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

Introduction: This study explores the use of imaging to evaluate brain aging to establish a model for predicting brain age in patients with Parkinson's disease. Methods: Structural brain MRI data from 345 healthy individuals were obtained from the IXI database, while data from 59 Parkinson's patients and 59 healthy controls were acquired from the PPMI database. A total of 1214 structural indicators were extracted, including information on the whole brain, cortex, subcortex, and white matter. This led to the development of a model for predicting brain age in Parkinson's patients. The model combined brain imaging features with a machine learning algorithm and the Shapley Additive Explanations (SHAP) interpretation model. Fifteen characteristic indicators most closely associated with Parkinson's brain aging were determined. Results: The XGBoost model + SHAP method framework, using the minimum mean absolute error for assessing brain aging within 4.21 years, was effective in predicting brain age in patients with Parkinson's disease. The superior temporal folding index and subcortical gray matter volume, left thalamus volume, and left and right vascular volumes had the most significant impact on the prediction results, suggesting their potential as clinical indicators for evaluating the extent of brain aging in Parkinson's patients. Discussion: These findings provide important clues for understanding the mechanisms underlying brain aging, as well as brain imaging evidence for the early diagnosis and treatment of Parkinson's disease.