Population aging is considered one of the most significant demographic trends around the world and has caused a rise in the prevalence of age-related cognitive impairment and dementia. An understanding of the neurobiology of cognitive aging has therefore become increasingly crucial. Functional magnetic resonance imaging (fMRI) is a powerful, non-invasive neuroimaging tool that has critically advanced our understanding of how aging impacts cognition. Blood-oxygenation-level-dependent (BOLD) imaging is one method used in fMRI for measuring changes in neural activity during rest or tasks that are correlated with brain function. However, the measured BOLD signal is often limited by a weak signal and confounded by various subject-dependent sources of artifact and noise that tend to be significantly greater among older age cohorts than younger age cohorts, particularly head motion and physiological processes such as respiration and cardiac pulsation.
These issues have led to the development of numerous different preprocessing steps (i.e., fMRI “pipeline”) that are designed to identify and reduce the effects of noise and artifacts of fMRI data prior to their analysis. However, choosing the optimal sequence of preprocessing steps continues to be a significant challenge, as they are rarely quantitatively validated in the neuroimaging literature due to the complex interactions between preprocessing algorithms. Too conservative of an approach reduces statistical power, while too flexible of an approach increases the likelihood of biased results.
In this literature review, we explore the different pipeline methodologies implemented in previous fMRI studies of cognition in older adults and identify reliable methods of fMRI preprocessing.