an embodiment of the present invention may include a deep learning model including a cardiac output calculation algorithm, and details of the cardiac output calculation model (16) according to an embodiment of the present invention are described later in FIG. 10.

[0061]

[0062] FIG. 2 is a block diagram showing the configuration of a cardiac output calculation device according to one embodiment of the present invention.

[0063] Referring to FIG. 2, the cardiac output calculation device (200) may include a memory interface (210), one or more processors (220), and a peripheral interface (230). The cardiac output calculation device (200) of FIG. 2 may be applied to the cardiac output calculation device (100) of FIG. 1.

[0064] Various components within the cardiac output calculation device (200) may be connected by one or more communication buses or signal lines.

[0065] The memory interface (210) is connected to the memory (250) and can transmit various data to one or more processors (220). Here, the memory (250) can include at least one type of storage medium among flash memory type, hard disk type, multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM, SRAM, ROM, EEPROM, PROM, network storage, cloud, and blockchain database.

[0066] In various embodiments, the memory (250) may store at least one of an operating system (251), a communication module (252), a graphical user interface module (GUI) (253), a sensor processing module (254), a telephone module (255), and an application module (256). Specifically, the operating system (251) may include instructions for processing basic system services and instructions for performing hardware operations. The communication module (252) may communicate with at least one of one or more other devices, computers, and servers. The graphical user interface module (GUI) (253) may process a graphical user interface. The sensor processing module (254) may process sensor-related functions (e.g., processing voice input received using one or more microphones (292). The telephone module (255) may process telephone-related functions. The application module (256) may perform various functions of a user application, such as electronic messaging, web browsing, media processing, navigation, imaging, and other processing functions. Additionally, the cardiac output calculation device (200) can store one or more software applications (256-1, 256-2) associated with one type of service in the memory (250).

[0067] In various embodiments, the memory (250) may store a digital assistant client module (257) (hereinafter, DA client module), and accordingly, may store commands for performing client-side functions of the digital assistant and various user data (258).

[0068] Meanwhile, the DA client module (257) can obtain the user's voice input, text input, touch input, and / or gesture input through various user interfaces (e.g., I / O subsystem (240)) provided in the cardiac output calculation device (200).

[0069] Additionally, the DA client module (257) can output data in audiovisual and tactile forms. For example, the DA client module (257) can output data consisting of a combination of at least two or more of voice, sound, notification, text message, menu, graphic, video, animation, and vibration. In addition, the DA client module (257) can communicate with a digital assistant server (not shown) using a communication subsystem (280).

[0070] In various embodiments, the DA client module (257) may collect additional information about the surroundings of the cardiac output calculation device (200) from various sensors, subsystems, and peripheral devices to construct a context associated with the user input. For example, the DA client module (257) may provide context information along with the user input to a digital assistant server to infer the user's intent. Here, the context information that may accompany the user input may include sensor information, such as lighting, ambient noise, ambient temperature, images of the surrounding environment, video, etc. As another example, the context information may include the physical state of the cardiac output calculation device (200) (e.g., device orientation, device location, device temperature, power level, speed, acceleration, motion pattern, cellular signal strength, etc.). As another example, the context information may include information related to the software status of the cardiac output calculation device (200) (e.g., processes running on the cardiac output calculation device (200), installed programs, past and present network activity, background services, error logs, resource usage, etc.).

[0071] In various embodiments, the memory (250) may include added or deleted instructions, and further, the cardiac output calculation device (200) may include additional configurations other than those illustrated in FIG. 2, or may exclude some configurations.

[0072] The processor (220) can control the overall operation of the cardiac output calculation device (200) and can execute various commands to calculate and / or output the user's cardiac output by running an application or program stored in the memory (250).

[0073] The processor (220) may correspond to a computing device such as a CPU (Central Processing Unit) or an AP (Application Processor). In addition, the processor (220) may be implemented in the form of an integrated chip (IC) such as a SoC (System on Chip) that integrates various computing devices such as an NPU (Neural Processing Unit).

[0074] In various embodiments, the processor (220) may receive at least one of first biosignal data including an electrocardiogram and a photoplethysmogram, and demographic information data.

