The ICeX framework: Brain age estimation from thalamic nuclei with conformalized and eXplainable random forest...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Gerontology

Background:

• Accurate brain age estimation is crucial for detecting early signs of neurological and cognitive decline.
• Traditional methods often overlook region-specific brain changes, limiting diagnostic precision.
• Thalamic nuclei, sensitive to aging, offer potential for more reliable brain age prediction.

Purpose of the Study:

• To develop a novel framework for brain age prediction utilizing thalamic nuclei volumes.
• To enhance model interpretability and reliability through feature attribution and uncertainty quantification.
• To provide a tool for precise exploration of individual aging trajectories.

Main Methods:

• A Random Forest Regression model was trained on thalamic nuclei volumes from 630 healthy adults.
• SHAP-based feature attribution and Conformal Prediction were integrated into the Individual Conformalized Explanations (ICeX) approach.
• ICeX provides subject-specific prediction intervals and insights into feature influence on brain age and uncertainty.

Main Results:

• The model achieved a mean absolute error of 2.77 years with 90.77% coverage.
• Key thalamic nuclei (e.g., Lateral Geniculate, Paratenial, Ventromedial) significantly contributed to prediction accuracy.
• ICeX successfully detailed individual feature contributions to both predicted brain age and its uncertainty.

Conclusions:

• The ICeX framework offers reliable and interpretable brain age predictions using region-specific biomarkers.
• This approach provides clinicians and researchers with a transparent tool for analyzing individual aging.
• Supports early detection and personalized interventions for age-related neurological conditions.