Method and apparatus for measuring blood glucose and glycated hemoglobin using artificial intelligence technology

Abstract

The present invention relates to technology for measuring blood glucose and glycated hemoglobin using a non-invasive PPG sensor and, more specifically, to a method and an apparatus for measuring blood glucose and glycated hemoglobin by using artificial intelligence technology in which pulse wave signals of a measurer are sensed using a PPG (pulse wave) sensor and analyzed by a machine learning algorithm to estimate blood glucose and glycated hemoglobin levels.

Description

Blood glucose and glycated hemoglobin measurement method and device using artificial intelligence technology

[0001] The present invention relates to a technology for measuring blood sugar and glycated hemoglobin using a non-blood-collecting PPG sensor, and relates to a blood sugar and glycated hemoglobin measuring method and measuring device using artificial intelligence technology that detects a pulse signal of a subject using a PPG (pulse wave) sensor and analyzes it using a machine learning algorithm to estimate blood sugar levels and glycated hemoglobin levels.

[0002] Diabetes is a metabolic disease characterized by hyperglycemia, caused by impaired insulin secretion or function, which is essential for blood sugar regulation. Chronic hyperglycemia caused by diabetes leads to damage and dysfunction in various organs. In particular, it can lead to microvascular complications, such as those affecting the retina, kidneys, and nerves, and macrovascular complications, such as arteriosclerosis, cardiovascular disease, and cerebrovascular disease, which increase mortality.

[0003] However, diabetes can worsen or increase the risk of complications due to blood sugar control, weight loss, and medication. Therefore, diabetic patients should regularly measure their blood sugar levels to manage their blood sugar levels and undergo regular tests for glycated hemoglobin (HbA1C), a treatment indicator as important as blood sugar levels.

[0004] The glycated hemoglobin (HbA1c) test measures the degree to which hemoglobin within red blood cells, which carry oxygen in the blood, has been glycated. It reflects blood sugar changes over the past three to four months, based on the average lifespan of red blood cells. Since glucose naturally exists in healthy individuals, some degree of glycation of hemoglobin in our blood is inherent. While the normal range varies depending on the testing method, a level of up to 5.6% is generally considered normal.

[0005] In diabetic patients, elevated blood glucose levels also lead to elevated levels of glycated hemoglobin (HbA1c). Therefore, these results, which clearly reveal the patient's current blood sugar control, are used to determine future treatment options.

[0006] Meanwhile, the conventional method of measuring glycated hemoglobin (HbA1c) involved collecting blood from a vein in the subject's arm or pricking the fingertip with a small, sharp needle to obtain a capillary blood sample, and measuring the concentration of glycated hemoglobin (HbA1c) using the obtained blood.

[0007] This invasive method of measuring glycated hemoglobin increases the burden of blood collection on the subjects and has the problem of providing inaccurate values ​​in cases of short red blood cell lifespan, pregnancy, or kidney disease.

[0008] As a conventional technology for measuring glycated hemoglobin, Patent Publication No. 10-2023-0044160 (Non-invasive glycated hemoglobin or blood sugar estimation method and device using machine learning) is disclosed. The patent describes a technology that estimates a subject's glycated hemoglobin or blood sugar level by inputting body mass index, blood oxygen saturation, and finger thickness information from the subject's biosignals into a machine learning model.

[0009] However, there is a problem that the accuracy is not high when estimating glycated hemoglobin or blood sugar levels using only body mass index, blood oxygen saturation, and finger thickness.

[0010] The present invention was created to solve the above-mentioned problem, and provides a blood sugar and glycated hemoglobin measuring method and measuring device that uses artificial intelligence technology to detect a pulse signal of a subject using a PPG (pulse wave) sensor, analyze it using a machine learning algorithm, and estimate blood sugar levels and glycated hemoglobin levels.

[0011] In addition, since the measurement is performed while the index finger is placed in the finger resting groove where the PPG sensor is installed while the hand of the measurer is placed, the pulse wave information of the measurer detected by the PPG (pulse wave) sensor can be accurately measured, and a blood sugar and glycated hemoglobin measuring method and measuring device using artificial intelligence technology that can measure blood sugar and glycated hemoglobin with higher accuracy than a watch type that is worn on the wrist in the past are provided.