[0075] The first bio-signal data may include electrocardiogram waveform data, photoplethysmography waveform data, and stroke volume data calculated based on arterial blood pressure calculated using APCO (Arterial Pressure Based Cardiac Output) equipment. The demographic information data may include at least one of the user's age, gender, height, and weight. The processor (220) may preprocess the first bio-signal data. In addition, the processor (220) may extract second bio-signal data from the preprocessed bio-signal data. In addition, the processor (220) may infer stroke volume using a deep learning model configured to input the first bio-signal data and the second bio-signal data and output stroke volume. The detailed description of the operation of the processor (220) will be described below with reference to FIG. 3.

[0076] The peripheral interface (230) can be connected to various sensors, subsystems, and peripheral devices to provide data so that the cardiac output calculation device (200) can perform various functions. Here, it can be understood that the function performed by the cardiac output calculation device (200) is performed by the processor (220).

[0077] The peripheral interface (230) can receive data from a motion sensor (260), a light sensor (optical sensor) (261), and a proximity sensor (262), through which the cardiac output calculation device (200) can perform orientation, light, and proximity detection functions, etc. For another example, the peripheral interface (230) can receive data from other sensors (263) (positioning system - GPS receiver, temperature sensor, biometric sensor), through which the cardiac output calculation device (200) can perform functions related to the other sensors (263).

[0078] In various embodiments, the cardiac output calculation device (200) may include a camera subsystem (270) connected to a peripheral interface (230) and an optical sensor (271) connected thereto, through which the cardiac output calculation device (200) may perform various photographing functions, such as taking pictures and recording video clips.

[0079] In various embodiments, the cardiac output calculation device (200) may include a communication subsystem (280) connected to a peripheral interface (230). The communication subsystem (280) may be comprised of one or more wired / wireless networks and may include various communication ports, radio frequency transceivers, and optical transceivers.

[0080] In various embodiments, the cardiac output calculation device (200) includes an audio subsystem (290) connected to the peripheral interface (230), the audio subsystem (290) including one or more speakers (291) and one or more microphones (292), such that the cardiac output calculation device (200) can perform voice-activated functions, such as voice recognition, voice duplication, digital recording, and telephone functions.

[0081] In various embodiments, the cardiac output calculation device (200) may include an I / O subsystem (240) connected to a peripheral interface (230). For example, the I / O subsystem (240) may control a touch screen (243) included in the cardiac output calculation device (200) via a touch screen controller (241). As an example, the touch screen controller (241) may detect a user's contact and movement or cessation of contact and movement using any one of a plurality of touch sensing technologies, such as capacitive, resistive, infrared, surface acoustic wave technology, or a proximity sensor array. As another example, the I / O subsystem (240) may control other input / control devices (244) included in the cardiac output calculation device (200) via other input controller(s) (242). As an example, the other input controller(s) (242) may control one or more buttons, rocker switches, thumb-wheels, infrared ports, USB ports, and pointer devices such as a stylus.

[0082]

[0083] Figure 3 is a flowchart schematically illustrating a method for calculating cardiac output according to one embodiment of the present invention.

[0084] Referring to FIG. 3, a cardiac output calculation device (100, FIG. 1) (or, transceiver (10, FIG. 1)) according to one embodiment of the present invention can receive first biosignal data including an electrocardiogram and photoplethysmogram of a user (S302).

[0085]

[0086] *The first biosignal data may include electrocardiogram waveforms, photoplethysmogram waveforms, and stroke volume calculated based on arterial blood pressure using APCO (Arterial Pressure Based Cardiac Output) equipment. For example, electrocardiogram waveforms and photoplethysmogram waveforms measured at a sample rate of 100 Hz, and stroke volume measured at 2-second intervals may be collected from the publicly available dataset VitalDB.

[0087] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can determine a stroke volume of a user based on the user's first bio-signal data by using a deep learning model configured to output stroke volume by inputting electrocardiogram and photoplethysmogram (S304).

[0088] The deep learning model may be a deep learning model trained using training data for multiple patients.

[0089] The learning data for a plurality of patients may include a first derivative PPG (VPG) waveform, which is a first derivative waveform of the photoplethysmogram, and a second derivative PPG (APG) waveform, which is a second derivative waveform of the photoplethysmogram. The learning data for a plurality of patients may include a PR interval, a QRS duration, a QRS amplitude, and an ST segment extracted from the electrocardiogram of each of the plurality of patients. In addition, the learning data for a plurality of patients may include a PTT (Pulse Transit Time) extracted from the electrocardiogram and the photoplethysmogram of each of the plurality of patients. The learning data for a plurality of patients may be referred to as second biosignal data.

