Influence of systemic low-frequency oscillations on fMRI identifiability -Implications for denoising and...

Systemic low-frequency oscillations (sLFOs) can significantly influence blood oxygen level-dependent (BOLD) signals in functional magnetic resonance imaging (fMRI), confounding neural activity assessments. This study evaluates the impact of sLFO correction on functional connectivity-based identifiability using data from the Human Connectome Project. We compare global signal regression (GSR) and partial global signal regression (pGSR), which consists in removing the portion of the global signal explained by peripheral recordings (cardiac and breathing), in their ability to isolate sLFOs while preserving neural low-frequency oscillations (nLFOs). Our results show that GSR achieves significantly higher identifiability than the null model and pGSR suggesting improved sLFO removal based on a previous report that identifiability is primarily driven by the neural component as opposed to the measurable noise components. Our findings support the use of GSR as an effective sLFOs denoising strategy in fMRI fingerprinting and connectivity studies. In situations where less aggressive denoising is required, we propose pGSR as an alternative to GSR whenever the partial global signal can be estimated accurately from the peripheral recordings.