Neural fuzzy semantic modelling of blood glucose under different dietary regime

• Main Author: Lai, Wei Song
• Other Authors: Quek Hiok Chai
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/62802
• Institution: Nanyang Technological University

Diabetes mellitus is a terminal disease that affects more than 9% of the world’s population as of 2014 and was directly responsible for causing 1.5 million deaths in 2012. Out of all patients suffering from diabetes mellitus, about 90% have type 2 diabetes mellitus (T2DM). Singapore is also reported to have the highest proportion of young T2DM patients with more than 20% of all T2DM patient under the age of 40. The objective of this research project is to look at a closed loop system that eliminates the need of having manual insulin injections for insulin therapy. The project first looked at prediction of glucose using available implementations like the Monte Carlo Evaluative Selection (MCES) for feature selection and the Self-adaptive Fuzzy Inference Network for building a prediction model. The project then looked at prediction of the insulin infusion which was based on the insulin distribution profile constructed using the activation profile of Actrapid, a fast-acting insulin. SaFIN and MCES was once again used to train a prediction model which was used to test for various diet regimes. The closed loop system is still at its infancy stage where more tests and experiments under various dietary regimes have to be carried out in order to validate and verify the protocols. More patient should also be brought in to verify if the system can be ported and personalized for others. Otherwise, the projects serves as a solid groundwork for the closed loop system which possess a huge amount of potential benefits for T2DM patients.