[0090] A deep learning model can determine a stroke volume of a user by inputting the user's first bio-signal data, second bio-signal data, and demographic information data. The second bio-signal data can include the VPG waveform, the APG waveform, the PR interval, the QRS duration, the QRS amplitude, the ST segment, and the PTT.

[0091]

[0092] FIG. 4 is a graph showing input data and output data of a cardiac output calculation device according to one embodiment of the present invention.

[0093] Referring to FIG. 4, a graph (400) representing input data and output data of a cardiac output calculation device is illustrated. The horizontal axis of the graph (400) represents time (t), and the vertical axis of the graph (400) represents stroke volume.

[0094] The input data may be electrocardiogram (ECG) data (30) and / or photoplethysmogram (PPG) data (32) of the patient. The patient's ECG data (30) and / or photoplethysmogram (PPG) data (32) may be part of the patient's biosignals collected during surgery. The patient's ECG data (30) and / or photoplethysmogram (PPG) data (32) may be waveform data for a predetermined period of time.

[0095] For example, the patient's electrocardiogram (ECG) data (30) and / or photoplethysmogram (PPG) data (32) may be waveform data for 20 seconds. For example, the patient's electrocardiogram (ECG) data (30) and / or photoplethysmogram (PPG) data (32) for a predetermined period immediately prior to an arbitrary time t may be used to calculate and output cardiac output at an arbitrary time t. The predetermined period may be 20 seconds.

[0096] For example, the patient's electrocardiogram (ECG) data (30) and / or photoplethysmogram (PPG) data (32) may be waveform data for 60 seconds. For example, the patient's electrocardiogram (ECG) data (30) and / or photoplethysmogram (PPG) data (32) for a predetermined period immediately prior to an arbitrary time t may be used to calculate and output cardiac output at an arbitrary time t. The predetermined period may be 60 seconds.

[0097] The output data may be stroke volume data (34). For example, the output data may be a value at a point in time when the stroke volume changes, such as the timing indicated by the arrow in Fig. 4.

[0098] Cardiac output can be calculated as the product of stroke volume and heart rate. The output data of the cardiac output calculation device according to one embodiment of the present invention may be cardiac output. The output data of the cardiac output calculation device according to another embodiment of the present invention may be stroke volume.

[0099]

[0100] Figure 5 is a flowchart illustrating a method for calculating cardiac output according to one embodiment of the present invention.

[0101] Referring to FIG. 5, a cardiac output calculation device (100, FIG. 1) (or a transceiver (10, FIG. 1)) according to an embodiment of the present invention can receive at least one of first bio-signal data including an electrocardiogram and a photoplethysmogram, and demographic information data (S502). The first bio-signal data can include electrocardiogram waveform data, photoplethysmogram waveform data, and stroke volume data calculated based on arterial blood pressure calculated using APCO (Arterial Pressure Based Cardiac Output) equipment. The demographic information data can include at least one of the user's age, gender, height, and weight.

[0102] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can preprocess first biosignal data (S504).

[0103] The cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can slice the first biosignal data into a predetermined length, extract sampling data from the sliced ​​first biosignal data, and remove noise data from the first biosignal data.

[0104] The cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can select a sample data set to be used for the collected electrocardiogram waveform data within predetermined criteria.

[0105] For example, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can extract a candidate data set of a first reference sample from the collected electrocardiogram waveform data, and the candidate data set of the first reference sample can be used as learning data. For example, the first reference can be 2 or more and 4.5 or less.

[0106] As another example, abnormal ECG waveforms can be excluded from the dataset only if the standard deviation of the RR interval is less than 30. The RR interval refers to the interval between R-peaks of the heartbeat.

[0107] The cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can select a sample data set to be used for the collected photoplethysmography data within predetermined criteria.

[0108] For example, the cardiac output calculation device (100, FIG. 1) (or the processor (14, FIG. 1)) can extract a candidate data set of second reference samples from the collected photoplethysmography waveform data, and the candidate data set of the second reference samples can be used as learning data. For example, the second reference can be 0 or more and 100 or less.

[0109] As another example, if the frequency of the photoplethysmography waveform is below a predetermined standard or if the photoplethysmography waveform is determined to be irregular, it may be excluded from the sample data set.

