The opportunities for using physiological data in AI-based closed-loop systems for young adults with T1D in the UK 

Type 1 diabetes (T1D) is a chronic condition characterised by insufficient insulin production, necessitating insulin injections for blood glucose control [1]. Recent advancements in T1D technology are leading to the automation of self-management through closed-loop systems. However, these systems face challenges in effectively accounting for the impact of physical activity on blood glucose levels. To address this issue, we propose incorporating physiological data from wearable devices to inform the closed-loop system about the user's activity.

To evaluate the feasibility of this approach, we plan to provide smartwatches to young adults in the UK who live with T1D. This age group has the lowest likelihood of meeting NHS targets for T1D management and a higher adoption of new technology [2][3]. Participants will upload their blood glucose readings, insulin doses, carbohydrate intake, and smartwatch data. They will also take part in interviews and focus groups to gather their experiences and thoughts on integrating smartwatches into T1D technology. While existing datasets cover some aspects of this work [4, 5], none comprehensively combine these elements, especially for young adults in the UK, or include qualitative insights.

The project aims to gather data on smartwatch usage in the context of T1D to enable analysis from both algorithmic and user perspectives within closed-loop systems. The project involves four milestones:
M1. Complete the data collection process, including interviews, focus groups, and the uploading of T1D and smartwatch data.
M2. Clean, anonymise, and integrate data from various sensors into a consistent time series format, and transcribe the interviews and focus groups.
M3. Conduct a data challenge using an online platform with a sample of the collected dataset to showcase its utility.
M4. Publicly release the dataset to benefit other researchers.

Possible challenges for this project include the need for careful cleaning and integration of multivariable time series data with varying sampling rates. Also, the requirement to bridge various research disciplines due to the project's multidisciplinary nature.