Neuroimaging and machine learning for brain age estimation

Abstract

2822 AGING The morphological correlates of brain aging in neuroimaging are characterized grosso modo by a decline in gray matter volume and a reduction in white matter integrity these changes are associated with cognitive decline including memory, attention, and executive functions and are thought to be the result of a combination of factors from predisposition and environment. There is converging evidence that regionally differential aging processes in the human brain exist, mainly affecting the frontal lobe and relatively sparing posterior and infratentorial areas [1]. Magnetic Resonance Imaging (MRI) is capable of providing high-resolution images of the brain that can be assessed for age-related changes [2]. Machine learning (ML), as an emerging artificial intelligence (AI)-based field in medicine, has been applied to quantify these MRI changes in the human brain, commonly accomplished by (1) extracting features from brain images and (2) training an ML model to predict the age from the extracted features [3]. Application to (advanced) MRI data obtains an age estimation that is considered to be representative of an individual’s brain age/health. The difference between estimated brain age and chronological age is called the brain age gap (BrainGAP) or brain predicted age difference (BrainPAD) and is thought to potentially serve as a biomarker for processes like accelerated/delayed brain aging. The BrainGAP was shown to correlate with other imaging biomarkers of brain aging, such as white matter hyperintensities [4], and appears to be related to cardiovascular risk factors associated with accelerated aging in general, such as systolic/diastolic blood pressure, smoking habits, and cardiac function.