A hybrid multimodal machine learning model for Detecting Alzheimer's disease

Area of Science:

• Neuroimaging and Machine Learning for Neurodegenerative Diseases

Background:

• Single neuroimaging modalities have limitations for accurate Alzheimer's disease (AD) diagnosis.
• Integrating complementary biomarkers from multiple sources can enhance diagnostic performance.
• Multimodal data fusion offers a promising avenue for improved characterization of AD.

Purpose of the Study:

• To develop and evaluate a multimodal machine learning framework for enhanced Alzheimer's disease diagnosis.
• To integrate magnetic resonance imaging (MRI), positron emission tomography (PET), and cerebrospinal fluid (CSF) data.
• To utilize a novel hybrid optimization and classification algorithm for feature selection and diagnosis.

Main Methods:

• Proposed a multimodal machine learning framework combining MRI, PET, and CSF data.
• Developed an enhanced Harris Hawks Optimization (ILHHO) algorithm for feature selection and Kernel Extreme Learning Machine (KELM) for classification.
• Evaluated the ILHHO-KELM model on 202 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Main Results:

• The ILHHO algorithm demonstrated superior optimization performance compared to other meta-heuristic algorithms.
• The multimodal approach (MRI + PET + CSF) achieved 99.2% accuracy in distinguishing AD from normal controls (NC).
• Multimodal classification significantly outperformed single-modality diagnostic accuracy.
• Discriminative feature analysis revealed complementary information from MRI and PET, highlighting neurodegeneration patterns.

Conclusions:

• Multimodal data fusion, integrating MRI, PET, and CSF, significantly improves Alzheimer's disease diagnostic accuracy.
• The synergistic ILHHO-KELM model effectively extracts sensitive imaging signatures for AD detection.
• Advanced feature learning techniques applied to complementary biomarkers are crucial for enhancing AD diagnosis.