Determining sleep stages and sleep events using sensor data

• US 10,321,871 B2
• Filed: 08/26/2016
• Issued: 06/18/2019
• Est. Priority Date: 08/28/2015
• Status: Active Grant

• Current Assignee: Awarables, Inc.
• Original Assignee: Awarables, Inc.
• Inventors: Bandyopadhyay, Amrit, Blankenship, Gilmer, Upender, Raghu, Upender, Madhvi, Giles, Chris
• Primary Examiner(s): Abreu, Michael J D
• Application Number:
  US15/249,108
• Publication Number:
  US 20170055898A1
• Time in Patent Office: 1,026 Days
• CPC Class Codes: A61B 2503/12 Healthy persons not otherwise provided for, e.g. subjects of a marketing survey
• CPC Class Codes: A61B 2560/0242 adapted to measure environmental factors, e.g. temperature, pollution
• CPC Class Codes: A61B 2562/0219 Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
• CPC Class Codes: A61B 2562/06 Arrangements of multiple sensors of different types
• CPC Class Codes: A61B 5/0022 Monitoring a patient using a global network, e.g. telephone networks, internet
• CPC Class Codes: A61B 5/0205 Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
• CPC Class Codes: A61B 5/02405 Determining heart rate variability
• CPC Class Codes: A61B 5/02416 using photoplethysmograph signals, e.g. generated by infrared radiation A61B5/14552 takes precedence
• CPC Class Codes: A61B 5/02438 with portable devices, e.g. worn by the patient
• CPC Class Codes: A61B 5/0816 Measuring devices for examining respiratory frequency measuring frequency of electric signals G01R23/00
• CPC Class Codes: A61B 5/0826 Detecting or evaluating apnoea events A61B5/4818 takes precedence
• CPC Class Codes: A61B 5/11 Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb for measuring pulse A61B5/02 ;A61B5/1038 takes precedence; motion detection to correct for motion artifacts in physiological signals A61B5/721
• CPC Class Codes: A61B 5/1118 Determining activity level
• CPC Class Codes: A61B 5/30 Input circuits therefor
• CPC Class Codes: A61B 5/303 Patient cord assembly, e.g. cable harness
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/318 Heart-related electrical modalities, e.g. electrocardiography [ECG]
• CPC Class Codes: A61B 5/352 Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/4064 Evaluating the brain A61B5/031, A61B5/369, A61B5/14553 take precedence
• CPC Class Codes: A61B 5/4806 : Sleep evaluation A61B5/4821 takes precedence; devices for inducing sleep A61M21/02
• CPC Class Codes: A61B 5/4812 : Detecting sleep stages or cycles
• CPC Class Codes: A61B 5/4818 : Sleep apnoea
• CPC Class Codes: A61B 5/486 : Bio-feedback A61B5/375 takes precedence
• CPC Class Codes: A61B 5/6804 : Garments; Clothes
• CPC Class Codes: A61B 5/681 : Wristwatch-type devices
• CPC Class Codes: A61B 5/6823 : Trunk, e.g., chest, back, abdomen, hip
• CPC Class Codes: A61B 5/6898 : Portable consumer electronic devices, e.g. music players, telephones, tablet computers
• CPC Class Codes: A61B 5/7246 : using correlation, e.g. template matching or determination of similarity
• CPC Class Codes: A61B 5/7257 : using Fourier transforms
• CPC Class Codes: A61B 5/7267 : involving training the classification device
• CPC Class Codes: A61B 5/7278 : Artificial waveform generation or derivation, e.g. synthesising signals from measured signals A61B5/327 takes precedence
• CPC Class Codes: A61B 5/742 : using visual displays A61B5/339 takes precedence
• CPC Class Codes: G16H 50/20 : for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining sleep stages and sleep events using sensor data. In some implementations, sensor data is obtained over a time period while a person is sleeping. The time period is divided into a series of intervals. Heart rate and changes in the heart rate are analyzed over the intervals. Based on the analysis of the heart rate changes, sleep stage labels are assigned to different portions of the time period. An indication of the assigned sleep stage labels is provided.

