Determining blood pressure using photoplethysmography (PPG)
A method for determining physiological characteristics is provided. The method involves causing light to be emitted by one or more light emitters toward a tissue of a user. The method involves obtaining samples of transmitted and/or reflected light. The method involves determining photoplethysmography (PPG) data based on the obtained samples. The method involves providing the PPG data as an input to a trained machine learning model, wherein the trained machine learning model has been trained, using a training set that comprises matched samples of electrocardiogram (ECG) data and PPG data. The method involves providing a blood pressure of the user based on an output of the trained machine learning model.
1 . A method for determining physiological characteristics, comprising:
causing light to be emitted by one or more light emitters toward a tissue of a user;
obtaining samples of transmitted and/or reflected light;
determining photoplethysmography (PPG) data based on the obtained samples;
providing the PPG data as an input to a single trained machine learning model, wherein the trained machine learning model has been trained using a training set that comprises matched samples of electrocardiogram (ECG) data and PPG data, wherein the matched samples comprise PPG data labeled, based at least in part on characteristics of the ECG data, to indicate timepoints associated with features relevant to blood pressure, such that the machine learning model is trained to identify features of the PPG data relevant to blood pressure; and
providing a blood pressure of the user based on an output of the trained machine learning model.
2 . The method of claim 1 , wherein the matched samples of ECG data and PPG data that were included in the training set comprise PPG data that has been labeled, using the ECG data, to indicate boundaries of pressure waveforms.
3 . The method of claim 1 , wherein motion data was provided as an input to the trained machine learning model.
4 . The method of claim 1 , wherein at least one item of demographic information associated with the user was provided as an input to the trained machine learning model.
5 . The method of claim 1 , wherein recent activity data associated with the user was provided as an input to the trained learning model.
6 . The method of claim 1 , wherein:
the one or more light emitters includes one or more light emitters of a wearable device worn by the user; and
the samples of transmitted and/or reflected light are obtained at one or more light sensors of the wearable device.
7 . The method of claim 1 , wherein providing the blood pressure of the user includes sending an instruction to present the blood pressure to the user at a wearable device worn by the user.
8 . The method of claim 1 , wherein:
the trained machine learning model is trained to identify features of the PPG data relevant to a systolic blood pressure and features of the PPG data relevant to a diastolic blood pressure; and
providing the blood pressure of the user based on the output of the trained machine learning model includes the systolic blood pressure and the diastolic blood pressure.
9 . The method of claim 1 , wherein the matched samples of ECG data and PPG data comprise data from multiple people other than the user.
10 . A system for determining physiological characteristics, the system comprising:
one or more processors configured to:
cause light to be emitted by one or more light emitters toward a tissue of a user;
obtain samples of transmitted and/or reflected light;
determine photoplethysmography (PPG) data based on the obtained samples;
provide the PPG data as an input to a single trained machine learning model, wherein the trained machine learning model has been trained, using a training set that comprises matched samples of electrocardiogram (ECG) data and PPG data, wherein the matched samples comprise PPG data labeled, based at least in part on characteristics of the ECG data, to indicate timepoints associated with features relevant to blood pressure, such that the machine learning model is trained to identify features of the PPG data relevant to blood pressure; and
provide a blood pressure of the user based on an output of the trained machine learning model.
11 . The system of claim 10 , wherein the matched samples of ECG data and PPG data that were included in the training set comprise PPG data that has been labeled, using the ECG data, to indicate boundaries of pressure waveforms.
12 . The system of claim 10 , wherein motion data was provided as an input to the trained machine learning model.
13 . The system of claim 10 , wherein at least one item of demographic information associated with the user was provided as an input to the trained machine learning model.
14 . The system of claim 10 , wherein:
the one or more light emitters includes one or more light emitters of a wearable device worn by the user; and
the samples of transmitted and/or reflected light are obtained at one or more light sensors of the wearable device.
15 . The system of claim 10 , wherein the matched samples of ECG data and PPG data comprise data from multiple people other than the user.
16 . A smart watch including one or more processors, the one or more processors configured to:
cause light to be emitted by one or more light emitters of the smart watch toward a tissue of a user;
obtain samples of transmitted and/or reflected light at one or more light sensors of the smart watch;
determine photoplethysmography (PPG) data based on the obtained samples;
provide the PPG data as an input to a single trained machine learning model, wherein the trained machine learning model has been trained, using a training set that comprises matched samples of electrocardiogram (ECG) data and PPG data, wherein the matched samples comprise PPG data labeled, based at least in part on characteristics of the ECG data, to indicate timepoints associated with features relevant to blood pressure, such that the machine learning model is trained to identify features of the PPG data relevant to blood pressure; and
provide a blood pressure of the user based on an output of the trained machine learning model.
17 . The smart watch of claim 16 , wherein the matched samples of ECG data and PPG data that were included in the training set comprise PPG data that has been labeled, using the ECG data, to indicate boundaries of pressure waveforms.
18 . The smart watch of claim 16 , wherein motion data was provided as an input to the trained machine learning model.
19 . The smart watch of claim 16 , wherein at least one item of demographic information associated with the user was provided as an input to the trained machine learning model.
20 . The smart watch of claim 16 , wherein providing the blood pressure of the user includes sending an instruction to present the blood pressure to the user at the smart watch.
21 . The smart watch of claim 16 , wherein the matched samples of ECG data and PPG data comprise data from multiple people other than the user.
22 . A non-transitory computer-readable storage medium including executable instructions that, when executed by one or more processors, cause the one or more processors to:
cause light to be emitted by one or more light emitters toward a tissue of a user;
obtain samples of transmitted and/or reflected light;
determine photoplethysmography (PPG) data based on the obtained samples;
provide the PPG data as an input to a single trained machine learning model, wherein the trained machine learning model has been trained, using a training set that comprises matched samples of electrocardiogram (ECG) data and PPG data, wherein the matched samples comprise PPG data labeled, based at least in part on characteristics of the ECG data, to indicate timepoints associated with features relevant to blood pressure, such that the machine learning model is trained to identify features of the PPG data relevant to blood pressure; and
provide a blood pressure of the user based on an output of the trained machine learning model.
23 . The non-transitory computer-readable storage medium of claim 22 , wherein the matched samples of ECG data and PPG data that were included in the training set comprise PPG data that has been labeled, using the ECG data, to indicate boundaries of pressure waveforms.
24 . The non-transitory computer-readable storage medium of claim 22 , wherein motion data was provided as an input to the trained machine learning model.
25 . The non-transitory computer-readable storage medium of claim 22 , at least one item of demographic information associated with the user was provided as an input to the trained machine learning model.
26 . The non-transitory computer-readable storage medium of claim 22 , wherein:
the one or more light emitters includes one or more light emitters of a wearable device worn by the user; and
the samples of transmitted and/or reflected light are obtained at one or more light sensors of the wearable device.
27 . The non-transitory computer-readable storage medium of claim 22 , wherein the matched samples of ECG data and PPG data comprise data from multiple people other than the user.