|
Abstract:
|
Brain age prediction as a potential biomarker for risk of neurodegenerative disorder, such as Alzheimer’s disease, has grown in popularity through utilizing neuroimaging data and machine learning models. However, longitudinal studies examining the temporal pattern of brain aging and its link with incident dementia are lacking. While many studies have analyzed brain age differences at discrete time points, the present study contributes to the literature by using a large longitudinal data set and multimodal brain imaging data shared by the Alzheimer’s Disease Neuroimaging Initiative (ADNI). A key issue in the field is accurate prediction of an individual’s likelihood in converting to dementia, a problem addressed using the longitudinal data set. A support vector regression (SVR) model was trained using MRI and PET data from healthy controls (test r=0.61, MAE=3.94), and used to predict brain age in groups with stable and progressive mild cognitive impairment (MCI). Multilevel modeling found that the rate of change in brain aging was significantly faster in the progressive MCI group, demonstrating the methods’ potential in the usefulness of brain age as a biomarker for dementia.
|
