Hidden Markov Model based Predicting of Alzheimer’s Disease with graph cut segmentation using MR Diffusion Tensor Imaging (DTI)

Abstract

Alzheimer’s disease (AD) is the predominant aetiology of dementia among the elderly population, accounting for about 60–70% of all instances of cognitive decline. Diffusion tensor imaging (DTI) is a contemporary methodology that enables the cartography of alterations in the microstructure of white matter (WM) in neurological diseases. Nevertheless, the effort of analysing substantial amounts of medical pictures poses significant challenges, prompting researchers to shift their focus towards machine learning. This approach encompasses a collection of computer algorithms that possess the ability to autonomously adjust their output to align with the desired goal. This work proposed the use of a combined approach using Hidden Markov Model (HMM) and MR-DTI, where Diffusion Tensor Imaging (DTI) is employed as a magnetic resonance imaging technique. The purpose of this method is to forecast the occurrence of AD. Furthermore, the statistical analysis demonstrated a significant correlation between microstructural WM changes with both output in the patient groups and cognitive functioning. This finding suggests that these abnormalities in WM might potentially serve as a biomarker for AD. The proposed method is named as Graphcut Hidden MorkovModel (Graph_HMM) is evaluated on ADNI database with statistical analysis and found that it achieves 99.8% of accuracy, 96.4% of sensitivity, 97.4% of specificity and 12.3% of MSE.