The key findings of this study are that (1) transient systemic blood pressure changes are reflected in brain NIRS measurements, and (2) the coupling appears to take place in both extracranial tissues and the brain itself, with different patterns. These results signal the need to monitor and correct for blood pressure effects when brain NIRS is used to investigate cortical haemodynamics.
Arm-raising generated transient increases in systemic blood pressure. There are a number of possible mechanisms for this including the central command driving local and systemic facilitatory cardiovascular responses in anticipation and in response to effort (Smith et al., 2000
), and the sudden shift between two hydrostatic states, i.e. when the arm is lowered all its blood volume is below the level of the shoulders whereas the opposite holds when it is raised. The magnitude of blood pressure responses to arm-raising (~6 mmHg) was larger than those typically evoked by individual visual stimuli (~2 mmHg; e.g., Minati et al., 2009
), and closer to values expected for extended blocks of demanding task performance (~5 mmHg; e.g., Tachtsidis et al., 2008
). The response time was, however, similar to that of a typical haemodynamic response evoked by a short stimulus, confirming the suitability of this manipulation for studying the effect of blood pressure changes during an event-related task.
Shallow recordings show that arm-raising evokes increases in O2
Hb concentration which significantly correlate with MAP. These clearly demonstrate that transient blood pressure manipulation affects the oxygenation and perfusion of subcutaneous extracranial tissues. Since O2
Hb was increased in the absence of a corresponding HHb decrease, the observed effect most likely reflects enlarged blood volume, consequential to vascular elasticity, rather than increased blood flow (Nishiyasu et al., 1999; Strangman et al., 2002; Hachiya et al., 2008
In the absence of blood pressure changes induced by arm-raising, deep (i.e. cortical) NIRS recordings showed increases in O2
Hb and decreases in HHb following pattern-reversal chequerboard stimulation consistent with neurally coupled haemodynamic responses. The latency of the O2
Hb response was correspondingly sensitive to stimulus duration. However, blood pressure manipulation by arm-raising induced a positive O2
Hb response, which was more than twice as large as that obtained through visual stimulation alone. Here, the correlation between the O2
Hb response and MAP was even stronger than for shallow recordings, implying that the coupling was also present intra-cranially. This is in line with literature on autonomic control of brain vasculature (Zhang et al., 2002
). The finding of a weak negative HHb response, however, suggests that a different mechanism may be at play in the brain, i.e. that increased blood pressure elevates blood flow, leading to increased O2
Hb concentration and dilution of HHb, as commonly observed for cortical haemodynamic responses (Strangman et al., 2002
). This decoupling is expected, considering that regional blood volume is more tightly and rapidly regulated in the brain than in the muscles (Zhang et al., 2002; Banaji et al., 2008
). Accordingly, the correlation between the HHb response and MAP was weaker than that observed for shallow recordings.
In arm-raising trials, the response latency difference observed between 1500 and 3000 ms visual stimuli was not detectable and interestingly, the stimulus-evoked O2
Hb responses were flattened, suggesting that increased systemic MAP does not additively enhance neurally driven haemodynamic activation. Our observations are in line with empirical data regarding non-linearity and time-dependence of cerebrovascular responses and support the notion of multiple interacting systems involved in the control of regional cerebral perfusion (Banaji et al., 2008
This study has several limitations that need to be considered. First, as with most other NIRS studies the wavepath and penetration depth were assumed a priori, on the basis of existing literature. More solid indication of the attained penetration depth would need to be obtained using Monte Carlo photon diffusion simulations, based on the segmentation of the individual structural scans (e.g., Mansouri et al., 2010
). Second, the number of participants was relatively small, especially for the shallow-recordings group; as a consequence, the findings should be interpreted as preliminary. Third, we are unable to directly confirm whether the source of the effect is wholly extra-cranial, or whether an intra-cranial effect is present as well. To this end, the experiment needs to be replicated using time-resolved spectroscopy, which enables one to gate the measurements to a specific penetration depth range (Torricelli et al., 2008
In spite of these limitations, these results extend previous correlational findings by unequivocally demonstrating that blood pressure changes can exert strong confounding effects on brain NIRS measurements. Our findings have specific relevance to the implementation of NIRS as a functional neuroscience method. Moreover, since autonomically mediated neurovascular effects are also present within the brain itself (Zhang et al., 2002
), they may also have more general relevance to other techniques including fMRI that depend on neurally coupled haemodynamic responses evoked by stimuli or responses that may also generate systemic effects. There are, in fact, a number of signal processing methodologies that can be used to address the issue of systemic confounds. One approach, discussed by Tachtsidis et al. (2010)
, is to measure and explicitly model the physiological confounds, inserting them as nuisance regressors. An alternative, discussed in Katura et al. (2008)
, is to use separation and classification techniques such as independent component analysis (ICA) and k
-means clustering to isolate task-related from physiology-related activity components, purely on the basis of the fNIRS or fMRI time series.
More work is needed to explore the effect of smaller blood pressure manipulations and to characterize the response function mapping different magnitudes of blood pressure change to their effects on O2Hb and HHb concentrations. This will enable inclusion of blood pressure-derived measurements as nuisance covariate in statistical analyses. Our findings do not definitively clarify the relative contribution of extra- vs. intra-cranial coupling, and the experiment needs to be replicated with time-resolved spectroscopy. Relevant confirmation of our findings should also be obtained simultaneously recording from deep- and shallow-separation optodes, through a NIRS system with a larger number of channels. Further investigations are necessary to clarify to what extent MAP changes may affect tasks structured in longer blocks rather than individual events. Even before such data becomes available, in light of our observations blood pressure effects should always be characterized during functional NIRS experiments.