Alzheimer's Imaging Consortium.

Area of Science:

• Neuroimaging
• Alzheimer's Disease Research
• Computational Neuroscience

Background:

• Predicting dementia risk is crucial for early Alzheimer's disease (AD) management.
• Brain structure volumes deviate from normal trajectories in individuals with AD.
• Analyzing deviations from expected age and sex-based volumetric measurements can predict AD progression risk.

Purpose of the Study:

• To predict individual Alzheimer's disease progression risk.
• To analyze deviations from age and sex-based volumetric measurements of brain structures.
• To identify individuals at high risk for AD conversion.

Main Methods:

• Utilized T1w MRI scans from multiple large-scale databases.
• Modeled healthy aging trajectories for 70 brain regions using Bayesian multilevel models and cubic b-splines.
• Employed a linear Survival Support Vector Machine with permutation importance to predict time-to-event for AD conversion.

Main Results:

• Achieved a concordance index of 0.82(0.04) in predicting AD conversion risk.
• Identified baseline diagnosis (Cognitively Normal or Mild Cognitive Impairment) as a key predictor.
• Highlighted the importance of baseline volumes in the Amygdala, Inferior lateral ventricles, Superior temporal gyrus, and Superior frontal gyrus.

Conclusions:

• Developed a library of healthy aging trajectories for 70 anatomical structures.
• Demonstrated the effectiveness of the method in identifying subjects at high risk of progressing to Alzheimer's Dementia.
• The approach offers a novel way to assess AD progression risk using neuroimaging data.