Fuzziness-Based Semi-Supervised Learning for Early Detection of Alzheimer’s Disease using MRI data

Alzheimer’s disease (AD) is a progressive neurodegenerative condition that erodes memory and cognitive abilities. It manifests as a gradual decline in cognitive functions, including memory loss, impaired decision-making and challenges in communication. As the disease advances, individuals may also experience changes in behavior and struggle with daily tasks. Detecting Alzheimer’s in its early stages is crucial, as intervention can slow brain damage. However, diagnosing it can be complex, especially when symptoms are subtle and resemble normal aging. This study’s focus is early AD detection, particularly identifying Mild Cognitive Impairment (MCI), a challenging clinical diagnosis from MRI data. A novel approach is proposed where deep features extracted from two different CNN architectures and ORBbased features are combined through feature-level fusion. After performing feature-level fusion, a semi-supervised methodology are incorporated in the concept of fuzziness. The classification task is executed using a SVM classifier, yielding an accuracy of 94.39% for MCI vs CN (Cognitively Normal) and 97.90% for AD vs CN. The proposed approach exhibits significant improvements in the classifier’s performance compared to existing methods on the ADNI dataset.