[0110] As another example, samples with a correlation within the beats of the photoplethysmogram of 0.9 or greater can be selected as candidate data sets, and samples with a standard deviation of 0 can be excluded.

[0111] The cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can perform minimum-maximum normalization of the photoplethysmogram.

[0112] The waveform of the photoplethysmogram measured by the photoplethysmogram measuring device (106) can be scaled in amplitude by the photoplethysmogram measuring device (106). Since the scaling factor(s) used in scaling cannot be determined from the user's perspective, minimum-maximum normalization can be performed to remove information about the inaccurate amplitude of the photoplethysmogram.

[0113] Additionally, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can remove noise from signals appearing in electrocardiogram waveforms and / or photoplethysmogram waveforms.

[0114] For example, the cardiac output calculation device (100, FIG. 1) (or the processor (14, FIG. 1)) can remove frequency components below 0.5 Hz from the electrocardiogram waveform and / or the photoplethysmogram waveform. As another example, frequency components below a cutoff frequency of 0.5 Hz and above 8 Hz can be blocked from the photoplethysmogram waveform.

[0115] In relation to this, Fig. 6 is a graph showing an example of noise in an electrocardiogram waveform and a photoplethysmogram waveform.

[0116] Referring to FIG. 6, a graph (600) showing an electrocardiogram waveform and a photoplethysmogram waveform received by a cardiac output calculation device (100, FIG. 1) is illustrated.

[0117] The cardiac output calculation device (100, FIG. 1) (or, transceiver (10, FIG. 1)) can receive first bio-signal data including electrocardiograms and photoplethysmograms of multiple patients measured during surgery in an operating room. Due to motion artifact, electrocautery, wall power, etc., the waveform (60) of the electrocardiogram collected in the operating room may include noise data. Due to motion artifact, nerve stimulation, respiration, etc., the waveform (62) of the photoplethysmogram collected in the operating room may include noise data such as baseline drift.

[0118] Referring again to S504 of FIG. 5, the step of preprocessing the first biosignal data may include a step of removing noise in the signal appearing in the electrocardiogram waveform and / or photoplethysmogram waveform that interferes with learning of the cardiac output calculation model (16, FIG. 1).

[0119] The stroke volume calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can select a sample data set to be used for the collected stroke volume data within predetermined criteria.

[0120] For example, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can extract a candidate data set of a third reference sample from the collected stroke volume data, and the candidate data set of the third reference sample can be used as training data. For example, the third reference can be 20 mmHg or more and 200 mmHg or less.

[0121] Since stroke volume calculated using APCO equipment is information calculated based on arterial blood pressure, the recorded arterial blood pressure value may have been recorded without being placed in the artery or may be a value due to artifact caused by flushing of the arterial tube. Therefore, selection according to the third criterion is to prevent this.

[0122] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can extract second biosignal data from preprocessed biosignal data (S506).

[0123] According to one embodiment of the present invention, a cardiac output calculation device (100, FIG. 1) (or a processor (14, FIG. 1)) can extract a VPG (first derivative PPG) waveform, which is a first derivative waveform of a photoplethysmogram, and an APG (second derivative PPG) waveform, which is a second derivative waveform of a photoplethysmogram, from a waveform of a preprocessed photoplethysmogram, and utilize them as secondary indices when calculating cardiac output. The VPG waveform and / or the APG waveform may be referred to as 'second biosignal data'.

[0124] In relation to this, Fig. 7 is a graph showing a photoplethysmography waveform and a secondary index extracted from the photoplethysmography waveform.

[0125] Referring to FIG. 7, graphs showing photoplethysmography (PPG) waveforms (70), VPG waveforms (72), and APG waveforms (74) are shown.

[0126] In order to compensate for the lack of information by removing information about the amplitude of the inaccurate photoplethysmogram waveform by performing min-max normalization of the photoplethysmogram waveform, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention may additionally utilize the VPG waveform (72) and / or the APG waveform (74) in addition to the photoplethysmogram (PPG) waveform (70) when calculating cardiac output.

[0127] Referring again to S506 of FIG. 5, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can extract PR interval, QRS duration, QRS amplitude, and ST segment from a preprocessed electrocardiogram waveform and use them as secondary indices when calculating cardiac output.

[0128] In relation to this, Fig. 7 is a graph showing an electrocardiogram waveform and a secondary index extracted from the electrocardiogram waveform.