Claims

1. A method performed by one or more computing devices, the method comprising:

obtaining, by the one or more computing devices, sensor data generated by one or more sensors over a time period while a person is sleeping;

dividing, by the one or more computing devices, the time period into a series of intervals;

analyzing, by the one or more computing devices, heart rate and the changes in the heart rate of the person indicated by the sensor data over the intervals, wherein the analysis comprises;

determining approximate entropy measures indicating approximate entropy of a heart rate of the person for each of the intervals, comprising determining a series of heart rate values corresponding to different times within each interval and assessing multiple sliding windows of the heart rate values within each interval;

generating a heart-rate variability (HRV) signal for each of the intervals, performing a frequency analysis of the HRV signals, and generating HRV ratios each indicating a ratio of low frequency components of the HRV signal to high frequency components of the HRV signal; and

determining a relative heart rate measure for each of the intervals, each relative heart rate measure being determined based on multiple different windows over a series of sub-interval epochs within the interval, wherein the relative heart rate measure for an interval is determined based on comparisons, for each particular epoch of at least some of the epochs, of a heart rate measure for the particular epoch with heart rate measures for epochs that occur before and after the particular epoch in the window corresponding to the particular epoch;

obtaining sleep stage likelihood scores determined by multiple different sleep stage analysis functions based on the approximate entropy measures, the HRV ratios, and the relative heart rate measures, comprising;

determining, based on one or more of the approximate entropy measures and one or more of the relative heart rate measures, a first score indicating a likelihood that the sleep stage for a particular portion of the time period is in a set consisting of a REM stage and a wake stage; and

determining a second score to discriminate between the REM stage and the wake stage, the second score being determined based on (i) a measure representing a comparison of a measure of heart rate during the particular portion with a resting reference heart rate for the person and (ii) a level of movement of the person during the particular portion of the time period;

assigning, by the one or more computing devices, sleep stage labels to different portions of the time period based on a combination of the likelihood scores from the multiple different sleep stage analysis functions; and

providing, by the one or more computing devices, an indication of the assigned sleep stage labels.

2. The method of claim 1, wherein obtaining the sensor data, dividing the time period into a series of intervals, and analyzing the heart rate and the changes in the heart rate are performed by a wearable, battery-operated device, wherein the wearable device includes sensors that detect the sensor data.

3. The method of claim 2, further comprising providing, by the wearable device, results of analyzing the heart rate and the changes in the heart rate to a second device for transmission to a server system;

and

wherein assigning the sleep stage labels and providing the indication of the assigned sleep stage labels are performed by the server system.

4. The method of claim 1, further comprising:

obtaining data sets indicating signals or function outputs, the data sets being labeled with sleep stage labels corresponding to the data sets; and

training a sleep stage classifier based on the data sets to produce output indicating likelihoods that input data corresponds to a sleep stage from among a set of sleep stages.

5. The method of claim 4, further comprising:

determining, for each of the signals or function outputs, a signal range histogram for the signal or function output; and

using the signal range histograms to train the sleep stage classifier.

6. The method of claim 1, wherein determining the relative heart rate measure for a particular interval comprises:

determining, for each particular epoch of multiple of the epochs in the particular interval;

an average heart rate for a particular epoch in the particular interval;

a first reference value based on average heart rates for a threshold number of epochs immediately before the particular epoch; and

a second reference value based on average heart rate for a threshold number of epochs immediately following the particular epoch;

determining a first difference between the average heart rate for the particular epoch and the first reference value for the particular epoch;

determining a second difference between the average heart rate for the particular epoch and the second reference value for the particular epoch; and

determining an epoch-level relative heart rate measure for the particular epoch based on the first difference and the second difference; and

determining, as the relative heart rate measure for the particular interval, an interval-level relative heart rate measure as an average of the epoch-level relative heart rate measures for multiple the multiple epochs in the particular interval.

7. The method of claim 6, wherein the first reference value is a minimum value from among average heart rates for multiple epochs in a series of intervals immediately preceding the particular epoch;

wherein the second reference value is a minimum value from among average heart rates for multiple epochs in a series of epochs immediately preceding the particular epoch; and

wherein the epoch-level relative heart rate measure for the particular epoch is an average of the first difference for the particular epoch and the second difference for the particular epoch.

8. The method of claim 1, wherein determining the approximate entropy measures comprises determining one or more approximate entropy factor scores that have a value within a predefined range and indicate a relationship of an approximate entropy distribution with respect to an approximate entropy reference value;

and

wherein determining the relative heart rate measures comprises determining one or more heart rate factor scores that have a value within a predefined range and indicate a relationship of a relative heart rate distribution with respect to a reference value.

9. The method of claim 8, wherein the predefined ranges for the one or more approximate entropy factor scores and the relative heart rate factor scores are a same predefined range;

and

wherein the approximate entropy factor scores and the relative heart rate factor scores are generated using a logistic function.

