Digital health · 2026
ACTIVE-GLU: Personalised modelling of physical activity-driven glucose dynamics in type 1 diabetes under free-living conditions.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Bilal Ahmad, Thabit Hood, Nutter Paul W, Harper Simon
Аннотация
OBJECTIVE: Physical activity (PA) lowers blood glucose (BG) levels in individuals with Type 1 Diabetes Mellitus (T1DM), with effects varying by intensity and duration. Predicting BG fluctuations under free-living conditions, beyond structured exercise, helps reduce hypoglycaemia risk. This study aimed to model and predict BG responses across varying PA levels using personalised, activity-informed temporal patterns derived from continuous glucose monitoring (CGM) data.
METHODS: Fourteen adults with T1DM were monitored under free-living conditions using Garmin smartwatches for PA tracking and CGM devices for BG measurement. PA was quantified using step count and maxMotion intensity, and categorised into four levels: low, medium, high, and very high. A 10-day rolling window (moving average) was applied to each participant's data to construct the personalised ACTIVE-GLU model, which estimates mean BG change at 15-minute intervals following PA.
RESULTS: The ACTIVE-GLU model accurately predicted BG changes across all PA intensities. The lowest average accuracy (81.67%) was observed under low PA, while the highest prediction error occurred during very high PA (MAE = 0.44 mmol/L, RMSE = 0.57 mmol/L, mean bound deviation = 0.38 mmol/L). Bland-Altman analysis confirmed close agreement between predicted and observed BG values, with negligible bias across intensity levels.
CONCLUSION: In this exploratory cohort study, quantifying daily PA within a personalised temporal framework enables accurate prediction of BG responses in T1DM under real-world conditions. By incorporating PA intensity into BG dynamics, ACTIVE-GLU provides a personalised, PA-aware framework that may support interpretation of BG trends and inform future decision-support strategies. Future work will validate the model using virtual participants to simulate adaptive dosing interventions.
METHODS: Fourteen adults with T1DM were monitored under free-living conditions using Garmin smartwatches for PA tracking and CGM devices for BG measurement. PA was quantified using step count and maxMotion intensity, and categorised into four levels: low, medium, high, and very high. A 10-day rolling window (moving average) was applied to each participant's data to construct the personalised ACTIVE-GLU model, which estimates mean BG change at 15-minute intervals following PA.
RESULTS: The ACTIVE-GLU model accurately predicted BG changes across all PA intensities. The lowest average accuracy (81.67%) was observed under low PA, while the highest prediction error occurred during very high PA (MAE = 0.44 mmol/L, RMSE = 0.57 mmol/L, mean bound deviation = 0.38 mmol/L). Bland-Altman analysis confirmed close agreement between predicted and observed BG values, with negligible bias across intensity levels.
CONCLUSION: In this exploratory cohort study, quantifying daily PA within a personalised temporal framework enables accurate prediction of BG responses in T1DM under real-world conditions. By incorporating PA intensity into BG dynamics, ACTIVE-GLU provides a personalised, PA-aware framework that may support interpretation of BG trends and inform future decision-support strategies. Future work will validate the model using virtual participants to simulate adaptive dosing interventions.