Work

Classification of mild cognitive impairment and alzheimer's disease patients: a multiscale and multivariate approach

Year: 2012
Type: dissertation
Abstract: [ANGLÈS] Alzheimer's Disease (AD) is the most common form of dementia and a growing health and socioeconomic problem. Moreover, the impact of the disease is expected to increase even more as the life expectancy is going to grow over the years. Consequently, a lot of research is focused on computer-aided diagnosis techniques that aim at quantitatively study Magnetic Resonance brain images of early stage patients. Early diagnosis could help in better future cure or disease- modifying treatments. An example of AD early stage is Mild Cognitive Impariment (MCI), as the 50% of the individuals who suffer from this pathology develop AD in three of four years. In this work, we use Support Vector Machines to classify subjects from AD, MCI and healthy control (CTL) groups. Our main objective is to study whether combining different anatomical scale brain regions and different image modalities could improve the classification accuracy. Thus, regional and global Grey Matter (GM) volumes (multiscale approach), Withe Matter (WM) regional volumes, Regional Asymmetry coefficients and T1- quantitative MRI data (multivariate) are combined. Our accuracies when comparing CTL vs AD and CTL vs MCI with large public databases (ADNI) are comparable to the results in the literature: 88.3% and 81.8% respectively. In this master thesis we study also smaller databases of MCI patients from Lausanne University Hospital. We pay special attention to the study of pre-processing steps: Intra Craneal Volume normalization and age correction. Our results show that for our small group of patients, better accuracies can be obtained when combining different types of features (multiscale and multivariate) than when only using classical GM region volumes. Moreover, the new region-based age-correction method proposed here presents encouraging results when applied prior to both CTL vsMCI and CTL vs AD classification.
Language: English
Field: Medicine
Open Access status: green
OpenAlex ID: W83162150
Cites: 54
Cited by:
Related to: 10