Real-time meal detection based on sensor glucose and estimated plasma insulin levels

1. A system for automatic real-time meal detection, the system comprising:

one or more processors; and

one or more processor-readable media storing instructions which, when executed by the one or more processors, cause performance of one or more operations including:

obtaining a plurality of glucose concentration values and a plurality of plasma insulin concentration estimations, each glucose concentration value of the plurality of glucose concentration values corresponding in time with a respective plasma insulin concentration estimation of the plurality of plasma insulin concentration estimations to form a plurality of time-corresponding pairs of glucose concentration values and plasma insulin concentration estimations;

generating an output based on applying a meal detection model to the plurality of glucose concentration values and the plurality of plasma insulin concentration estimations; and

determining, based on the output, that a glucose concentration value of the plurality of glucose concentration values corresponds to an ongoing glycemic response to a meal.

2. The system of claim 1 , wherein the instructions further cause performance of one or more operations including causing delivery of insulin to at least partially counteract the ongoing glycemic response to the meal.

3. The system of claim 2 , wherein the causing the delivery of insulin comprises communicating one or more signals to an insulin delivery device.

4. The system of claim 2 , wherein the causing the delivery of insulin comprises temporarily decreasing a target glucose value for a closed-loop mode of an insulin delivery device, the closed-loop mode configured for automated delivery of the insulin based on the target glucose value.

5. The system of claim 2 , wherein the causing the delivery of insulin comprises communicating one or more signals to an actuator associated with an insulin reservoir.

6. The system of claim 2 , wherein the delivered insulin comprises a meal bolus.

7. The system of claim 2 , wherein the delivered insulin comprises one or more correction boluses constituting at least part of a meal bolus.

8. The system of claim 1 , wherein the instructions further cause performance of one or more operations including causing generation of a notification indicative of the meal detection.

9. The system of claim 1 , wherein the plurality of time-corresponding pairs are provided to the meal detection model as a first input vector and a second input vector, the first input vector comprising at least a portion of the plurality of glucose concentration values, the second input vector comprising at least a portion of the plurality of plasma insulin concentration estimations, each of the first input vector and the second input vector including a predetermined number of past data points within a moving window of time.

10. The system of claim 9 , wherein the moving window of time is sized to include a fasting period and to exclude a period of postprandial glucose level increase prior to the fasting period.

11. The system of claim 1 , wherein determining that the glucose concentration value of the plurality of glucose concentration values corresponds to the ongoing glycemic response to the meal comprises:

identifying an earliest glucose concentration value that was identified, based on applying the meal detection model to the plurality of glucose concentration values, as being part of the ongoing glycemic response to the meal;

evaluating backward in time from the earliest glucose concentration value to identify an onset of an increasing glucose trend within three glucose concentration values from the earliest glucose concentration value identified; and

determining that the onset of the increasing glucose trend corresponds to an onset of the ongoing glycemic response to the meal.

12. The system of claim 1 , wherein the obtaining the plasma insulin concentration estimations comprises applying an insulin absorption model to insulin delivered exogenously to a user.

13. The system of claim 1 , wherein generating the output comprises:

obtaining a probability that at least a subset of the plurality of glucose concentration values corresponds to the ongoing glycemic response to the meal;

comparing the probability to a probability threshold; and

making a binary determination based on the comparing.

14. The system of claim 1 , wherein:

the meal detection model comprises a machine learning model operable as a classifier; and

the classifier comprises a logistic regression model or a long short-term memory (LSTM) recurrent neural network model.

15. The system of claim 1 , wherein the obtaining the plurality of plasma insulin concentration estimations comprises normalizing the plurality of plasma insulin concentration estimations based on a median insulin total daily dose.

16. The system of claim 1 , wherein the obtaining the plurality of glucose concentration values comprises, responsive to determining that a threshold number of the plurality of glucose concentration values are valid, replacing at least one invalid glucose concentration value with at least one valid or estimated glucose concentration value.

17. The system of claim 1 , wherein each glucose concentration value is a blood glucose concentration estimation based on an interstitial glucose measurement.

18. A processor-implemented method comprising:

obtaining a plurality of glucose concentration values and a plurality of plasma insulin concentration estimations, each glucose concentration value corresponding in time with a respective plasma insulin concentration estimation to form a plurality of time-corresponding pairs of glucose concentration values and plasma insulin concentration estimations;

generating an output based on applying a meal detection model to the plurality of glucose concentration values and the plurality of plasma insulin concentration estimations; and

determining, based on the output, that a glucose concentration value of the plurality of glucose concentration values corresponds to an ongoing glycemic response to a meal.

19. The method of claim 18 , further comprising causing delivery of insulin to at least partially counteract the ongoing glycemic response to the meal.

20. One or more processor-readable media storing instructions which, when executed by one or more processors, cause performance of:

obtaining a plurality of glucose concentration values and a plurality of plasma insulin concentration estimations, each glucose concentration value corresponding in time with a respective plasma insulin concentration estimation to form a plurality of time-corresponding pairs of glucose concentration values and plasma insulin concentration estimations;

generating an output based on applying a meal detection model to the plurality of glucose concentration values and the plurality of plasma insulin concentration estimations; and

determining, based on the output, that a glucose concentration value of the plurality of glucose concentration values corresponds to an ongoing glycemic response to a meal.