[0012] The blood sugar and glycated hemoglobin measurement method of the present invention, to which the artificial intelligence technology of the present invention is applied for solving the above-mentioned problem, comprises the steps of: receiving a pulse signal of a subject measured through a biosignal measuring device (100); extracting bioinformation from the pulse signal, calculating each average, calculating blood sugar levels and glycated hemoglobin levels, and inputting the extracted bioinformation into a pre-learned machine learning algorithm to calculate a disease prediction probability of the subject; A step of outputting an alarm level by determining whether the calculated disease prediction probability falls within a set threshold range; wherein the bio-information includes the age, sex, HR (heart rate), HRV (heart rate variability), PI (an index related to the oxygen saturation of blood pressure and the quality of blood pressure signals), AC (an index indicating changes in heart rate and blood pressure by measuring variability with changing blood pressure signals), SDNN (standard deviation of normal RR (heartbeat cycles per minute) intervals in the entire ballistocardiography record), RMSSD (a value expressed as the square root of the average of the sum of the square roots of the differences in adjacent HRV intervals), PNN50 (a ratio of the number of intervals in which the difference in consecutive RR intervals exceeds 50 ms), LF (low frequency in the band of 0.04 to 0.15 Hz), HF (high frequency in the band of 0.15 to 0.4 Hz), and HF_LF (a frequency in the band of 0.5 to 2 MHz obtained by dividing LF by HF).

[0013] The method further includes a step of removing noise from the collected pulse signal before calculating the blood sugar level and glycated hemoglobin level.

[0014] The above warning levels are classified into peace of mind, concern, caution, and suspicion, and when the disease prediction probability is 0% to 10%, it is judged as “peace of mind,” when the disease prediction probability is 11% to 30%, it is judged as “concern,” when the disease prediction probability is 31% to 50%, it is judged as “caution,” and when the disease prediction probability is 51% or higher, it is judged as “suspicion.”

[0015] The blood glucose and glycated hemoglobin measuring device to which the artificial intelligence technology of the present invention is applied comprises: a bio-signal measuring device (100) for acquiring a pulse signal of a subject; a terminal (200) connected to the bio-signal measuring device (100) by wire or wirelessly to receive the pulse signal acquired from the bio-signal measuring device, extract bio-information, estimate blood glucose levels and glycated hemoglobin levels, and store a machine learning algorithm for predicting a disease; or a server (210) in which the machine learning algorithm is stored, through a wireless communication network; wherein the bio-signal measuring device (100) comprises: a measuring device body (110) having an outer surface that is convexly curved from top to bottom so that a subject's hand can be held thereon, a display portion (111) formed at the upper portion to display measurement information, and an operation portion (112) formed at the lower portion of one side; a finger receiving groove (120) formed at a predetermined depth on the upper portion of the outer surface of the measuring device body (110) so that a subject's index finger can be inserted and received; It includes a PPG sensor (130) installed on the inner surface of the finger mounting groove (120) so that the inner surface of the index finger of the measurer is in contact with the inner surface of the index finger of the measurer to detect the pulse signal of the measurer; a control unit (140) that receives the pulse signal detected from the PPG sensor (130) inside the measuring device body (110) and transmits the pulse information to the terminal (200) through wired or wireless communication;

[0016] According to the present invention having the above configuration, the pulse wave information of the subject can be detected and analyzed using a machine learning algorithm to estimate blood sugar levels and glycated hemoglobin levels, and since the measurement is performed while the subject's hand is placed and the index finger is placed in a finger resting groove in which a PPG sensor is installed, pulse wave information with higher accuracy than in the past can be acquired.

[0017] Figure 1 is a configuration diagram of a blood sugar and glycated hemoglobin measuring device to which artificial intelligence technology according to the present invention is applied.

[0018] Figure 2 is a perspective view of a biosignal measuring device according to the present invention;

[0019] Figure 3 is a perspective view viewed from the opposite side of Figure 2;

[0020] Figure 4 is a schematic diagram for explaining the blood sugar and glycated hemoglobin estimation process according to the present invention.

[0021] Hereinafter, with reference to the attached drawings, a blood sugar and glycated hemoglobin measuring method and measuring device using artificial intelligence technology according to a preferred embodiment of the present invention will be described in detail.

