Two layered hidden Markov model for studying type 2 diabetes

This study uses a layered Hidden Markov Model (LHMM) to examine the complex interactions between dietary habits, insulin use, and glucose levels in people with diabetes. The forward algorithm calculates the probability of an emission state sequence. In addition, the study determined the probability mass function for abnormal and normal emission states in one-, two-, and three-state sequences. The three fundamental HMM algorithms were presented and confirmed using manual calculations. The research demonstrates a large likelihood of starting with an unbalanced nutritional status, emphasizing the necessity for dietary adjustments. Transition probability matrices depict dynamic behaviours in food and insulin states, with emission probabilities emphasizing the importance of insulin therapy in maintaining normal glucose levels. Probability mass functions show a disturbing tendency of rising abnormal states over time. Simulation results support the model’s usefulness, implying that it can be used in clinical settings to monitor diabetes control. The study provides insights into diabetes care and highlights the importance of continuous monitoring and dietary modifications.