[0129] Referring to FIG. 7, a graph showing an electrocardiogram waveform (800), PR interval, QRS duration, QRS amplitude, and ST segment is shown.

[0130] The PR interval can be defined as the time interval from the beginning of the P wave to the beginning of the QRS complex on the electrocardiogram, and it represents the time during which nerve conduction takes place in the heart. The QRS duration can be defined as the time it takes for the QRS complex to appear on the electrocardiogram, and it represents the time during which the heart contracts. The QRS amplitude represents the maximum difference in polarity of the QRS complex, and it can be used to diagnose the condition of the heart by measuring the strength of the electrical activity occurring in the heart. The ST segment can be defined as the time interval from the end of the QRS complex to the beginning of the T wave, and the condition of the heart can be diagnosed by the polarity changes that occur between the contraction and relaxation of the heart.

[0131] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can additionally utilize PR interval, QRS duration, QRS amplitude, and ST segment in addition to the electrocardiogram waveform (800) when calculating cardiac output. For example, a cardiac output calculation model (16, FIG. 1) can learn the characteristics of cardiac contraction using secondary indices extracted from the ECG.

[0132] Referring again to S506 of FIG. 5, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can extract PTT (Pulse Transit Time) from the preprocessed electrocardiogram waveform and photoplethysmogram and use it as a secondary index when calculating cardiac output.

[0133] In relation to this, Fig. 9 is a graph showing secondary indices extracted from electrocardiogram waveforms and photoplethysmogram waveforms.

[0134] Referring to FIG. 9, a graph showing an electrocardiogram (ECG), a photoplethysmogram (PPG), and a PTT (90) is shown.

[0135] Pulse Transit Time (PTT) can be defined as the time between the peak of the QRS complex and a specific reference point on the PPG waveform measured at the fingertip. PTT can be interpreted as the pulse generation time, i.e., the time it takes for blood to contract from the heart and be delivered to the peripheral vascular measurement site. PTT is an indicator used to more efficiently analyze and evaluate blood pressure and can also be used as an indicator of vascular elasticity.

[0136] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can additionally utilize PTT (90) as well as an electrocardiogram (ECG) and a photoplethysmogram (PPG) waveform when calculating cardiac output.

[0137] Referring again to FIG. 5, the cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) according to one embodiment of the present invention can infer stroke volume using a deep learning model configured to output stroke volume by inputting first biosignal data and second biosignal data (S508).

[0138] The processor (14, FIG. 1) of the cardiac output calculation device (100, FIG. 1) can match and store electrocardiogram waveforms, photoplethysmogram waveforms, secondary indices extracted from electrocardiogram waveforms and / or photoplethysmogram waveforms, stroke volume information, and demographic information including age, gender, height, and weight for a specific patient. The stored data described above can be used as learning data.

[0139] The processor (14, FIG. 1) of the cardiac output calculation device (100, FIG. 1) can train a cardiac output calculation model (16, FIG. 1). The cardiac output calculation model (16, FIG. 1) may be a deep learning model having weight values ​​obtained by extracting and combining various features.

[0140] The processor (14, FIG. 1) of the cardiac output calculation device (100, FIG. 1) can smooth learning data. The smoothed learning data can be input into a cardiac output calculation model (16, FIG. 1), and the cardiac output calculation model (16, FIG. 1) can be trained by extracting and combining various features to update weights.

[0141] A cardiac output calculation device (100, FIG. 1) (or processor (14, FIG. 1)) can infer stroke volume of a specific patient using biometric data including electrocardiogram and photoplethysmogram, stroke volume information, demographic information including age, gender, height and weight, and a cardiac output calculation model (16, FIG. 1).

[0142]

[0143] FIG. 10 is a block diagram showing the architecture of a cardiac output calculation model according to one embodiment of the present invention.

[0144] Referring to FIGS. 1 and 10, the cardiac output calculation model (1000) can be applied to the cardiac output calculation model (16) of FIG. 1.

[0145] Referring to FIG. 10, the input data of the cardiac output calculation model (1000) may include bio signal data (1002) including an electrocardiogram waveform, a photoplethysmogram waveform, a VPG waveform, and an APG waveform, and demographic data (Demographics) (1004) including age, gender, height, and weight.

