Objective
Objective 1: Develop and implement a search engine that can find Continuous Glucose Monitoring (CGM) traces similar to a user's observed pattern (Search Query Pattern (SQP)).
Objective 2: Create an intelligent filter that retrieves insulin profiles associated with similar CGM traces obtained by the search engine, based on the SQP.
Objective 3: Develop a personalized forecast module (PFM) to estimate upcoming insulin requirements. This module will be integrated into the UVA-AID algorithm.
Background Rationale
Continuous glucose monitoring (CGM) technology has revolutionized the management of diabetes by providing valuable data on blood sugar levels and led to the development of automated insulin delivery (AID) systems. The UVA Center for Diabetes Technology (CDT) has been at the forefront of this research since its beginning in 2006, when the first JDRF exploratory AID studies were initiated. Since then, the CDT: (i) introduced the first computer simulator of the human metabolic system accepted by the FDA as a substitute to animal trials in the development of AID strategies (2008), (ii) developed the first wearable AID system running on a smart phone (DiAs, 2012), and (iii) designed the algorithm and run the pivotal trials for an AID system now known as Control-IQ (Tandem Diabetes Care), approved by the FDA in 2019. We therefore have in-depth knowhow of the advantages and pitfalls of AID: while most AID system show immediate improvement over other therapies, these systems soon reach a “ceiling” that they cannot exceed. This is primarily due to the lack of “learning” and adaptation of the system to the changing behavior and physiology of the user. Therefore, this project is proposing a technique that continually learns from a person’s glycemic patterns and the patterns of others stored in databases, and thereby adapts to improve AID performance.
Description of Project
This application comes from the developers of the University of Virginia (UVA) Automated Insulin Delivery (AID) System, which in 2012 became the first wearable artificial pancreas running on a smart phone, and in 2017 was built in Tandem’s t:slim X2 insulin pump to become Control-IQ – an AID system now used by over 500,000 people with type 1 diabetes worldwide. After this successful clinical translation, our academic mandate now is to open new horizons, design and test new AID algorithms, and train the next generation of diabetes researchers.
In this project, we propose to integrate a Personalized Forecast Module (PFM) in the UVA-AID, which will achieve personalized adaptive insulin delivery that continues to improve its performance over time. A most innovative aspect of this project is the capability of this Module to learn from the continuous glucose monitoring (CGM) and insulin delivery patterns of the person who uses it, and from the patterns of others stored in population databases; thus, unlike any other AID algorithm, the PFM will use both a person’s own history and the history of others, to forecast and adapt to metabolic disturbances.
The first step towards this development is creating an efficient CGM search engine to quickly find similar CGM/insulin patterns, which will allow the PFM to be informed by the individual's past glucose and insulin data, as well as data from other people with similar patterns, to predict changes in glucose levels and adjust insulin delivery accordingly. The UVA Center for Diabetes Technology already has a large database of CGM/insulin profile from over 20,000 people using the UVA-AID technology; thus, we will implement and test the concept in this project in this database before proceeding to a clinical trial. The goal is to create a system that can personalize insulin delivery in real-time based on patterns and outcomes from similar cases in the database.
Anticipated Outcome
The anticipated outcome of this project is to provide valuable tools and insights for both individuals with diabetes and healthcare professionals. The project aims to achieve the following objectives:
1. Efficient Data Retrieval: The project will develop an artificial intelligence framework that allows for quick and easy retrieval of similar Continuous Glucose Monitoring (CGM) and corresponding insulin delivery data from large Automated Insulin Delivery (AID) databases. This will enable individuals and healthcare professionals to access relevant and comparable data efficiently;
2. Personalization and Adaptation of AID Systems: By analyzing patterns in individual and population data, the project will facilitate the design of personalized and adaptable AID systems. This means that the systems can be tailored to meet the specific needs of individuals with type 1 diabetes, taking into account their unique patterns and preferences, as well as similar patterns stored in databases;
3. Clinician Support: The project will provide clinicians with a system that allows them to search for, and interpret, individual CGM and insulin delivery data within the context of similar historical records. This will enable clinicians to derive actionable information, and make informed decisions and recommendations for patient care.
Relevance to T1D
The relevance of this project to Type 1 diabetes lies in its focus on improving diabetes management specifically in the following ways:
1. Personalized Diabetes Management: The project aims to facilitate the development of personalized and adaptable AID systems. This is crucial for individuals with Type 1 diabetes, as their insulin needs, and glucose patterns can vary widely. By analyzing individual and population data, the project can help tailor diabetes management approaches to meet the specific needs of individuals with Type 1 diabetes.
2. Data Analysis and Insights: The project focuses on analyzing CGM and simultaneous insulin delivery patterns. This analysis can provide valuable insights into glucose control and insulin usage in individuals with Type 1 diabetes. By understanding these patterns, individuals and healthcare professionals can make informed decisions regarding insulin dosing, diet, exercise, and overall diabetes management.
3. Clinician Support and Actionable Information: The project aims to provide clinicians with a system to search for and interpret individual data within the context of similar historical records. This can help healthcare professionals better understand a patient's diabetes management progress and make more accurate assessments and recommendations. By offering actionable information, the project can assist clinicians with optimizing treatment plans and improving outcomes for individuals with Type 1 diabetes.