Casamitjana A. Study of early stages of Alzheimer’s disease using magnetic resonance imaging. Vilaplana V. Signal Theory and Communications. [Barcelona]: Universitat Politècnica de Catalunya; 2019.


Alzheimer's disease (AD) is a neurodegenerative disorder that constitutes the most common dementia pathology. It represents a global epidemic that expands exponentially as the life expectancy increases with no yet useful treatment. Currently, it represents a huge social and economic burden for our societies and it is expected to tension public health infraestructures and finances in the near future.

AD is characterized by amyloid plaque deposition and neurofibrillary tangles measured by ex-vivo examination of the brain. Recent developments in fluid biomarkers and brain imaging allow in-vivo quantification of pathophysiological processes of amyloid deposition or tau tangles formation in the brain, providing the community with highly sensitive and specific in-vivo biomarkers for Alzheimer's disease diagnosis. Abnormal levels of these biomarkers are thought as the initiating event to a cascade of subsequent events that continue with synapse loss, cell death, memory impairment, functional dysfunction and cognitive decline. All these events constitute the Alzheimer's continuum which can be broadly split into two main parts: an initial long and silent preclinical stage characterized by abnormal AD biomarkers and cognition within the normal range that could last from 15 to 30 years and a posterior clinical stage where subjects develop dementia symptoms.

The etiology of AD is still poorly understood even though several risk factors are identified. Large observational studies can help the study of AD and its related biomarkers and risk factors. In this thesis we provide methodological tools for the analysis of Alzheimer's disease using magnetic resonance imaging (MRI). We focus on the study of subjects within the preclinial stage of AD by using statistical learning and pattern recognition frameworks to perform inferential statistics and develop predictive models.

The main outcomes of this thesis are three-fold: firstly, we develop an open-source toolbox for nonlinear neuroimage analysis in population studies. While nonlinear association between medical images and several factors is already known, standard neuroimaging softwares only provide linear statistical frameworks that limit the analyses. Secondly, we study the relationship between brain structure using MRI and the underlying Alzheimer's pathology along the disease continuum and at different stages. The close relationship between MRI and clinical symptoms has been widely studied but describing AD using biomarkers instead of clinical phenotypes allows us to study preclinical stages of AD. Finally, we present a framework to predict cognitively unimpaired and amyloid positive subjects using MR imaging and machine learning. We report the results in a cross-sectional study and in a longitudinal study that compares the volumetric rate-of-change between subjects with different amyloid status. We further test the proposed methodology as a part of the triaging process in clinical trials showing great potential benefits.