Brain functional connectivity analysis of fMRI-based Alzheimer's disease data.

The prevalence of Alzheimer's disease (AD) poses a significant public health challenge. Distinguishing AD stages remains a complex process due to ambiguous variability within and across AD stages. Manual classification of such multifaceted and massive data of brain volumes is operationally inefficient and vulnerable to human errors. Here, we propose a precise and systematic framework for AD stages classification. The core of this framework discovers and analyzes functional connectivity among regions of interest (ROIs) of a human brain. Multivariate Pattern Analysis (MVPA) is applied to extract features that reveal complex functional connectivity patterns in the brain. These features are then used as inputs for an Extreme Learning Machine (ELM) model to classify AD stages. The model's performance is assessed through comprehensive evaluation metrics to ensure robustness and reliability. Applying this framework on datasets which contain meticulously validated fMRI scans such as the OASIS and AD Neuroimaging Initiative datasets, we validate the merit of this proposed work. The framework's results show improvement in the collective performance of two-class and multi-class classification. Feeding ELM with MVPA features yield decent outcomes given a generalizable and computationally-efficient model. This study underscores the effectiveness of the proposed approach in accurately distinguishing AD stages, offering potential improvements in AD and AD stages detection.