HyPER: Region-specific hypersampling of fMRI to resolve low-frequency, respiratory, and cardiac pulsations, revealing age-related differences

Abstract

Resting-state functional MRI (fMRI) signals capture physiological processes, including systemic low-frequency oscillations (LFOs), respiration, and cardiac pulsations. These physiological oscillations—often treated as noise in functional connectivity analysis—reflect fundamental aspects of brain physiology and have recently been recognized as key drivers of brain waste clearance. However, these critical physiological signals are obscured in fMRI data due to slow sampling rates (typical repetition time (TR) > 0.8 s), which cause cardiac signal to alias into lower frequencies. To resolve physiological signals in fMRI datasets, we leveraged fast cross-slice sampling within each TR to hypersample the fMRI signal. A key novelty of this study is the development of a region-specific hypersampling approach, called HyPER (Hypersampling for Physiological signal Extraction in a Region-specific manner). HyPER enhances temporal resolution within coherently pulsating vascular and tissue compartments, including the major cerebral arteries, the superior sagittal sinus (SSS), gray matter (GM), and white matter (WM). This study is structured in three parts: (1) We developed and validated the HyPER approach using fast fMRI from a local dataset in four regions of interest: the major cerebral arteries, SSS, GM, and WM. (2) We applied this approach to the publicly available Human Connectome Project-Aging (HCP-A) dataset (ages 36–90 years), increasing the resolvable frequency by ninefold—from 0.625 Hz to 5.625 Hz—enabling clear separation of cardiac, respiration, and LFO oscillations. (3) We investigated how brain physiological pulsations change with age. Our findings revealed an age-related increase in cardiac and respiratory pulsations across all brain regions, likely reflecting an increased vessel stiffness and reduced dampening of high-frequency pulsations along the vascular network. In contrast, LFO pulsations generally declined with age, suggesting reduced vasomotion in the older brain. In summary, we demonstrated the feasibility and reliability of a region-specific hypersampling technique to resolve physiological pulsations in fMRI. This method can be broadly applied to existing fMRI datasets to uncover hidden physiological pulsations and advance our understanding of brain physiology and disease-related alterations.