Classification of Alzheimer's Disease Leveraging Multi-task Machine Learning Analysis of Speech and Eye-Movement Data

Area of Science:

• Neuroscience
• Computer Science
• Ophthalmology

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting cognition.
• Early AD detection is crucial for timely intervention and management.
• Subtle changes in eye movements and language may precede cognitive decline.

Purpose of the Study:

• To develop and evaluate a machine learning model for Alzheimer's disease classification using multimodal data.
• To assess the diagnostic utility of novel tasks: pupil fixation and description of a pleasant past experience.
• To compare the performance of individual tasks with fused multimodal data for AD detection.

Main Methods:

• Machine learning analysis of a novel multimodal dataset.
• Inclusion of data from pupil fixation, pleasant past experience descriptions, picture description, and paragraph reading.
• Dataset comprised 79 memory clinic patients (AD, MCI, SMC) and 83 older adult controls.

Main Results:

• Individual novel tasks demonstrated discriminative ability comparable to established tasks.
• Fusing multimodal data across all tasks achieved the highest classification accuracy.
• The combined dataset yielded an Area Under the Curve (AUC) of 0.83 ± 0.01.

Conclusions:

• Novel tasks, including pupil fixation and past experience description, are valuable for AD classification.
• Multimodal data fusion significantly enhances the accuracy of Alzheimer's disease detection.
• This approach offers a promising avenue for early and accurate diagnosis of Alzheimer's disease.