A predictive model of Parkinsonian brain aging based on brain imaging features

Area of Science:

• Neuroimaging
• Machine Learning
• Gerontology

Background:

• Brain aging is a complex process.
• Parkinson's disease (PD) is associated with accelerated brain aging.
• Predictive models for brain age in PD are needed.

Purpose of the Study:

• To develop and validate a brain imaging-based model for predicting brain age in Parkinson's disease patients.
• To identify key neuroimaging indicators associated with brain aging in PD.
• To explore the utility of machine learning for assessing brain aging in neurodegenerative diseases.

Main Methods:

• Utilized structural brain MRI data from healthy individuals (IXI database) and Parkinson's patients (PPMI database).
• Extracted 1214 structural indicators across brain regions.
• Developed a predictive model using XGBoost machine learning and SHAP (Shapley Additive Explanations) for interpretation.

Main Results:

• The XGBoost + SHAP model accurately predicted brain age in Parkinson's patients with a mean absolute error of 4.21 years.
• Key predictors included superior temporal folding index, subcortical gray matter volume, left thalamus volume, and left/right vascular volumes.
• Identified 15 characteristic indicators most associated with brain aging in Parkinson's disease.

Conclusions:

• The developed model effectively predicts brain age in Parkinson's disease.
• Specific neuroimaging features show significant potential as clinical biomarkers for brain aging in PD.
• Findings support the use of neuroimaging and machine learning for early diagnosis and understanding PD pathogenesis.