Granted Patent B2

US 12,201,443 · App. 17/549,293 · Granted Jan 21, 2025

Inventors: Dimitrios Mavroeidis (Utrecht, NL); Ulf Grossekathoefer (Eindhoven, NL); Aki Sakari Härmä (Eindhoven, NL)

Assignee: KONINKLIJKE PHILIPS N.V.

A61B5/4812A61B5/38A61B5/4809A61B5/7221A61B5/7267

Abstract

Claims (24)

1. A computer-implemented method for determining a sleep state of a user in a system that includes one or more physiological sensors comprising one or more of a polysomnogram (PSG) sensor, an electroencephalogram (EEG) sensor, or an electrocardiogram (ECG) sensor, one or more peripheral devices comprising one or more of an inertia sensor, a temperature sensor, a strain sensor or a skin conductance sensor, and a processor configured to implement a deep neural network having a plurality of layers comprising convolutional layers or recurrent layers, an output layer, and a detection model developed for detection of conditions or sources of signal detection in signals generated by the one or more physiological sensors that employs a time-varying vector-valued sensor uncertainty function, the method comprising:

receiving one or more physiological signals from the one or more physiological sensors used by the user;

determining the sleep state of the user based on the received one or more physiological signals and without data from the one or more peripheral sensors using the deep neural network, wherein the output layer generates the determined sleep state using a softmax function;

calculating a reliability value indicative of a reliability of the determination of the sleep state using the deep neural network based on the received one or more physiological signals and without an output from the one or more peripheral sensors, the reliability value being calculated based on a comparison of the received one or more physiological signals with one or more historic physiological signals each associated with a historic user and a historic sleep state previously determined for the historic user that is the same as the determined sleep state, wherein the output layer generates the reliability value using the softmax function and a Bayesian deep leaning method;

responsive to determining that the reliability value is lower than a first threshold value, preventing output of the sleep state determination;

responsive to determining that the reliability value is greater than a second threshold value that is greater than the first threshold value, outputting the sleep state determination; and

responsive to determining that the first reliability value is greater than or equal to the first threshold value and less than the second threshold value, requesting and receiving data from the one or more peripheral devices, determining an updated sleep state of the user and an updated reliability value for the updated sleep state using the detection model and the time-varying vector-valued sensor uncertainty function of the deep neural network based on the received one or more physiological signals and the received data from the one or more peripheral devices, and outputting the updated sleep state determination only if the updated reliability value is greater than the second threshold value.

2. The method according to claim 1 , further comprising storing determination results for the sleep state determination or the updated sleep state determination and training the neural network using the stored determination results.

3. The method according to claim 1 , further comprising causing auditory stimulation of the user based on the sleep state determination.

4. A system for determining a sleep state of a user, comprising:

one or more physiological sensors comprising one or more of a polysomnogram (PSG) sensor, an electroencephalogram (EEG) sensor, or an electrocardiogram (ECG) sensor;

one or more peripheral devices comprising one or more of an inertia sensor, a temperature sensor, a strain sensor or a skin conductance sensor; and

a processor configured to implement a deep neural network having a plurality of layers comprising convolutional layers or recurrent layers, an output layer, and a detection model developed for detection of conditions or sources of signal detection in signals generated by the one or more physiological sensors that employs a time-varying vector-valued sensor uncertainty function, the processor being further configured to:

receive one or more physiological signals from the one or more physiological sensors used by the user;

determine the sleep state of the user based on the received one or more physiological signals and without data from the one or more peripheral sensors using the deep neural network, wherein the output layer generates the determined sleep state using a softmax function;

calculate a reliability value indicative of a reliability of the determination of the sleep state using the deep neural network based on the received one or more physiological signals and without an output from the one or more peripheral sensors, the reliability value being calculated based on a comparison of the received one or more physiological signals with one or more historic physiological signals each associated with a historic user and a historic sleep state previously determined for the historic user that is the same as the determined sleep state, wherein the output layer generates the reliability value using the softmax function and a Bayesian deep leaning method;

determine whether the reliability value is lower than a first threshold value;

responsive to determining that the reliability value is lower than the first threshold value, prevent output of the sleep state determination;

determine whether the reliability value is greater than a second threshold value that is greater than the first threshold value;

responsive to determining that the reliability value is greater than a second threshold value, output the sleep state determination; and

determine whether the first reliability value is greater than or equal to the first threshold value and less than the second threshold value;

responsive to determining that the first reliability value is greater than or equal to the first threshold value and less than the second threshold value, request and receive data from the one or more peripheral devices, determine an updated sleep state of the user and an updated reliability value for the updated sleep state using the detection model and the time-varying vector-valued sensor uncertainty function of the deep neural network based on the received one or more physiological signals and the received data from the one or more peripheral devices, and output the updated sleep state determination only if the updated reliability value is greater than the second threshold value.

5. The system according to claim 4 , wherein the processor is adapted to store determination results for the sleep state determination or the updated sleep state determination and train the neural network using the stored determination results.

6. A non-transitory computer-readable medium storing a program which, when executed on a computer, is configured to cause the computer to perform a method according to claim 1 in a system according to claim 1 .

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (21)