10. The method of claim 1, wherein determining the first score for a particular interval comprises determining a mean of (i) the relative heart rate measure for the particular interval and (ii) a first approximate entropy factor score indicating a relationship of approximate entropy values for the particular interval with respect to an average of approximate entropy values;

wherein obtaining the sleep stage likelihood scores further comprises determining a third score indicating a likelihood that the particular portion of the time period represents a slow wave sleep stage as;

a score derived using one or more HRV ratios for the particular interval;

or

a mean of (i) the score derived using one or more HRV ratios for the particular interval and (ii) a second approximate entropy factor score indicating a relationship of approximate entropy values with respect to a reference indicating a lowest percentage of approximate entropy values for the particular interval.

11. The method of claim 10, wherein the slow wave sleep score is the mean of the HRV ratio and the second approximate entropy factor score;

wherein the second approximate entropy factor score is determined based on (i) a mean of approximate entropy values for the particular interval from a lowest approximate entropy value up to a particular percentile of the approximate entropy values and (ii) a mean of approximate entropy values for the particular interval from a lowest approximate entropy value up to a percentile of the approximate entropy values that is greater than the particular percentile.

12. The method of claim 1, wherein the sleep stage labels include labels corresponding to stages for wake, rapid eye movement (REM) sleep, light sleep, and deep or slow-wave sleep.

13. The method of claim 1, wherein assigning the sleep stage labels comprises:

determining, for each of the intervals, (i) a likelihood that the interval corresponds to REM sleep, and (ii) an indication whether the interval is classified as REM sleep;

determining, for each of the intervals, (i) a likelihood that the interval corresponds to light sleep, and (ii) an indication whether the interval is classified as light sleep;

determining, for each of the intervals, (i) a likelihood that the interval corresponds to slow-wave sleep, and (ii) an indication whether the interval is classified as slow-wave sleep;

or

determining, for each of the intervals, (i) a likelihood that the interval corresponds to a wake stage, and (ii) an indication whether the interval is classified as a wake stage.

14. The method of claim 1, wherein the sensor data comprises at least one selected from a group consisting of:

(i) EKG signal data from EKG sensor, and (ii) heartbeat data from a heartbeat sensor.

15. The method of claim 1, wherein dividing the time period into a series of intervals comprises diving the time period into adjacent periods each having a same duration.

16. The method of claim 1, wherein analyzing the heart rate of the person comprises evaluating the value of a heart rate signal in each interval with respect to the values in a neighborhood of the interval, wherein the neighborhood for each interval is a time window that extends from a first time threshold that precedes the interval to a second time threshold that follows the interval.

17. The method of claim 1, wherein the obtained sensor data corresponds to a sleep session of the person;

and

wherein analyzing the heart rate of the person comprises evaluating an absolute value of the heart rate signal with respect to the rest of the heart rate values for the sleep session and predetermined thresholds.

18. The method of claim 1, wherein assigning the sleep stage labels comprises assigning a sleep stage label to a particular portion of the time period, comprising:

providing multiple signals that are separately correlated to sleep stages as input to a sleep stage classifier; and

obtaining a single sleep stage label for the particular portion of the time period based on output of the sleep stage classifier.

19. The method of claim 1, comprising detecting heartbeats by:

detecting peaks in an EKG signal;

determining a derivative signal from of the EKG signal by determining, for each sample of the EKG signal, a difference of between a current sample and an immediately previous sample;

identifying R waves based on applying one or more thresholds to the derivative signal; and

detecting the heartbeats based on the identified R waves.

20. The method of claim 1, wherein analyzing the heart rate and the changes in the heart rate comprises determining, for each of the intervals, (i) multiple approximate entropy factors corresponding to mean values of different ranges of the determined approximate entropy scores, (ii) multiple HRV ratio factors corresponding to mean values of different ranges of the determined HRV ratios;

and

wherein obtaining sleep stage likelihood scores comprises determining, for each of the intervals, a probability of one or more sleep stages based on a combination of one of the approximate entropy factors for the interval and one of the HRV ratio factors for the interval.

21. The method of claim 1, wherein analyzing the heart rate and the changes in the heart rate comprises determining an additional heart rate measure for an interval computed as an exponential average the series of heart rate values for the interval minus a median of the series of heart rate values;

and

wherein the sleep stage likelihood scores are obtained based on the additional heart rate measure.

22. The method of claim 1, wherein generating the approximate entropy measures comprises:

counting, for an interval, a number of incidences of sliding windows for which a maximum difference in corresponding elements is less than a comparison threshold; and

averaging logarithmic sums of the counts.

