University of Twente Student Theses

Login

Modelling fluctuations of blood glucose levels based on food intake and physical activity in patients with Diabetes Mellitus Type 2 : development of a controlled protocol which administer perturbations in daily life situations

Sawaryn, B. (2020) Modelling fluctuations of blood glucose levels based on food intake and physical activity in patients with Diabetes Mellitus Type 2 : development of a controlled protocol which administer perturbations in daily life situations.

Full text not available from this repository.

Full Text Status:Access to this publication is restricted
Abstract:Introduction - Lifestyle interventions are a key part of Type 2 Diabetes Mellitus (T2DM) treatment, while diet and physical activity not only can increase insulin receptor sensitivity and therefore improve blood glucose levels, but can also greatly reduce the risk of cardiovascular complications. This requires models to predict changes in blood glucose levels, such that the effect of interventions can be analysed to give patients insight of the effect on their blood glucose levels or ultimately advise interventions when hypo- or hyperglycemic events are expected. Development of these models have been investigated at the University of Twente before. The data used from the DIALECT (DIAbetes and LifEstyle Cohort Twente) project has been acquired in unaltered daily living conditions, as a consequence the researchers concluded that the derived models were very insensitive to variations in input parameters. For improved system identification, measurements need to be done under conditions where the subjects' behaviour is sufficiently perturbed with respect to food intake and physical activity under daily living conditions. Methods - The pREdictive Modelling IN Diabetes (REMIND) study was performed on T2DM patients in the outpatient setting, using standardized meals and letting participants perform physical activities which were aimed to induce changes in blood glucose levels. These datasets were used to develop subject-specific autoregressive (AR) and autoregressive-moving-average with exogenous inputs (ARMAX) models and were assessed on prediction accuracy and sensitivity. Each participant received each change in dietary intake and physical activity in duplo. To verify whether the REMIND study really resulted in more accurate and sensitive models, the entire analysis was repeated for each subject with the data from the DIALECT project and compared with the results from the REMIND study. Results - At a prediction horizon of 120 minutes, on average the AR and ARMAX models respectively performed best with a root mean square error (RMSE) of 1.95 and 1.64 mmol/L. The final subject specific models were on average able to predict in clinically acceptable levels 98.9\% and 99.4\% of the time for the AR and ARMAX models respectively. Finally, the sensitivity analysis showed that the developed ARMAX models on average were sensitive to specific types of perturbations, mainly carbohydrates and steps. Compared to training and testing on the dataset of DIALECT, performance is not greatly improved and does not result in more clinically acceptable or sensitive predictions. Conclusion - These and other models should be explored more thoroughly and more data should be acquired. If comparable or better significant results are achieved, these types of models could be used in a coaching system to aid T2DM patients for improved glycemic control.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:42 biology, 44 medicine, 54 computer science
Programme:Biomedical Engineering MSc (66226)
Link to this item:http://purl.utwente.nl/essays/80996
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page