Systems and methods for continuous glucose monitoring outcome predictions
Inventors
Iyer, Anand K. • KUMBARA, Abhimanyu • SHOMALI, Mansur E. • Luo, Junjie • GAO, Guodong
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12283381-B2
Publication Date
2025-04-22
Expiration Date
Abstract
Methods and devices include predicting future glucose and engagement levels for a user by receiving the user's glucose levels collected by a continuous glucose monitoring (CGM) device over a time period, receiving engagement data associated with the user, wherein the engagement data are associated with the user's medication intake, diet, physical activity, laboratory results, and education activity, determining a first glycemia risk index (GRI) value, determining, using a machine learning model and responsive to the user's glucose levels and the engagement data collected over the time period, one or more predictions for future glucose levels for the user including a prediction that a future GRI value is greater than or less than the first GRI value, and determining, using the machine learning model and responsive to the user's engagement data collected over the time period, one or more predictions for future engagement levels.
Core Innovation
The document describes sensing a user’s glucose levels using a continuous glucose monitoring (CGM) device over a time period, and receiving the glucose levels collected by the CGM device. It further receives engagement data associated with the user, collected by a computing device over the same time period, and associated with medication activity, diet activity, physical activity, laboratory results, education activity, and CGM device usage.
The document determines a first glycemia risk index (GRI) value based on a first amount of time the user is hypoglycemic during the time period and a second amount of time the user is hyperglycemic during the time period. It also determines a time in range (TIR) value based on an amount of time the user’s glucose level is within a threshold band over the time period, where the threshold band is determined based on lifestyle, habits, and medical test results of the user.
It then inputs the user’s glucose levels and the engagement data into a machine learning model, which outputs predictions for future glucose levels responsive to the user’s glucose levels and the engagement data collected over the time period. The predictions include future GRI and future TIR relations, future GRI zone and threshold relations, and predictions for future engagement levels, including CGM device engagement level, high engagement state or low engagement state, and future manual engagement level.
In addition, the document describes calculating a timing and a dosing amount of basal insulin at regular intervals and a timing and a dosing amount of bolus insulin corresponding to mealtimes, and administering via an insulin pump in response to an output of the machine learning model. It also includes providing notifications to the user based on the predictions of future engagement levels, and synchronizing administration via the insulin pump in response to the model’s output.
Claims Coverage
The partial content provides three independent claims: clm-00001 (method), clm-00007 (training method), and clm-00011 (system). Across these independent claims, the inventive features center on CGM-derived glycemia risk index (GRI) and time in range (TIR) together with engagement data, future predictions using a machine learning model, and integration of insulin pump control and engagement-based notifications.
Future predictions from CGM glucose
Sensing a user’s glucose levels using a continuous glucose monitoring (CGM) device over a time period; receiving the user’s glucose levels collected by the CGM device; inputting the user’s glucose levels into a machine learning model; outputting one or more predictions for future glucose levels responsive to the user’s glucose levels and the engagement data collected over the time period.
Engagement data associated with medication, diet, activity, labs, education, and CGM usage
Receiving engagement data associated with the user, the engagement data collected by a computing device over the time period, wherein the engagement data is associated with the user’s medication activity, diet activity, physical activity, laboratory results, education activity, and CGM device usage, wherein at least some of the engagement data is collected using one or more sensors associated with the user.
GRI determination from hypoglycemic and hyperglycemic time
Determining a first glycemia risk index (GRI) value based on a first amount of time the user is hypoglycemic during the time period and a second amount of time the user is hyperglycemic during the time period.
TIR determination from a threshold band determined by lifestyle, habits, and medical test results
Determining a time in range (TIR) value of the user’s glucose level based on an amount of time the user’s glucose level is within a threshold band over the time period, where the threshold band is determined based on lifestyle, habits, and medical test results of the user.
Machine learning model outputs future GRI and future TIR relations
Outputting, by the machine learning model and responsive to the user’s glucose levels and the engagement data collected over the time period, predictions for future glucose levels for the user including future GRI and future TIR relations, GRI zone relations, and threshold relations.
Future engagement predictions including CGM engagement level and engagement states
Outputting, by the machine learning model and responsive to the user’s engagement data collected over the time period, predictions for future engagement levels, including future CGM device engagement level, high engagement state or low engagement state, and future manual engagement level.
Basal and bolus insulin calculations and insulin pump administration responsive to model outputs
Calculating, by the machine learning model, a timing and a dosing amount of basal insulin for the user at regular intervals and a timing and a dosing amount of bolus insulin for the user corresponding to mealtimes; administering, via an insulin pump, in response to an output of the machine learning model, the dosing amounts of basal insulin and bolus insulin; synchronizing administration, via the insulin pump, in response to an output of the machine learning model.
Engagement-based notifications tied to predicted future engagement levels
Providing, by the machine learning model, notifications to the user based on the one or more predictions of future engagement levels.
Training with first, second, and third glucose and corresponding engagement across time periods
Receiving first, second, and third glucose levels of the user collected by a CGM device over first, second, and third time periods; receiving first, second, and third engagement data associated with the user over the same time periods; deriving one or more sets of features and sub features; training a machine learning model based on a machine learning algorithm and training data comprising the glucose levels, engagement data, and derived sub features; determining one or more patterns in the training data; and receiving, from the trained machine learning model, predictions for future glucose levels and future engagement levels.
Across the independent claims, the document provides a computer-implemented method and system that sense CGM glucose and receive engagement data associated with medication, diet, physical activity, laboratory results, education activity, and CGM device usage. It determines GRI from hypoglycemia and hyperglycemia time and determines TIR from time within a threshold band defined by lifestyle, habits, and medical test results, and then uses a machine learning model to output future glucose predictions together with future engagement predictions. The independent claims further integrate insulin pump basal and bolus calculations and administration responsive to the model output and provide notifications based on predicted future engagement levels.
Stated Advantages
Not explicitly described in patent.
Documented Applications
Not explicitly described in patent.
Interested in licensing this patent?