US-12616423-B2 - Apparatus and method for estimating physiological variables

Abstract

An apparatus for estimating physiological variables includes: a sensor configured to measure a bio-signal from an object; and a processor. The processor is configured to: configure input data for use in a neural network-based physiological variable estimation model based on bio-signal data measured at different times by the sensor, and input the configured input data to the physiological variable estimation model to obtain a variation in physiological variables over a time period from a calibration time to a current time.

Inventors

• Dae Geun Jang

Assignees

• SAMSUNG ELECTRONICS CO., LTD.

Dates

Publication Date

20260505

Application Date

20230314

Priority Date

20221007

Claims (3)

1. 1 . An apparatus for estimating blood pressure, the apparatus comprising: a sensor configured to measure a bio-signal a plurality of times from an object; a memory configured to store the measured bio-signals; a processor comprising a neural network-based blood pressure estimation model configured to: collect the bio-signal measured at a current time and additional bio-signals measured at least one or more times at a calibration time from the memory, as input data, input the input data into the neural network-based blood pressure estimation model to obtain a blood pressure variation over a time period from the calibration time to the current time, and estimate blood pressure by combining the obtained blood pressure variation and a blood pressure at the calibration time, and a display configured to display the estimated blood pressure and warning information based on the estimated blood pressure, wherein the processor is configured to adjust each of the bio-signals to be equal in length by normalizing the bio-signals by (i) adding a predetermined padding value to each of the bio-signals having a shorter length in comparison to a predetermined length, and (ii) cutting a predetermined region of bio-signals having a longer length in comparison to the predetermined length, wherein the blood pressure variation includes at least one of a difference and a ratio between the blood pressure at the calibration time and a blood pressure at the current time.
2. 2 . The apparatus of claim 1 , wherein the bio-signal comprises at least one of photoplethysmogram (PPG), Electrocardiography (ECG), Electromyography (EMG), impedance plethysmogram (IPG), Pressure wave, video plethysmogram (VPG), Speckle-plethysmogram (SPG), Magnetic-plethysmograph (MPG), Ballistocardiogram (BCG), or Seismocardiogram (SCG).
3. 3 . The apparatus of claim 1 , wherein the processor is further configured to collect user information as the input data.

Description

CROSS-REFERENCE TO RELATED APPLICATION(S) This application claims priority from Korean Patent Application No. 10-2022-0128917, filed on Oct. 7, 2022, in the Korean Intellectual Property Office, the entire disclosure of which is incorporated herein by reference for all purposes. BACKGROUND 1. Field The following description relates to technology for configuring input data for a physiological variable estimation model based on bio-signals measured non-invasively, and for estimating physiological variables by using the input data. 2. Description of Related Art Research on information technology (IT)—medical convergence technology, in which IT and medical technology are combined, is being recently carried out to address medical challenges such as the aging population structure, rapid increase in medical expenses, and shortage of specialized medical service personnel. Monitoring of the health condition of the human body may not be limited to a fixed place, such as a hospital, but may expand to include a mobile healthcare sector for monitoring a user's health condition at any time and any place in daily life at home and office. Electrocardiography (ECG), photoplethysmogram (PPG), and electromyography (EMG) signals are examples of bio-signals that may indicate an individual's health condition. A variety of signal sensors are being developed to measure such signals in daily life. Particularly, by using a PPG sensor, it is possible to estimate blood pressure of the human body by analyzing the shape of pulse wave that reflects cardiovascular status. SUMMARY An apparatus for estimating physiological variables may include: a sensor configured to measure a bio-signal from an object; and a processor. The processor may be configured to: configure input data for use in a neural network-based physiological variable estimation model based on bio-signal data measured at different times by the sensor, and input the configured input data to the physiological variable estimation model to obtain a variation in physiological variables over a time period from a calibration time to a current time. The apparatus may further include a memory configured to store the bio-signal data corresponding to the calibration time. When the sensor measures a bio-signal at the current time for estimating the physiological variables, the processor may be configured to collect the bio-signal data corresponding to the calibration time from the memory. The bio-signal may include at least one of photoplethysmogram (PPG), Electrocardiography (ECG), Electromyography (EMG), impedance plethysmogram (IPG), Pressure wave, video plethysmogram (VPG), Speckle-plethysmogram (SPG), Magnetic-plethysmograph (MPG), Ballistocardiogram (BCG), or Seismocardiogram (SCG). The bio-signal data may include at least one of: raw data of the bio-signal measured by the sensor, preprocessed data obtained by preprocessing the raw data, representative waveform data extracted from the raw data or the preprocessed data, multi-dimensional data converted from the raw data or the preprocessed data, or feature data extracted from the raw data or the preprocessed data, the feature data being associated with the physiological variables. The processor being configured to configure input data may include being configured to use the bio-signal data to configure the input data as multi-channel input data to be input to an input layer of the neural network-based physiological variable estimation model. In response to the bio-signal data differing in length, the processor may be configured to adjust each of the bio-signal data to be equal in length by normalizing the bio-signal data by (i) adding a predetermined padding value to each of the bio-signal data having a shorter length in comparison to a predetermined length, and (ii) cutting a predetermined region of bio-signal data having a longer length in comparison to the predetermined length. The processor being configured to configure the input data may include being configured to: extract a feature associated with the physiological variables from the measured bio-signal, normalize the extracted feature associated with the physiological variables based on a calibration feature at the calibration time, and configure the input data based further on the normalized value. The processor being configured to configure the input data may be based further on at least one of a physiological variable value at the calibration time or user information. The variation in physiological variables may include at least one of a difference and a ratio between the physiological variable value at the calibration time or a physiological variable value at a current time. Based on the obtained variation in physiological variables, the processor may be configured to estimate the physiological variables including at least one of blood pressure, blood glucose, body temperature, antioxidant level, or triglyceride level. The processor may be configured to estimate the physi