[0146] The biosignal data (1002) input to the cardiac output calculation model (1000) may be continuous data sliced ​​by a predetermined period. The predetermined period may be referred to as a "sliding window." The patient's electrocardiogram waveform and / or photoplethysmogram waveform may be input to the cardiac output calculation model (1000) by sliding by a predetermined sliding window (e.g., 20 seconds).

[0147] The input data of the cardiac output calculation model (1000) may include data (Extracted Features) (1006) extracted from the waveform of an electrocardiogram and / or the waveform of a photoplethysmogram.

[0148] Data (1006) extracted from the waveform of the electrocardiogram and / or the waveform of the photoplethysmogram may include PR interval, QRS duration, QRS amplitude, and ST segment extracted from the electrocardiogram waveform, and PTT extracted from the electrocardiogram waveform and the photoplethysmogram waveform. Data (1006) extracted from the waveform of the electrocardiogram and / or the waveform of the photoplethysmogram input to the cardiac output calculation model (1000) may be an average of values ​​calculated over a predetermined period. For example, an average value of each of the PR interval, QRS duration, QRS amplitude, ST segment, and PTT values ​​calculated over 20 seconds or 60 seconds may be input to the cardiac output calculation model (1000).

[0149] Bio signal data (1002) including an electrocardiogram waveform, a photoplethysmogram waveform, a VPG waveform, and an APG waveform can be normalized for training a cardiac output calculation model (1000). In addition, demographic data (Demographics) (1004), data (1006) extracted from an electrocardiogram waveform and / or a photoplethysmogram waveform can be normalized for training a cardiac output calculation model (1000). Accordingly, bio signal data (1002), demographic data (Demographics) (1004), data (1006) extracted from an electrocardiogram waveform and / or a photoplethysmogram waveform can be normalized and input to the cardiac output calculation model (1000).

[0150] A cardiac output calculation model (1000) according to one embodiment of the present invention can be trained to enable more accurate prediction of cardiac output using high-resolution biosignal data (1002). In addition, by extracting and utilizing secondary indicators related to cardiac output (e.g., APG waveform, VPG waveform, PR interval, QRS duration, QRS amplitude, ST segment, PTT) from high-resolution electrocardiogram waveforms and photoplethysmogram waveforms, the performance of the cardiac output calculation model (1000) can be improved.

[0151] A cardiac output calculation model (1000) according to one embodiment of the present invention may have a CNN (Convolution Neural Network) structure including at least one SE (Squeeze and Excitation) block and at least one Residual block.

[0152] For example, the cardiac output calculation model (1000) may include a convolutional neural network of the SE-ResNet structure.

[0153] ResNet can be defined as an algorithm that overcomes the vanishing gradient problem that can occur as layers become deeper by learning the residual, which is the difference between the input and output of each layer. SE can be defined as a module that performs operations that squeeze and excite information. SE calculates the importance of each channel by considering the interdependence between the channels in the feature map resulting from each layer, and recalibrates by assigning weights to important channels.

[0154] The cardiac output calculation model (1000) can perform convolution using convolution filters of various sizes while passing through multiple layers, and can simultaneously perform residual learning. In addition, the cardiac output calculation model (1000) can use max pooling and / or GAP (Global Average Pooling) as a squeeze operation.

[0155] The cardiac output calculation model (1000) can infer the stroke volume of a specific patient using the input data (1002, 1004, 1006) described above and the learned CNN structure including the SE block and the Residual block. The output data (1008) of the cardiac output calculation model (1000) may be the stroke volume. For example, the cardiac output calculation model (1000) may output the stroke volume once every 8 seconds.

[0156] The cardiac output calculation model (1000) may be an algorithm of any structure including a convolutional neural network, and the technical scope of the present invention is not limited or restricted by the embodiment of the cardiac output calculation model (1000) described above.

[0157]

[0158] According to one embodiment of the present disclosure, a data set used for development and training of a cardiac output calculation model (1000) includes biometric data including electrocardiograms and photoplethysmograms collected during 3,970 surgeries at a specific hospital, stroke volume information, and demographic information including age, gender, height, and weight.

[0159] Specifically, a dataset collected using the time series holdout method during 3,970 surgeries at a specific hospital was divided into a training dataset, a validation dataset, and a test dataset. The training dataset, validation dataset, and test dataset were organized in a ratio of 6:2:2. The training dataset included 741,150 samples collected during 2,777 surgeries, and the test dataset included 196,686 samples collected during 694 surgeries.

