Cerebral autoregulation (CA) refers to the brain’s ability to maintain constant perfusion (CBF) in the presence of fluctuations in mean arterial pressure (MAP). Lower body negative pressure (LBNP) is a means of inducing orthostatic stress by way of central hypovolemia (reduced central blood volume). Studies have demonstrated that MAP is not reduced during LBNP due to peripheral vasoconstriction; yet significant reductions in blood flow velocity (CBFv) measured in large cerebral arteries using Transcranial Doppler (TCD) still occur [1].

Thus, the exact mechanisms of CA are obscure, but MRI methods provide an opportunity for more comprehensive blood flow measurements. We perform an orthostatic challenge in healthy humans using a custom made MRI compatible LBNP chamber (Fig. 1), and use multi-TI ASL to infer changes in arterial cerebral blood volume (aCBV).

A 3T GE HDx scanner equipped with an eight-channel receiver head coil was used to acquire data from eight subjects (all male) using a Pulsed Arterial Spin Labelling (PASL) sequence with a gradient-echo spiral readout and multiple short inversion times (TIs =150,300,450,600ms, TE=3ms, TR=variable 64x64 matrix, 12 slices, 20cm tag width, 1cm tag/slice gap, 40 tag/control pairs per TI). Subjects were placed in a custom built LBNP chamber and data collected for a baseline run (0mmHg, i.e. atmospheric pressure) and a negative pressure run (-40 mmHg, below atmospheric pressure). Separate scans were acquired to estimate arterial blood equilibrium magnetization (M_0a) from M₀ of CSF. Concurrent beat-to-beat blood pressure (Caretaker, BIOPAC) and end-tidal partial pressure CO2 (P_ETCO₂) measurements were also acquired.

Volumes were registered to the first volume of the baseline run and difference image time-series (ΔM) were obtained via tag/control subtraction for all TIs. ΔM time-series were extracted from a large artery mask, and an intravascular signal model was fit to the data using a least-squares brute-force approach. The model assumes plug flow of tagged blood not yet perfused into tissue and includes bolus arrival time (BAT) and aCBV parameters [2](Fig. 2). An amended form of the model that accommodates dispersion of the tagged blood due to laminar flow [3] was also fit to each voxel time-series, and the Akaike information criterion was used to determine the best model (Fig. 2).

There was no significant change in either MAP (mean difference=0.68±9.23(SD)) or P_ETCO₂ (mean difference=-0.47±3.03(SD))) between baseline and -40mmHg LBNP. As seen in Fig. 1 there was a significant difference in the large artery mask averaged M_0a normalised signal between baseline and -40mmHg in the first and last TIs (p<0.01, Bonferroni corrected). Parameter estimates for the fitted models elucidate the physiological basis for this difference, and Fig. 3 shows the distribution of voxel BAT and aCBV values for baseline and -40mmHg, as well as the difference between conditions. Qualitatively, a change in the distribution of aCBV values can be observed Fig. 3B, and a linear model reveals a highly significant (p ~ 10^-11) effect of baseline aCBV on the change in aCBV between baseline and -40mmHg (Fig. 3D). No significant change was seen in a BAT linear model, but there was a trend (p<0.05, uncorrected) for an increase in BAT in the smallest vessels (aCBVbaseline=1%) during LBNP (Fig. 3C).

LBNP is a promising technique for obtaining MRI measurements of the cerebrovascular mechanisms of autoregulation, and their role in health and disease. These findings are consistent with the TCD literature showing reduced large artery CBFv during LBNP [1], but the additional small artery information rectifies the apparent contradictory nature of these reports with regard to CA, by suggesting downstream vasodilation in smaller arteries preserves CBF to the brain parenchyma during central hypovolemia. In the largest arteries a reduction in aCBV of approximately 40% was calculated, compared with an increase of approximately 100% in the smallest arteries.

The relatively coarse temporal resolution of our multi-TI data means that BAT and aCBV changes are only roughly estimated, yet this novel finding is encouraging for the use of MRI and LBNP as means to probe CA in further detail than previously afforded by TCD. Further LBNP experiments utilizing multi-TI ASL sequences with higher temporal resolutions, and including longer TIs to quantify perfusion to the capillary bed, will elucidate the relationship between systemic and cerebral blood volume state.