Early Detection of Alzheimer's Disease From Cortical and Hippocampal Local Field Potentials Using an Ensembled...

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Early diagnosis of Alzheimer's disease (AD) remains a significant challenge.
• Data-driven methods for AD detection benefit from multimodal data integration.
• Decoding complex cognitive functions from neuronal signals requires sophisticated approaches.

Purpose of the Study:

• To develop an explainable machine learning model (EXML) for early Alzheimer's disease detection.
• To identify subtle patterns in local field potential (LFP) signals as potential early AD markers.
• To leverage multimodal data, including LFP, electrocardiogram, and respiration signals, for enhanced diagnostic accuracy.

Main Methods:

• Acquisition of LFP signals from healthy and AD animal models using multielectrode probes.
• Integration of electrocardiogram and respiration signals to validate LFP data.
• Generation of feature sets from LFPs across temporal, spatial, and spectral domains.
• Application of an ensembled machine learning model with late fusion for classification.

Main Results:

• The EXML model achieved a high overall accuracy of 99.4% in detecting early-stage AD.
• The model identified subtle network activity patterns indicative of amyloid plaque deposition at 3 months post-onset.
• Both individual and ensemble models demonstrated robustness against channel artefacts.

Conclusions:

• The proposed EXML model offers a promising tool for the early and accurate diagnosis of Alzheimer's disease.
• Multimodal data integration and explainable AI enhance the reliability of detecting early neuropathological changes.
• This approach provides valuable insights into the progression of AD at the network activity level.