Predicting Age Using Neuroimaging: Innovative Brain Ageing Biomarkers

Brain age can be predicted in individuals based on neuroimaging data using machine learning approaches to model trajectories of healthy brain ageing. The predicted brain age for a new individual can differ from his or her chronological age; this difference appears to reflect advanced or delayed brain ageing. Brain age has been shown to relate to cognitive ageing and multiple aspects of physiological ageing and to predict the risk of neurodegenerative diseases and mortality in older adults. Various diseases, including HIV, schizophrenia, and diabetes, have been shown to make the brain appear older. Further, brain age is being used to identify possible protective or deleterious factors for brain health as people age. Brain age is being actively developed to combine multiple measures of brain structure and function, capturing increasing amounts of detail on the ageing brain. The brain changes as we age and these changes are associated with functional deterioration and neurodegenerative disease. It is vital that we better understand individual differences in the brain ageing process; hence, techniques for making individualised predictions of brain ageing have been developed. We present evidence supporting the use of neuroimaging-based ‘brain age’ as a biomarker of an individual’s brain health. Increasingly, research is showing how brain disease or poor physical health negatively impacts brain age. Importantly, recent evidence shows that having an ‘older’-appearing brain relates to advanced physiological and cognitive ageing and the risk of mortality. We discuss controversies surrounding brain age and highlight emerging trends such as the use of multimodality neuroimaging and the employment of ‘deep learning’ methods. The brain changes as we age and these changes are associated with functional deterioration and neurodegenerative disease. It is vital that we better understand individual differences in the brain ageing process; hence, techniques for making individualised predictions of brain ageing have been developed. We present evidence supporting the use of neuroimaging-based ‘brain age’ as a biomarker of an individual’s brain health. Increasingly, research is showing how brain disease or poor physical health negatively impacts brain age. Importantly, recent evidence shows that having an ‘older’-appearing brain relates to advanced physiological and cognitive ageing and the risk of mortality. We discuss controversies surrounding brain age and highlight emerging trends such as the use of multimodality neuroimaging and the employment of ‘deep learning’ methods. a biological measurement that gives an estimate of an organism’s biological age based on the biological age of an organ, tissue, or cell. the hypothetical underlying age of an organism, defined by measuring some aspect of the organism’s biology. Biological age may differ from the organism’s chronological age and be a better indicator of residual lifespan, functional capacity, and risk of age-associated changes. the predicted age of an individual derived using high-dimensional neuroimaging data in a machine learning framework. Brain age potentially represents a biomarker of the underlying ‘age’ of the brain, whereby an ‘older’ brain in adults indicates increased risks of neurodegenerative diseases and mortality. changes to the human brain that generally accompany ageing. These changes occur at molecular, cellular, and tissue levels and have characteristic functional and behavioural consequences (Box 1). an extension of machine learning based on artificial neural networks. ‘Deep’ refers to the multiple layers of neural networks used, including one or more ‘hidden’ layers. Each layer is used to transform input data into a different format that encodes something salient about the features contained in the data. a variable used in a machine learning algorithm or an aspect of a dataset that is of some relevance. In the context of brain age, features are local measures of brain structure or function (e.g., grey matter volume). the scientific study of the old and the ageing process. a statistical approach derived from the study of artificial intelligence based on the concept that statistical models should be able to make accurate predictions from new ‘unseen’ data (either categorical, e.g., group membership, or continuous, e.g., age, IQ). a medical imaging technique that capitalises on the inherent physical properties of biological tissues when inside powerful magnetic fields. Particularly, hydrogen atoms contained in water within biological tissue behave in characteristic ways when the magnetic fields are manipulated and release energy in the form of radiofrequency (RF) pulses that can be recorded. These RF pulses can be transformed into 3D images that give information on brain volume, blood flow, brain function, and white matter microstructure, to name but a few biological characteristics. a volume element, the 3D equivalent of a pixel. Voxels are the unit of resolution for MRI scans of the brain. voxel-wise maps of the brain where each voxel contains a numeric representation of the statistical model learned by a machine learning algorithm.