Inspired by the groundbreaking paper titled "Accurate brain‐age models for routine clinical MRI examinations", this project develops a predictive model using volumetric features extracted via SynthSeg (FreeSurfer) from 3D T1-w brain MRIs. It estimates brain age, detecting deviations between chronological and biological brain age, showing that Alzheimer’s Disease links to accelerated brain ageing.
Using datasets from ADNI, AIBL, and OASIS, I curated a collection of 7545 MRI scans from 2227 unique patients for analysis. These scans belong exclusively to two diagnostic categories:
- Cognitively Normal
CN68.68% - Alzheimer's Disease
AD31.32%
A polynomial regression model was trained via Cross-Validation, achieving a MAE of 3.99 years, with age as the dependent variable and other features as predictors. Trained on CN data, it predicted brain age for CN patients, with a Brain PAD of -0.12 years, closely matching chronological age. For AD patients, however, it showed a Brain PAD of +7.48 years, indicating that AD’s association with accelerated brain ageing.