23. The method of claim 1, comprising detecting heartbeats, comprising:

identifying local maxima and local minima in an EKG signal;

computing ranges between the maximum value of the EKG signal and minimum value of the EKG signal within each of multiple windows of the EKG signal; and

assigning, to each of the multiple windows, a probability of the window representing a beat.

24. A system comprising:

one or more computing devices;

one or more computer-readable media storing instructions that, when executed by the one or more computing devices, cause the one or more computing devices to perform operations comprising;

obtaining, by the one or more computing devices, sensor data generated by one or more sensors over a time period while a person is sleeping;

dividing, by the one or more computing devices, the time period into a series of intervals;

analyzing, by the one or more computing devices, heart rate and the changes in the heart rate of the person indicated by the sensor data over the intervals, wherein the analysis comprises;

determining approximate entropy measures indicating approximate entropy of a heart rate of the person for each of the intervals, comprising determining a series of heart rate values corresponding to different times within each interval and assessing multiple sliding windows of the heart rate values within each interval;

generating a heart-rate variability (HRV) signal for each of the intervals, performing a frequency analysis of the HRV signals, and generating HRV ratios each indicating a ratio of low frequency components of the HRV signal to high frequency components of the HRV signal; and

determining a relative heart rate measure for each of the intervals, each relative heart rate measure being determined based on multiple different windows over a series of sub-interval epochs within the interval, wherein the relative heart rate measure for an interval is determined based on comparisons, for each particular epoch of at least some of the epochs, of a heart rate measure for the particular epoch with heart rate measures for epochs that occur before and after the particular epoch in the window corresponding to the particular epoch;

obtaining sleep stage likelihood scores determined by multiple different sleep stage analysis functions based on the approximate entropy measures, the HRV ratios, and the relative heart rate measures, comprising;

determining, based on one or more of the approximate entropy measures and one or more of the relative heart rate measures, a first score indicating a likelihood that the sleep stage for a particular portion of the time period is in a set consisting of a REM stage and a wake stage; and

determining a second score to discriminate between the REM stage and the wake stage, the second score being determined based on (i) a measure representing a comparison of a measure of heart rate during the particular portion with a resting reference heart rate for the person and (ii) a level of movement of the person during the particular portion of the time period;

assigning, by the one or more computing devices, sleep stage labels to different portions of the time period based on a combination of the likelihood scores from the multiple different sleep stage analysis functions; and

providing, by the one or more computing devices, an indication of the assigned sleep stage labels.

25. One or more computer-readable media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

obtaining, by the one or more computing devices, sensor data generated by one or more sensors over a time period while a person is sleeping;

dividing, by the one or more computing devices, the time period into a series of intervals;

analyzing, by the one or more computing devices, heart rate and the changes in the heart rate of the person indicated by the sensor data over the intervals, wherein the analysis comprises;

determining approximate entropy measures indicating approximate entropy of a heart rate of the person for each of the intervals, comprising determining a series of heart rate values corresponding to different times within each interval and assessing multiple sliding windows of the heart rate values within each interval;

generating a heart-rate variability (HRV) signal for each of the intervals, performing a frequency analysis of the HRV signals, and generating HRV ratios each indicating a ratio of low frequency components of the HRV signal to high frequency components of the HRV signal; and

determining a relative heart rate measure for each of the intervals, each relative heart rate measure being determined based on multiple different windows over a series of sub-interval epochs within the interval, wherein the relative heart rate measure for an interval is determined based on comparisons, for each particular epoch of at least some of the epochs, of a heart rate measure for the particular epoch with heart rate measures for epochs that occur before and after the particular epoch in the window corresponding to the particular epoch;

obtaining sleep stage likelihood scores determined by multiple different sleep stage analysis functions based on the approximate entropy measures, the HRV ratios, and the relative heart rate measures, comprising;

determining, based on one or more of the approximate entropy measures and one or more of the relative heart rate measures, a first score indicating a likelihood that the sleep stage for a particular portion of the time period is in a set consisting of a REM stage and a wake stage; and

determining a second score to discriminate between the REM stage and the wake stage, the second score being determined based on (i) a measure representing a comparison of a measure of heart rate during the particular portion with a resting reference heart rate for the person and (ii) a level of movement of the person during the particular portion of the time period;

assigning, by the one or more computing devices, sleep stage labels to different portions of the time period based on a combination of the likelihood scores from the multiple different sleep stage analysis functions; and

providing, by the one or more computing devices, an indication of the assigned sleep stage labels.