Early-Stage Alzheimer's Disease Detection Using an Explainable Deep Learning Method and Functional Magnetic Resonance Imaging (fMRI)

Alzheimer’s disease (AD) is a degenerative condition that alters the brain’s biology and structural integrity. Degeneration can lead to memory loss and social and cognitive deterioration. An individual’s death might happen depending on how severe their AD is. Since there is no treatment for it, early detection of AD is crucial to slowing its progression. A noninvasive diagnostic technology called functional magnetic resonance imaging (fMRI) may detect minute variations in blood flow that are brought on by brain activity. It can also be utilized to determine AD in its early stages. In addition, deep learning (DL) approaches have been employed to detect different stages of AD using fMRI data. There exist certain constraints in the setup and interpretation of 4D fMRI data as input for DL models. Certain methodologies rely on specific techniques to decrease the dimensionality of data, rendering it suitable for utilization in DL models. Yet, in the majority of instances, noteworthy information is forfeited as a result of this reduction in dimensionality. Furthermore, DL is sometimes viewed as a "black box," and it is crucial to make sure that judgments about it are influenced by factors that are suitable for the settings in which they are utilized, ranging from law to healthcare. In this thesis, we try to put in consideration regarding these issues and utilize a data shape keeping the most spatial and temporal information, in addition to hire an explainable deep learning approach to detect Mild Cognitive Impairment (MCI) using fMRI data.