Evaluating Models of the Ageing BOLD Response.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Aging Research

Background:

• Functional magnetic resonance imaging (fMRI) infers neural activity from hemodynamic responses.
• Existing hemodynamic models often focus on young, healthy populations, limiting applicability to aging and dementia studies.
• Validating and comparing different hemodynamic models across the adult lifespan is crucial for accurate interpretation of fMRI data in diverse populations.

Purpose of the Study:

• To evaluate the validity of various hemodynamic models across the healthy adult lifespan.
• To compare descriptive features of the Blood-Oxygen-Level-Dependent (BOLD) response with parameters from nonlinear and biophysical models.
• To assess the predictive validity of model parameters for estimating age.

Main Methods:

• Utilized a large sample size and a sensorimotor task optimized for BOLD response detection.
• Characterized age-related effects on BOLD response features (e.g., peak amplitude, latency).
• Compared linear convolution models, nonlinear HRF fitting, and biophysical generative models (HDM) in key brain regions (auditory, visual, motor cortices).
• Employed cross-validated multiple regression to test the predictive validity of model parameters for age.

Main Results:

• Models with three free parameters effectively capture age-related BOLD response differences.
• Biophysical models (HDM) demonstrate comparable predictive validity to common models.
• HDM suggests age effects on BOLD are primarily due to altered vasoactive signal decay and blood transit rates, not solely neural changes.
• HDM identified specific physiological mechanisms underlying age-related BOLD signal variations.

Conclusions:

• Hemodynamic models with three parameters adequately represent age-related BOLD response changes.
• Biophysical models provide mechanistic insights beyond descriptive features, enhancing understanding of aging effects on brain function.
• Age-related BOLD signal alterations are largely explained by physiological factors (vasoactivity, blood flow) rather than intrinsic neural activity changes.
• While HDM offers valuable mechanistic information, unique parameter interpretation from fMRI alone remains challenging, especially in certain brain regions or task contexts.