Spatial and Subject Variability of Long-term Respiration Effects in fMRI

INTRODUCTION Physiological fluctuations due to respiration are the dominant source of confounding variability in BOLD fMRI at high fields [1-4]. Short-term respiration effects result from B0 modulation due to bulk susceptibility changes from thoracic organ movement and gas volume as well as head motion [3-5]. Recently low-frequency fluctuations in the depth and the rate of respiration have been identified as a source of long-term respiration effects, mediated by the carbon dioxide content of arterial blood [1,2]. Long-term effects are particularly problematic in resting-state functional connectivity analysis since their frequency range overlaps with the frequencies of fluctuations believed to reflect resting brain activity [1,6]. Therefore, it is vital to precisely characterize the impact of breathing on the BOLD signal. Respiration volume per time (RVT) and respiration response function (RRF) have been proposed as global models of respiration effects [1,7]. The global models implicitly assume that the respiration effects are space-invariant and subject-invariant. This assumption is probably not strictly met. A technique was previously introduced to estimate voxelspecific physiological impulse response function (PIRF) for characterizing the impact of respiration on the BOLD signal [8]. The goal of this study is to examine space and subject variability of long-term respiration effects by clustering the voxel-specific PIRFs.