[0160] In order to evaluate the validity of the stroke volume (i.e., predicted value) output from the cardiac output calculation model and the stroke volume (i.e., actual value) calculated based on arterial blood pressure, the cardiac output calculation model according to one embodiment of the present invention used Percentage Error, MAPE (mean absolute percentage error), and MAE (Mean Absolute error).

[0161] Percentage error can be interpreted as an indicator to evaluate whether the variation in the difference between the actual value and the predicted value is at an acceptable level.

[0162] MAPE (Mean Absolute Percentage Error) is the average of the percentage errors between actual and predicted values, expressed as a percentage of MAE. For example, if a cardiac output calculation model has an MAE of 12 ml / beat, it means that the model incorrectly predicted 12 ml / beat on average.

[0163] MAE (Mean Absolute Error) is the average of the absolute errors between the actual and predicted values. For example, a MAPE of 12% for a cardiac output calculation model means that the actual and predicted values ​​differ by approximately 12%.

[0164] Percentage Error, MAPE, and MAE indicate better performance when their values ​​are lower.

[0165] The performance evaluation results of the cardiac output calculation model according to one embodiment of the present invention can be expressed as shown in Table 1 below.

[0166] Percentage error of 10min median (%)(10min interval)Percentage Error (%)MAPE (%)MAE (ml / beat)Result37.641.516.112.2

[0167] Referring to Table 1, since stroke volume is not a value that actually changes frequently, a median filter was applied to the predicted values ​​and actual values ​​measured at 10-minute intervals to calculate the percentage error, and the percentage error was measured to be 37.6%. FIG. 11 shows graphs comparing the predicted values ​​and actual values ​​of the cardiac output calculation model according to an embodiment of the present invention.

[0168] Referring to FIG. 11, graphs for continuous cardiac output (1101, 1102, 1103) and graphs for cardiac output measured at 10-minute intervals (1104, 1105, 1106) are shown.

[0169] Referring to Figure 11, various indicators of actual and predicted values ​​were compared in the form of a scatter plot (1101, 1104), a Bland-Altman plot (1102, 1105), and a four quadrant plot (1103, 1106).

[0170] Referring to the scatter plot (1101, 1104), it can be confirmed that there is a linear relationship between the continuous cardiac output (Target SV (mL)) calculated based on arterial blood pressure and the cardiac output (Predicted SV (mL)) output from the cardiac output calculation model.

[0171] Referring to the Bland-Altman plot (1102, 1105), we can see that most of the points are located around the bias.

[0172] A Bland-Altman plot is a scatterplot that calculates the mean and difference between each pair of measurements for the same subject (stroke volume derived from arterial blood pressure and stroke volume output from a cardiac output calculation model). The x-axis of the Bland-Altman plot represents the mean (Average of Differences (mL)), and the y-axis represents the difference (Diff. of Differences (mL)).

[0173] A Bland-Altman plot can display three horizontal dashed lines parallel to the x-axis, which can be interpreted as values ​​summarizing the degree of discrepancy. Among the three horizontal dashed lines, the upper line represents the upper limit of the 95% agreement limit, the center line represents the difference from the mean, and the lower line represents the lower limit of the 95% agreement limit. According to one embodiment of the present disclosure, the Percentage Error can be calculated as shown in the following mathematical expression 1.

[0174] [Mathematical Formula 1]

[0175] Percentage Error = 100*((1.96*SD) / mean of Stroke volume)

[0176]

[0177] Referring to the Four Quadrant Plot (1106) for cardiac output measured at 10-minute intervals, it can be confirmed that the concordance rate between the actual and predicted values ​​is 79.2%, indicating a high proportion of points with positive correlation.

[0178]

[0179] Although the embodiments of the present invention have been described in more detail with reference to the attached drawings, the present invention is not necessarily limited to these embodiments, and various modifications may be implemented without departing from the technical spirit of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical spirit of the present invention, but to explain it, and the scope of the technical spirit of the present invention is not limited by these embodiments. Therefore, it should be understood that the embodiments described above are illustrative in all aspects and not restrictive. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within a scope equivalent thereto should be interpreted as being included in the scope of the rights of the present invention.