[0022] As illustrated, the blood sugar and glycated hemoglobin measuring device using artificial intelligence technology according to the present invention acquires pulse wave information of the subject using a PPG (pulse wave) sensor and transmits it to a terminal, and the terminal analyzes the received pulse wave information to extract biometric information of the subject and estimates blood pressure and glycated hemoglobin levels using the same. Since the device measures while the subject places his or her hand and places only the index finger in the finger resting groove, it is a measuring device that provides reliable results with high accuracy.

[0023] The blood sugar and glycated hemoglobin measuring device of the present invention transmits pulse wave information acquired by a terminal connected by wired or wireless communication, extracts biometric information from the terminal and inputs the extracted information into a machine learning algorithm to estimate blood sugar levels and glycated hemoglobin levels, and the blood sugar levels and glycated hemoglobin levels estimated in this manner can be confirmed on the terminal or displayed on a display unit (11) so that the user can visually check his or her blood sugar levels and glycated hemoglobin levels.

[0024] A terminal is a terminal that is connected to a blood glucose and glycated hemoglobin measuring device by wire or wireless communication (short-range wireless communication), and is equipped with a memory for storing programs, a microprocessor for executing programs, and calculations and control.

[0025] For example, a terminal is a terminal capable of communication, such as a personal computer (PC), laptop, personal digital assistant (PDA), mobile communication terminal, tablet terminal, etc. For convenience of use, a smart pad or smart phone that the measurer always carries can be used as a mobile communication terminal.

[0026] In order to analyze pulse information, a dedicated app (application) is installed on the terminal (50) to connect to the server (55) and then used.

[0027] A machine learning algorithm is installed in the server (55), and the terminal (50) accesses the server to extract bio-information from pulse information received from the control unit (40), inputs this into the machine learning algorithm, analyzes it through machine learning, and estimates blood sugar levels and glycated hemoglobin levels. Measurement information including the estimated blood sugar levels and glycated hemoglobin levels is transmitted to the terminal (50) and displayed on the screen.

[0028] Accordingly, when the dedicated app is run on the terminal, the user places the hand of the user on the measuring device of the present invention while connecting to the server (55) and places the index finger on the finger resting groove to measure the pulse wave, the measured pulse wave information is used to extract bio-information (variables) for estimating blood sugar and glycated hemoglobin, and then the extracted bio-information is input into an algorithm to estimate blood sugar levels and glycated hemoglobin levels, and then the alarm level is determined and the result is displayed on the screen. The user checks the health status, such as whether there is a disease, through the blood sugar and glycated hemoglobin levels based on the blood sugar level and glycated hemoglobin level information displayed on the screen of the terminal.

[0029] Meanwhile, in the present invention, a machine learning algorithm is installed in the server (55) and configured to be accessed through a dedicated app from the terminal (50). However, if the server (55) is not separately provided, a machine learning algorithm program may be installed in the terminal (50) to extract the biometric information of the subject using pulse information from the terminal itself and predict blood sugar levels and glycated hemoglobin.

[0030] The variables for estimating blood sugar and glycated hemoglobin are divided into dependent and independent variables. The dependent variables are blood sugar and glycated hemoglobin, and the independent variables are age, sex, HR (heart rate), HRV (heart rate variability), PI (an index related to the quality of oxygen saturation and blood pressure signals), AC (an index representing changes in heart rate and blood pressure by measuring variability with changing blood pressure signals), SDNN (standard deviation of normal RR (heart rate cycles per minute) intervals in the entire ballistic cardiogram record), RMSSD (a value expressed as the square root of the average of the sum of the square roots of the differences in adjacent HRV intervals), PNN50 (the ratio of the number of intervals in which the difference in consecutive RR intervals exceeds 50 ms), LF (low frequency in the 0.04-0.15 Hz band) and HF (high frequency in the 0.15-0.4 Hz band), and HF_LF (LF divided by HF, frequency in the 0.5-2 MHz band) as variables. Specify.

[0031] The blood glucose and glycated hemoglobin measuring device of the present invention is composed of a measuring device body (10), a finger mounting groove (20), a PPG sensor (30), and a control unit (40) that communicates with a terminal (50) by wire or wireless communication.

