Calibrated BOLD using direct measurement of changes in venous oxygenation

Abstract

Calibration of the BOLD signal is potentially of great value in providing a closer measure of the underlying changes in brain function related to neuronal activity than the BOLD signal alone, but current approaches rely on an assumed relationship between cerebral blood volume (CBV) and cerebral blood flow (CBF). This is poorly characterised in humans and does not reflect the predominantly venous nature of BOLD contrast, whilst this relationship may vary across brain regions and depend on the structure of the local vascular bed. This work demonstrates a new approach to BOLD calibration which does not require an assumption about the relationship between cerebral blood volume and cerebral blood flow. This method involves repeating the same stimulus both at normoxia and hyperoxia, using hyperoxic BOLD contrast to estimate the relative changes in venous blood oxygenation and venous CBV. To do this the effect of hyperoxia on venous blood oxygenation has to be calculated, which requires an estimate of basal oxygen extraction fraction, and this can be estimated from the phase as an alternative to using a literature estimate. Additional measurement of the relative change in CBF, combined with the blood oxygenation change can be used to calculate the relative change in CMRO(2) due to the stimulus. CMRO(2) changes of 18 ± 8% in response to a motor task were measured without requiring the assumption of a CBV/CBF coupling relationship, and are in agreement with previous approaches.

Fig. 1

(a) A schematic diagram of R_2,rest* and R_2,act* plotted against venous blood oxygenation, indicating the BOLD signal change on normoxia (open arrow) and hyperoxia (solid arrow). The linear fits shown are used to calculate rvCBV and q_act as described in the Theory section. (b) Simulated relationship between SaO₂ and SvO₂ (= 1 − Q) and PaO₂. The broken lines indicate arterial and venous oxygen saturation for 110 mm Hg (~ 21% O₂, normoxia) and 500 mm Hg (~ 60% O₂). Values used in this simulation are PaO_2,0 = 110 mm Hg, OEF = 0.4, ϕ = 1.34 ml(O₂)/g, [Hb] = 15 g/dl_blood and ε = 0.0031 ml/(dl_blood·mm Hg).

Fig. 2

An illustration of the combined hyperoxia and motor task for (A) the 30 s ON/30 s OFF paradigm (Paradigm A) and (B) the 30 s ON/60 s OFF paradigm (Paradigm B). The motor task is shown in grey. Trials marked with an * were used in the analysis for (B).

Fig. 3

Example data illustrating the phase-based calculation of Q₀. Average (a) normoxia and (b) hyperoxia phase maps (range − 1 to 1 rad), with the selected line profile position highlighted in red. (c) Line profiles for normoxia (black) and hyperoxia (red). The fitted ratio a of hypercapnia phase to normocapnia phase is shown for this subject.

Fig. 4

Example timecourses from a single subject for (a) BOLD, (b) P_ETO₂ and (c) P_ETCO₂ (BOLD motor activation mask).

Fig. 5

(a) Average motor trial %BOLD timecourses during hyperoxia (red dashed line) and normoxia (black line). (b) Example of the linear fit between R₂* and (1 + q_h). (c) The %CBF response to the motor task, averaged over trials. Data from subject 8, formed from the ‘combined mask’ (the intersection of BOLD and CBF motor activation masks).

Fig. 6

Voxelwise maps of M, rvCBV and q_act across the combined mask (the intersection of BOLD and CBF motor activation masks).