Claims

1. A method of operating a device for calculating cardiac output noninvasively, A step of receiving first biosignal data including an electrocardiogram and photoplethysmogram of a user; A step of determining the stroke volume of the user based on the first bio-signal data of the user, using a deep learning model configured to output stroke volume by inputting electrocardiogram and photoplethysmogram; comprising; A method of operating a device for noninvasively calculating cardiac output.

2. In paragraph 1, The above deep learning model is trained using training data for multiple patients, The above learning data for the above multiple patients is, Including the VPG (first derivative PPG) waveform, which is the first derivative waveform of the photovolume pulse wave, and the APG (second derivative PPG) waveform, which is the second derivative waveform of the photovolume pulse wave. A method of operating a device for noninvasively calculating cardiac output.

3. In paragraph 2, The above learning data for the above multiple patients is, Including the PR interval, QRS duration, QRS amplitude, and ST segment extracted from the electrocardiogram of each of the plurality of patients. A method of operating a device for noninvasively calculating cardiac output.

4. In paragraph 3, The above learning data for the above multiple patients is, Including the PTT (Pulse Transit Time) extracted from the electrocardiogram and the photoplethysmogram of each of the plurality of patients, A method of operating a device for noninvasively calculating cardiac output.

5. In paragraph 4, Further comprising the step of receiving demographic information data of the user; The deep learning model is configured to determine the stroke volume of the user by using the first biosignal data, the second biosignal data, and the demographic information data of the user as inputs, The above demographic information data includes at least one of the user's age, gender, height, and weight, The second biosignal data includes the VPG waveform, the APG waveform, the PR interval, the QRS duration, the QRS amplitude, the ST segment, and the PTT. A method of operating a device for noninvasively calculating cardiac output.

6. As a device for calculating cardiac output noninvasively, memory; A transceiver for receiving at least one of first biosignal data including electrocardiogram and photoplethysmogram, and demographic information data; and A processor including a deep learning model; The above processor, Preprocessing the first bio-signal data, extracting second bio-signal data from the preprocessed bio-signal data, and inferring stroke volume using a deep learning model configured to output stroke volume using the first bio-signal data and the second bio-signal data as inputs. A device that noninvasively calculates cardiac output.

7. In paragraph 6, The above processor, It is further configured to slice the first biosignal data into a predetermined length, extract sampling data from the sliced ​​first biosignal data, and remove noise data of the first biosignal data. A device that noninvasively calculates cardiac output.

8. In paragraph 7, The above second biosignal data is, The first derivative PPG (VPG) waveform, which is the first derivative waveform of the photoplethysmogram, and the second derivative PPG (APG) waveform, which is the second derivative waveform of the photoplethysmogram, PR interval, QRS duration, QRS amplitude, and ST segment extracted from the above electrocardiogram, At least one of the electrocardiogram and the PTT (Pulse Transit Time) extracted from the photoplethysmogram, A device that noninvasively calculates cardiac output.

9. In paragraph 8, The above processor, Perform minimum-maximum normalization of the above photoplethysmogram, It is further configured to extract the VPG waveform and the APG waveform of the photoplethysmographic wave, The above VPG waveform and the above APG waveform are input to the deep learning model in the form of time series data having the predetermined length. A device that noninvasively calculates cardiac output.

10. In paragraph 8, The PR interval, the QRS duration, the QRS amplitude, the ST segment, and the PTT are During a time corresponding to the above-mentioned predetermined length, the average of each value measured multiple times is input to the deep learning model. A device that noninvasively calculates cardiac output.

11. In paragraph 6, The above learning data is, Including the first bio-signal data, the second bio-signal data, stroke volume data, and the demographic information data of multiple patients, The above demographic information data includes at least one of age, gender, height, and weight. A device that noninvasively calculates cardiac output.

12. In paragraph 11, The above memory is, configured to store a set of the learning data matched for each of the plurality of patients, A device that noninvasively calculates cardiac output.

13. In paragraph 6, The above deep learning model is, Having a CNN (convolutional neural network) structure including at least one SE (squeeze and excitation) block and at least one residual block, A device that noninvasively calculates cardiac output.

14. In paragraph 6, The transceiver is further configured to receive the first bio-signal data and the second bio-signal data of the user corresponding to a period from t seconds (where t is a positive integer) to (t+n) seconds (where n is a positive integer); and The processor is further configured to calculate the stroke volume of the user at the (t+n) second using the deep learning model; A device that noninvasively calculates cardiac output.

Images