[0032] The measuring device body (10) is formed to wrap around the measuring device body while the measuring device is held by the measuring user's hand, and a display portion (11) is formed on the upper part to display measurement information, etc.

[0033] The display unit (11) displays the on / off status and measurement status, as well as measurement information, and can also receive and display blood sugar levels and glycated hemoglobin levels analyzed by the terminal.

[0034] The finger mounting groove (20) is formed on the upper part of the outer surface of the measuring instrument body (10), and is formed with a certain depth and an area slightly larger than the area of ​​the finger so that the index finger of the measuring instrument can be inserted and mounted.

[0035] The outer surface of the measuring device body (10) on which the finger mounting groove (20) is formed is formed to be curved convexly toward the bottom, thereby allowing the user to hold the hand in the most natural state possible, thereby improving finger contact during the measurement process, increasing convenience of use, and reducing stress.

[0036] An operating section (12), such as a power button for power supply or a measurement button used when measuring pulse information, is formed on the lower side of one side of the measuring device body (10), and a charging terminal is installed.

[0037] The above power button and measurement button are formed as push-type buttons that are easy to press and use.

[0038] Additionally, a lamp that flashes when power is supplied due to operation of the operating unit (12) may be installed in the measuring device body (10).

[0039] In addition, the above operation unit (12) may be equipped with a reset button for resetting or initializing when an error occurs.

[0040] The above PPG sensor (30) is installed on the inner surface of the finger mounting groove (20) to be connected to the control unit (40) by a circuit. Of course, in order to prevent the surface of the index finger of the measurer from directly contacting the PPG sensor (30), a transparent panel is separately installed on the outside of the PPG sensor in the hole where the PPG sensor is installed.

[0041] The above PPG sensor (30) detects the pulse signal of the subject when the subject places the index finger on the finger placement groove (20) and the inner surface of the index finger comes into contact with it, and the detected pulse signal of the subject is transmitted to the control unit (40).

[0042] The control unit (40) is configured to include a control board and a microcomputer (MCU), and receives pulse wave information detected by the PPG sensor (30) and transmits the pulse wave information to a terminal (50) connected via wired or wireless communication.

[0043] The control unit (40) may be configured to directly transmit the pulse signal detected by the PPG sensor to the terminal (50), or may be configured to temporarily store the pulse signal in a separate memory and then transmit it to the terminal (50) via wired or wireless communication.

[0044] In order to supply power to the above control unit (40) and PPG sensor (30), the measuring device body (10) may be equipped with a connection terminal to which a conventional wired cable is connected, or a battery other than a commercial power supply may be equipped to supply power.

[0045] Batteries can be mercury cells, accumulators, or rechargeable batteries. For convenience, it's recommended to use a battery rather than a wired cable to supply power. If a battery is used, a separate storage space can be provided on one side of the control board to facilitate battery replacement.

[0046] A method for measuring blood sugar and glycated hemoglobin using the artificial intelligence technology of the present invention is described.

[0047] The blood sugar and glycated hemoglobin measurement method using the artificial intelligence technology of the present invention is a method of receiving a pulse signal of a subject from a PPG sensor to extract biometric information, inputting the extracted biometric information into a machine learning algorithm to estimate the output blood sugar level and glycated hemoglobin level, and predicting the probability of a disease through the estimated value to confirm whether or not the disease has developed.

[0048] The blood sugar and glycated hemoglobin measurement method using the artificial intelligence technology of the present invention is performed by a computer, and the computer stores a computer program that performs the method for measuring blood sugar and glycated hemoglobin. The computer refers to a broad computing device that includes not only general personal computers, including servers, but also smart devices such as smartphones and tablet PCs.

[0049] The blood sugar and glycated hemoglobin measurement method of the present invention includes a step of receiving a pulse signal from a PPG sensor, a step of calculating a disease prediction probability using blood sugar levels and glycated hemoglobin levels, and a step of determining and outputting an alarm level.

[0050] Specifically, when a pulse signal is input from a PPG sensor, noise in the input signal is removed. To remove noise, the signal size of the input pulse signal is scaled.

[0051] A pulse signal corresponding to a preset minimum root mean square (RMS) value and maximum root mean square (RMS) value range is extracted from the scaled pulse signal, and a signal corresponding to a preset minimum amplitude value and maximum amplitude value is extracted from the extracted pulse signal, thereby completing noise removal of the final input pulse signal.

