Alzheimer's Disease stage identification using deep learning models.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Alzheimer's Disease (AD) diagnosis and staging are crucial for effective patient management.
• Current methods for monitoring AD progression can be invasive or resource-intensive.
• Objective staging of AD using objective patient data is an unmet need.

Purpose of the Study:

• To develop a deep learning model for identifying Alzheimer's Disease stages using mobility data.
• To leverage smartphone-collected accelerometer data for non-invasive AD patient monitoring.
• To improve the accuracy of AD stage identification compared to traditional methods.

Main Methods:

• Collected accelerometer data from 35 Alzheimer's Disease patients over one week using smartphones.
• Processed time-series mobility data, labeling it with disease stages (early, middle, late).
• Utilized a Convolutional Neural Network (CNN) model to recognize patterns indicative of each AD stage.

Main Results:

• The CNN-based approach achieved 90.91% accuracy in identifying AD stages.
• An F1-score of 0.897 was obtained, demonstrating robust performance.
• Significantly outperformed traditional feature-based classifiers in accuracy.

Conclusions:

• Mobility data is a valuable resource for monitoring Alzheimer's Disease progression.
• The developed CNN model enhances the accuracy of AD stage identification.
• This approach offers a promising non-invasive tool for AD patient management and treatment optimization.