Dementia Care Research and Psychosocial Factors

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Alzheimer's disease (AD) involves progressive cognitive decline and neurodegeneration.
• This study analyzes stable and transition stages of AD progression using longitudinal data.
• A multi-domain approach integrates clinical, memory, and imaging biomarkers to predict disease progression.

Purpose of the Study:

• To predict whether a patient's condition will progress to the next stage or remain stable.
• To enable timely and personalized therapeutic interventions for Alzheimer's disease.
• To assess disease progression by deriving rate-of-change measures.

Main Methods:

• Longitudinal analysis of 1,416 patients in stable (CN, MCI, AD) and transition (CN-to-MCI, MCI-to-AD) groups.
• Data imputation for missing values (<5%) and exclusion of patients with >10% missing data.
• Random Forest Classifier trained and tested on stratified data, evaluated using Accuracy, ROC-AUC, and sensitivity-specificity.

Main Results:

• Significant trends observed in cognitive, memory, and imaging biomarkers across disease stages.
• MCI-to-AD transitions showed steepest changes in cognitive and memory measures, and structural atrophy in imaging.
• Random Forest Classifier achieved high AUC values (0.76-0.98), particularly for MCI-to-AD transitions.

Conclusions:

• Machine learning with multi-domain biomarkers effectively predicts Alzheimer's disease progression.
• Key biomarkers like CDRSB_change and Hippocampus_change identify transitions.
• This approach supports early, personalized interventions for Alzheimer's disease.