[0052] The preset minimum root mean square (RMS) value, maximum root mean square (RMS) value, minimum amplitude value, and maximum amplitude value are average values ​​extracted from the pulse signal of the measurer, and can be set in various ways depending on the measurer's hand movement or index finger bending behavior.

[0053] The pulse signal with noise removed is input into a machine learning algorithm to estimate blood sugar levels and glycated hemoglobin levels, and the disease prediction probability is output.

[0054] The biometric information of the pulse signal is designated as independent variables, including age, sex, HR (heart rate), HRV (heart rate variability), PI (an index related to the oxygen saturation of blood pressure and the quality of blood pressure signals), AC (an index representing changes in heart rate and blood pressure by measuring the variability of changing blood pressure signals), SDNN (standard deviation of normal RR (heartbeat cycles per minute) intervals in the entire ballistocardiographic record), RMSSD (the square root of the average of the sum of the square roots of the differences in adjacent HRV intervals), PNN50 (the ratio of the number of intervals in which the difference in consecutive RR intervals exceeds 50 ms), LF (low frequency in the 0.04–0.15 Hz band), HF (high frequency in the 0.15–0.4 Hz band), and HF_LF (LF divided by HF, which is the frequency in the 0.5–2 MHz band).

[0055] In the present invention, by additionally setting variables for age, sex, heart rate, heart rate variability, PI (an index related to the oxygen saturation of blood pressure and the quality of blood pressure signals), AC (an index indicating changes in heart rate and blood pressure by measuring variability with changing blood pressure signals), SDNN (standard deviation of normal RR (heart rate cycles per minute) intervals in the entire ballistocardiographic record), RMSSD (a value expressed as the square root of the average of the sum of the square roots of the differences in adjacent HRV intervals), and PNN50 (the ratio of the number of intervals in which the difference in consecutive RR intervals exceeds 50 ms), it is possible to estimate blood glucose and glycated hemoglobin levels more accurately than in the past, thereby significantly increasing the accuracy.

[0056] That is, in the past, when estimating the glycated hemoglobin level, the accuracy of measuring glycated hemoglobin was bound to be low because various variables were not taken into consideration because it was simply estimated using blood oxygen saturation. However, in the present invention, the accuracy of measurement is increased by setting the indicators of changes in heart rate and blood pressure as variables by measuring the variability with the changing blood pressure signal as well as the indicators of the quality of the blood oxygen saturation and blood pressure signal while using the indicators of the standard deviation of the normal RR (heartbeat cycle per minute) interval in the entire ballistocardiography record as a standard, while basically measuring the variability with the changing blood pressure signal.

[0057] SDNN is a measure of heart rate variability and represents the standard deviation of heart rate cycle (RR) interval data over a one-minute period. SDNN measures heart rate cycle variability over a full minute, representing the standard deviation of normal RR intervals across all recordings detected by the PPG sensor over that one-minute period.

[0058] A decrease in SDNN may predict an increased risk of left ventricular dysfunction and ventricular tachycardia.

[0059] RMSSD is one of the measures of heart rate variability, and is the average of the square roots of the differences between consecutive heartbeat cycles (RR intervals).

[0060] The higher the RMSSD value, the healthier the condition.

[0061] PNN50 is expressed as the percentage of RR intervals in which the difference between consecutive RR intervals exceeds 50 ms.

[0062] The smaller the PNN50 value, the healthier the condition.

[0063] The average value of the detected biometric information, excluding age and gender, is calculated, and each calculated average value is input into the machine learning algorithm.

[0064] The above machine learning algorithm is an algorithm that allows a computer to learn on its own through input data, and the present invention uses a machine learning algorithm that can classify certain data into one of two types.

[0065] In detail, the above machine learning algorithm can use classification technique algorithms such as regression analysis, logistic regression analysis, decision tree, Bayesian classification, artificial neural network, SVM (Supporter Vector Machine), and K-nearest neighbor.

[0066] In the present invention, 10,000 sets of learning biometric data were collected from diabetic patients and normal people, and the artificial neural network and logistic regression analysis algorithm were trained, respectively.

