Development of a Two-State Gaussian Hidden Markov Model for Modelling Dementia Progression in Patients with Mild Cognitive Impairment

Dementia is characterized by a progressive deterioration of brain function affecting mental processes, such as memory, problem-solving, and concentration. Modelling of dementia progression can provide important insights into the disease process through the visualization and evaluation of disease trajectories. To demonstrate dynamic changes that can take place during the course of a disease trajectory in patients with mild cognitive impairment (MCI), we propose a Gaussian Hidden Markov Model (HMM) that incorporates non-invasive markers commonly used for assessing cognitive and functional changes in dementia diagnostics. We train our model in an unsupervised manner using the Baum-Welch Expectation-Maximization method and evaluate its performance by comparing HMM states with the corresponding ground-truth labels, not used during the model development phase. We believe our HMM framework can contribute to a better understanding of conversion to dementia and be used to support clinical decisions in real world situations.