Abstract
Alzheimer's disease (AD) is a complex, progressive, and irreversible neurodegenerative disorder marked by cognitive decline and memory loss. Early diagnosis is the most effective strategy to slow the disease's progression. Mild Cognitive Impairment (MCI) is frequently viewed as a crucial stage before the onset of AD, making it the ideal period for therapeutic intervention. AD is marked by the buildup of amyloid-beta (Aβ) plaques and tau neurofibrillary tangles (NFTs), which are believed to cause neuronal loss and cognitive decline. Both Aβ plaques and NFTs accumulate for many years before the clinical symptoms become apparent in AD. As a result, in this study, CerebroSpinal Fluid (CSF) biomarker information is combined with hippocampal volumes to differentiate between MCI and AD. For this, a novel two-stage hybrid learning model that leverages 3D CNN and the notion of a Fuzzy and Machine learning model is proposed. A 3D-CNN architecture is employed to segment the hippocampus from the structural brain 3D-MR images and quantify the hippocampus volume. In stage 1, the hippocampus volume is passed through thirteen machine learning models and fuzzy clustering for classifying symptomatic AD and healthy brain (Normal Control - NC). The CSF data is fuzzified to capture the inherent uncertainty and overlap in clinical data. The identified symptomatic AD data in the stage1 are further classified into MCI and AD with the aid of a fuzzified CSF biomarker in stage 2. The experimental work presented in this study utilized the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. The proposed hybrid model achieved an average accuracy of 93.6% for distinguishing between NC and symptomatic AD and 93.7% for discriminating between MCI and AD. This approach enhances diagnostic accuracy and provides a more comprehensive assessment, allowing for earlier and more targeted therapeutic interventions.