Converse or reverse? Machine-learning modeling for disease progression: A study based on Alzheimer's disease...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomarker Discovery

Background:

• Understanding the progression from healthy aging to dementia is complex.
• Identifying reliable markers for cognitive decline trajectories is crucial.

Purpose of the Study:

• To evaluate machine learning models for classifying health-to-dementia trajectories.
• To identify key predictive markers across the cognitive continuum.

Main Methods:

• Five machine learning approaches were assessed using the ADNI cohort (ADNI1, ADNIGO, ADNI2, ADNI3).
• Participants with stable, convertible, and reverse progression trajectories were analyzed.
• Feature importance was evaluated for predictive markers.

Main Results:

• Random Forest demonstrated superior performance with 70.8% sensitivity and 96.8% specificity.
• Key predictors included visuospatial/memory deficits, amyloid uptake variations, plasma APOE4, and neurofilament light chain levels.
• These markers effectively classified participant groups across different disease stages.

Conclusions:

• Machine learning holds significant potential for classifying disease trajectories.
• The study identified crucial clinical and neuroimaging biomarkers for predicting cognitive decline.