Systems and methods of analyte measurement analysis
Disclosed are systems for non-invasively determining a measurement of an analyte. The systems include an electrocardiogram sensor and a processing device operatively coupled to the electrocardiogram sensor. The processing device can execute instructions to receive electrocardiogram data from the electrocardiogram sensor and apply a machine learning model, wherein the machine learning model has been trained based on previous electrocardiogram data associated with a subject and source of an analyte measurement associated with the subject. The system may also determine an indication of a level of the analyte based on the electrocardiogram data.
1. A system comprising:
an electrocardiogram sensor; and
at least a processing device operatively coupled to the electrocardiogram sensor, wherein the at least a processing device is configured to:
train an untrained machine learning model, wherein training the untrained machine learning model comprises:
receiving previous electrocardiogram data of a user from the electrocardiogram sensor;
receiving a first measurement of a level of an analyte within the user at a first time;
generating a set of estimated values of the level of the analyte in the user over a time period subsequent to the first time based in part on the level of the analyte within the user at the first time; and
training the untrained machine learning model using the previous electrocardiogram data and one or more of the set of estimated values;
receive current electrocardiogram data of the user from the electrocardiogram sensor;
determine an indication of a current level of the analyte in the user based on the current electrocardiogram data using the trained machine learning model, wherein the current electrocardiogram data is provided to the trained machine learning model as an input to output the current level of the analyte in the user; and
generate an alert to the user and a recommendation to contact a doctor or physician if the indication of the current level of the analyte in the user is above or below a threshold, wherein the at least a processor is further configured to, if the indication of the current level of the analyte in the user is above or below the threshold, provide a request to the user to retest the current electrocardiogram data.
2. The system of claim 1 , wherein to train the untrained machine learning model, the at least a processing device is configured to:
input an interval of electrocardiogram data of the previous electrocardiogram data received from the electrocardiogram sensor into the untrained machine learning model, wherein the interval of electrocardiogram data was measured by the electrocardiogram sensor during the time period subsequent to the first time; and
update the untrained machine learning model based on a comparison of an output of the untrained machine learning model based on the interval of electrocardiogram data and the one or more estimated values of the set of estimated values of the level of the analyte in the user.
3. The system of claim 1 , wherein the at least a processing device generates the set of estimated values of the level of the analyte in the user based further in part on data about an action that causes the level of the analyte in the user during the time period subsequent to the first time to vary in a predictable manner.
4. The system of claim 3 , wherein the data comprises information about variance in the level of the analyte in the user after administration to the user of a drug that accumulates in a bloodstream of the user over time.
5. The system of claim 3 , wherein the data comprises information about variance in the level of the analyte in the user while performing dialysis on the user.
6. The system of claim 1 , wherein to train the untrained machine learning model, the at least a processing device is further configured to:
receive a second measurement of the level of the analyte within the user at a second time.
7. The system of claim 6 , wherein the at least a processing device generates the set of estimated values of the level of the analyte in the user based on the level of the analyte within the user at the first time and the level of the analyte within the user at the second time.
8. The system of claim 1 , wherein the at least a processor is further configured to generate additional alerts to other individuals if the indication of the current level of the analyte in the user is above or below the threshold.
9. The system of claim 1 , wherein the at least a processor is further configured to pre-process multiple electrocardiogram signals to form the previous electrocardiogram data that is used to train the untrained machine learning model.
10. The system of claim of claim 9 , wherein the at least a processor is configured to pre-process the multiple electrocardiogram signals by:
averaging multiple electrocardiogram waveforms to form one or more sets of single averaged electrocardiogram waveforms; and
performing smoothing and/or noise reduction on the one or more single averaged electrocardiogram waveforms to generate one or more inputs that are used to train the untrained machine learning model.
11. A method comprising:
training, by a least a processing device, an untrained machine learning model, wherein training the untrained machine learning model comprises:
receiving previous electrocardiogram data of a user from an electrocardiogram sensor;
receiving a first measurement of a level of an analyte within the user at a first time;
generating a set of estimated values of the level of the analyte in the user over a time period subsequent to the first time based in part on the level of the analyte within the user at the first time; and
training the untrained machine learning model using the previous electrocardiogram data and one or more of the set of estimated values;
receiving, by the at least a processor, current electrocardiogram data of the user from the electrocardiogram sensor;
determining, by the at least a processor, an indication of a current level of the analyte in the user based on the current electrocardiogram data using the trained machine learning model, wherein the current electrocardiogram data is provided to the trained machine learning model as an input to output the current level of the analyte in the user; and
generating, by the at least a processing device, an alert to the user and a recommendation to contact a doctor or physician if the indication of the current level of the analyte in the user is above or below a threshold, wherein the at least a processor is further configured to, if the indication of the current level of the analyte in the user is above or below the threshold, provide a request to the user to retest the current electrocardiogram data.
12. The method of claim 11 , wherein training the untrained machine learning model comprises:
inputting an interval of electrocardiogram data of the previous electrocardiogram data received from the electrocardiogram sensor into the untrained machine learning model, wherein the interval of electrocardiogram data was measured by the electrocardiogram sensor during the time period subsequent to the first time; and
updating the untrained machine learning model based on a comparison of an output of the untrained machine learning model based on the interval of electrocardiogram data and the one or more estimated values of the set of estimated values of the level of the analyte in the user.
13. The method of claim 11 , wherein the set of estimated values of the level of the analyte in the user is generated based further in part on data about an action that causes the level of the analyte in the user during the time period subsequent to the first time to vary in a predictable manner.
14. The method of claim 13 , wherein the data comprises information about variance in the level of the analyte in the user after administration to the user of a drug that accumulates in a bloodstream of the user over time.
15. The method of claim 13 , wherein the data comprises information about variance in the level of the analyte in the user while performing dialysis on the user.
16. The method of claim 8 , wherein training the untrained machine learning model further comprises:
receiving a second measurement of the level of the analyte within the user at a second time.
17. The method of claim 16 , wherein the set of estimated values of the level of the analyte in the user is generated based on the level of the analyte within the user at the first time and the level of the analyte within the user at the second time.
18. A non-transitory computer-readable medium having instructions stored thereon which, when executed by at least a processing device, cause the at least a processing device to:
train an untrained machine learning model, wherein training the untrained machine learning model comprises:
receiving previous electrocardiogram data of a user from an electrocardiogram sensor; and
receiving a first measurement of a level of an analyte within the user at a first time;
generating a set of estimated values of the level of the analyte in the user over a time period subsequent to the first time based in part on the level of the analyte within the user at the first time; and
training the untrained machine learning model using the previous electrocardiogram data and one or more of the set of estimated values;
receive current electrocardiogram data of the user from the electrocardiogram sensor;
determine an indication of a current level of the analyte in the user based on the current electrocardiogram data using the trained machine learning model, wherein the current electrocardiogram data is provided to the trained machine learning model as an input to output the current level of the analyte in the user; and
generate an alert to the user and a recommendation to contact a doctor or physician if the indication of the current level of the analyte in the user is above or below a threshold, wherein if the indication of the current level of the analyte in the user is above or below the threshold, provide a request to the user to retest the current electrocardiogram data.