Covariance-based vs. correlation-based functional connectivity dissociates healthy aging from Alzheimer disease

Area of Science:

• Neuroscience
• Neuroimaging
• Alzheimer Disease Research

Background:

• Resting state functional connectivity (FC) studies often use correlation-based methods (e.g., SBC, ICA) insensitive to signal amplitude.
• Spontaneous BOLD signal amplitude is increasingly recognized as physiologically significant.
• Previous research has not fully differentiated the effects of aging and Alzheimer disease on FC amplitude and structure.

Purpose of the Study:

• To systematically compare covariance-based FC (amplitude-sensitive) and correlation-based FC (amplitude-insensitive) in aging and Alzheimer disease.
• To investigate the distinct impacts of healthy aging, autosomal dominant Alzheimer disease (ADAD), and late-onset Alzheimer disease (LOAD) on resting state BOLD signal characteristics.
• To elucidate the relationship between signal amplitude and correlation structure in the context of aging and Alzheimer disease.

Main Methods:

• Functional connectivity was computed across 222 regions of interest using both covariance and correlation methods.
• Group differences were analyzed by projecting components onto a lower-dimensional space.
• Two cohorts were studied: individuals with ADAD, LOAD, and age-matched controls.

Main Results:

• Healthy aging is associated with a global, uniform decrease in resting state fMRI signal amplitude, leading to reduced covariance FC but preserved correlation FC.
• Symptomatic ADAD and LOAD exhibit a significant loss of spontaneous activity amplitude.
• Both ADAD and LOAD demonstrate severely degraded correlation structure in resting state BOLD signals.

Conclusions:

• A double dissociation exists between aging and Alzheimer disease concerning signal amplitude and correlation structure.
• Aging primarily affects signal amplitude, while Alzheimer disease impacts both amplitude and the spatial organization of functional connectivity.
• The degradation of correlation structure in Alzheimer disease may stem from an increased proportion of locally restricted variance compared to widely shared variance.