Conv-eRVFL: Convolutional Neural Network Based Ensemble RVFL Classifier for Alzheimer's Disease Diagnosis

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Alzheimer's disease (AD) presents a growing global health challenge.
• Early identification of AD and mild cognitive impairment (MCI) is difficult but crucial for timely intervention.
• Neuroimaging techniques like MRI and PET reveal structural and metabolic brain changes indicative of AD years before symptom onset.

Purpose of the Study:

• To propose a novel image fusion approach for enhanced Alzheimer's disease detection.
• To integrate structural (MRI) and metabolic (PET) neuroimaging data using advanced machine learning.
• To improve the classification accuracy of Alzheimer's disease (AD) and mild cognitive impairment (MCI) using a fusion-based ensemble model.

Main Methods:

• A wavelet packet transform-based fusion technique was developed for MRI and PET scans.
• An eight-layer Convolutional Neural Network (CNN) extracted multi-layer features from fused images.
• Features were processed by an ensemble of Random Vector Functional Link (RVFL) networks with s-membership fuzzy activation functions for outlier robustness.

Main Results:

• The proposed fusion approach demonstrated effective feature extraction and classification.
• The ensemble RVFL model achieved notable performance in distinguishing between healthy controls (CN), Alzheimer's disease (AD), and mild cognitive impairment (MCI).
• Experimental results on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset validated the model's effectiveness.

Conclusions:

• The developed wavelet packet transform-based fusion method shows promise for early AD and MCI detection.
• Ensemble learning with RVFL networks offers a robust approach for analyzing complex neuroimaging data.
• This fusion-based strategy enhances diagnostic capabilities for neurodegenerative diseases.