Methods and systems for labeling sleep states

Patent metadata

Field	Value
Publication number	US-2022265208-A1
Application number	US-202217690369-A
Country	US
Kind code	A1
Filing date	Mar 9, 2022
Priority date	Sep 6, 2016
Publication date	Aug 25, 2022
Grant date	—

Abstract

A system, computer-readable storage medium, and a method capable of, directly or indirectly, estimating sleep states of a user based on sensor data from movement sensors and/or optical sensors.

First claim

1 - 98 . (canceled)

99 . A computer-implemented method for labeling stages of sleep, the computer-implemented method comprising:
obtaining motion data for a user during a time window in which the user is asleep; obtaining cardiopulmonary data for the user during the time window; processing the motion data and the cardiopulmonary data with a classifier model to generate a label for one or more time periods within the time window, the label indicating the user is awake or in one of a plurality of different stages of sleep during the one or more time periods, wherein the classifier model is trained using sleep stage classifications determined by a human for one or more sleep study subjects and based on motion data for the one or more sleep study subjects and cardiopulmonary data for the one or more sleep study subjects.

100 . The computer-implemented method of claim 99 , wherein:
obtaining motion data for the user includes obtaining the motion data from a first sensor; and obtaining cardiopulmonary data for the user includes obtaining the cardiopulmonary data from a second sensor that is different than the first sensor.

101 . The computer-implemented method of claim 100 , wherein the first sensor includes an accelerometer and the second sensor includes an optical sensor.

102 . The computer-implemented method of claim 101 , wherein the optical sensor includes a photoplethysmogram (PPG) sensor.

103 . The computer-implemented method of claim 99 , further comprising:
extracting a first feature from the motion data for the user; and extracting a second feature from the cardiopulmonary data for the user, wherein processing the motion data for the user and the cardiopulmonary data for the user with the classifier model includes processing the first feature and the second feature to generate the label.

104 . The computer-implemented method of claim 103 , wherein the first feature includes an amount of time that has lapsed since the motion data last indicated a threshold amount of motion by the user.

105 . The computer-implemented method of claim 103 , wherein the second feature includes a difference between a minimum heart rate of the user during a threshold amount of time and a maximum heart rate of the user during the threshold amount of time.

106 . The computer-implemented method of claim 103 , wherein the second feature includes:
a variability of an envelope of inter-beat intervals, a variability of a de-trended respiration rate extracted from the cardiopulmonary data, an inter-percentile spread of a heart rate extracted from the cardiopulmonary data or from the inter-beat intervals, a normalized de-trended heart rate extracted from the cardiopulmonary data or from the inter-beat intervals, or a cross-correlation of each pulse shape of one or more pulse shapes in the cardiopulmonary data with a previous pulse shape in the cardiopulmonary data, wherein the pulse shapes are normalized to a common duration prior to the cross-correlation.

107 . The computer-implemented method of claim 99 , wherein the plurality of different stages of sleep include a light sleep stage, a deep sleep stage, and a rapid eye movement (REM) sleep stage.

108 . The computer-implemented method of claim 99 , wherein the classifier model includes a random forest classifier or a linear discriminant classifier model.

109 . A system for labeling stages of sleep, the system comprising:
a plurality of sensors; and one or more processors configured to perform operations, the operations including:
obtaining, via one or more sensors of the plurality of sensors, motion data for a user during a time window in which the user is asleep;
obtaining, via one or more sensors of the plurality of sensors, cardiopulmonary data for the user during the time window; and
processing the motion data and the cardiopulmonary data with a classifier model to generate a label for one or more time periods within the time window, the label indicating the user is awake or in one of a plurality of different stages of sleep during the one or more time periods,
wherein the classifier model is trained using sleep stage classifications determined by a human for one or more sleep study subjects and based on motion data for the one or more sleep study subjects and cardiopulmonary data for the one or more sleep study subjects.

110 . The system of claim 109 , wherein:
obtaining motion data for the user includes obtaining, via a first sensor of the plurality of sensors, the motion data; and obtaining cardiopulmonary data includes obtaining, via a second sensor of the plurality of sensors, the cardiopulmonary data, the second sensor being different than the first sensor.

111 . The system of claim 110 , wherein:
the first sensor includes an accelerometer; and the second sensor includes an optical sensor.

112 . The system of claim 109 , wherein the operations further include:
extracting a first feature from the motion data for the user; and extracting a second feature from the cardiopulmonary data for the user, wherein processing the motion data for the user and the cardiopulmonary data for the user with the classifier model includes processing the first feature and the second feature to generate the label.

113 . The system of claim 112 , wherein the first feature includes an amount of time that has lapsed since the motion data last indicated a threshold amount of motion by the user.

114 . The system of claim 112 , wherein the second feature includes a difference between a minimum heart rate of the user during a threshold amount of time and a maximum heart rate of the user during the threshold amount of time.

115 . The system of claim 109 , wherein the plurality of different stages of sleep include a light sleep stage, a deep sleep stage, and a rapid eye movement (REM) sleep stage.

116 . The system of claim 109 , wherein the classifier model includes a random forest classifier or a linear discriminant classifier model.

117 . The system of claim 109 , wherein processing the motion data and the cardiopulmonary data using the classifier model includes:
determining a first confidence score for the user being awake during the one or more time periods; and determining a second confidence score for the user being in a first stage of the plurality of different stages of sleep during the one or more time periods; determining a third confidence score for the user being a second stage of the plurality of different sleep stages during the one or more time periods; and labeling the one or more time periods with a label corresponding to the highest of the first confidence score, the second confidence score, and the third confidence score.

Assignees

• Fitbit Inc

Inventors

• Heneghan Conor Joseph
• Arnold Jacob Anthony
• Beattie Zachary Todd
• Pantelopoulos Alexandros A
• Russell Allison Maya
• Foeckler Philip
• Tucker Adrienne M
• Lopez Delisa
• Lafon Belen
• Ghoreyshi Atiyeh

Classifications

• A61B5/7267

  involving training the classification device · CPC title

• A61B5/11

  Measuring movement of the entire body or parts thereof, e.g. head or hand tremor or mobility of a limb {(A61B5/1038 takes precedence; motion detection to correct for motion artifacts in physiological signals A61B5/721)} · CPC title

• A61B2562/0219

  Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches · CPC title

• A61B5/02438

  with portable devices, e.g. worn by the patient · CPC title

• A61B5/02416

  using photoplethysmograph signals, e.g. generated by infrared radiation (A61B5/14552 takes precedence) · CPC title