[0067] The above artificial neural network and logistic regression analysis algorithm ultimately output blood sugar levels and glycated hemoglobin levels, respectively, and calculate the disease prediction probability and determine the alert level based on the output values.

[0068] The above warning level can be classified into “safety”, “concern”, “caution”, and “suspicion”, and when the disease prediction probability is 0% to 10%, it is judged as “safety”, when the disease prediction probability is 11% to 30%, it is judged as “concern”, when the disease prediction probability is 31% to 50%, it is judged as “caution”, and when the disease prediction probability is 51% or more, it is judged as “suspicion” and output.

[0069] Through these steps, the onset of the subject's disease can be diagnosed, and the output results can be monitored on the terminal screen.

Claims

1. A step of receiving a pulse signal of a subject measured through a bio-signal measuring device (100); A step of extracting bio-information from the pulse signal, calculating each average, calculating blood sugar levels and glycated hemoglobin levels, and inputting the calculated average into a pre-learned machine learning algorithm to calculate the disease prediction probability of the subject; A step of outputting an alarm level by determining whether the calculated disease prediction probability falls within a set threshold range; A method for measuring blood glucose and glycated hemoglobin using artificial intelligence technology, characterized in that the above bio-information includes the age and sex of the measurer, HR (heart rate), HRV (heart rate variability), PI (an index related to the oxygen saturation of blood pressure and the quality of blood pressure signals), AC (an index representing changes in heart rate and blood pressure by measuring variability with changing blood pressure signals), SDNN (standard deviation of normal RR (heartbeat cycles per minute) intervals in the entire ballistocardiography record), RMSSD (a value expressed as the square root of the average of the sum of the square roots of the differences between adjacent HRV intervals), PNN50 (a ratio of the number of intervals in which the difference between consecutive RR intervals exceeds 50 ms), LF (low frequency in the range of 0.04 to 0.15 Hz), HF (high frequency in the range of 0.15 to 0.4 Hz), and HF_LF (a frequency in the range of 0.5 to 2 MHz obtained by dividing LF by HF).

2. In paragraph 1, A blood sugar and glycated hemoglobin measurement method using artificial intelligence technology, characterized in that it further includes a step of removing noise from collected pulse signals before calculating the blood sugar level and glycated hemoglobin level.

3. In paragraph 1, The above alert levels are categorized as Safe, Concern, Caution, and Suspicion. A blood sugar and glycated hemoglobin measurement method using artificial intelligence technology, characterized in that when the disease prediction probability is 0% to 10%, it is judged as "safe", when the disease prediction probability is 11% to 30%, it is judged as "concern", when the disease prediction probability is 31% to 50%, it is judged as "caution", and when the disease prediction probability is 51% or more, it is judged as "suspicion" and output.

4. In paragraph 1, A method for measuring blood sugar and glycated hemoglobin using artificial intelligence technology, wherein the machine learning algorithm is an artificial neural network (ANN) and a logistic regression analysis algorithm.

5. A biosignal measuring device (100) for acquiring a pulse signal of a subject; It consists of a terminal (200) that is connected to the bio-signal measuring device (100) by wire or wirelessly, receives pulse signals acquired from the bio-signal measuring device, extracts bio-information, estimates blood sugar levels and glycated hemoglobin levels, and stores a machine learning algorithm that predicts diseases, or is connected to a server (210) where the machine learning algorithm is stored via a wireless communication network; The above biosignal measuring device (100) is formed with an outer surface that is curved convexly from top to bottom so that the measurer's hand can be raised and held, a display portion (111) is formed on the upper side to display measurement information, and a control portion (112) is formed on the lower side of one side. A finger mounting groove (120) formed at a certain depth on the upper outer surface of the measuring instrument body (110) so that the index finger of the measuring instrument can be inserted and mounted; A PPG sensor (130) installed so that the inner surface of the index finger of the measurer is in contact with the inner surface of the finger mounting groove (120) to detect the pulse signal of the measurer; A blood sugar and glycated hemoglobin measuring device using artificial intelligence technology, characterized in that it includes a control unit (140) that receives a pulse signal detected from a PPG sensor (130) inside the measuring device body (110) and transmits the pulse information to the terminal (200) through wired or wireless communication.

Images