Global intravascular and local hyperoxia contrast phase-based blood oxygenation measurements

Abstract

The measurement of venous cerebral blood oxygenation (Yv) has potential applications in the study of patient groups where oxygen extraction and/or metabolism are compromised. It is also useful for fMRI studies to assess the stimulus-induced changes in Yv, particularly since basal Yv partially accounts for inter-subject variation in the haemodynamic response to a stimulus. A range of MRI-based methods of measuring Yv have been developed recently. Here, we use a method based on the change in phase in the MR image arising from the field perturbation caused by deoxygenated haemoglobin in veins. We build on the existing phase based approach (Method I), where Yv is measured in a large vein (such as the superior sagittal sinus) based on the field shift inside the vein with assumptions as to the vein's shape and orientation. We demonstrate two novel modifications which address limitations of this method. The first modification (Method II), maps the actual form of the vein, rather than assume a given shape and orientation. The second modification (Method III) uses the intra and perivascular phase change in response to a known change in Yv on hyperoxia to measure normoxic Yv in smaller veins. Method III can be applied to veins whose shape, size and orientation are not accurately known, thus allowing more localised measures of venous oxygenation. Results demonstrate that the use of an overly fine spatial filter caused an overestimation in Yv for Method I, whilst the measurement of Yv using Method II was less sensitive to this bias, giving Yv = 0.62 ± 0.03. Method III was applied to mapping of Yv in local veins across the brain, yielding a distribution of values with a mode of Yv = 0.661 ± 0.008.

Keywords: Hyperoxia; MR susceptometry; MRI; Oxygen extraction fraction (OEF); Oxygen saturation; Phase imaging.

Fig. 1

Example normoxia phase data after spatial filtering with a complex Hanning filter diameter of D = 16, D = 32 and D = 64 pixels in k-space.

Fig. 2

Example intravascular (red) and reference (blue) ROI, based on the manually drawn sagittal sinus ROI (green).

Fig. 3

The analysis steps for the forward field calculation. A manually drawn superior sagittal sinus (SSS) binary mask (b, green ROI) is used to form a field perturbation map (c), which is then spatial filtered (d) using the same filter as applied to the phase data (a). The intravascular (red) and reference (blue) ROIs in (d) are the same as in Fig. 2.

Fig. 4

Mean Y_v across subjects (SEM error bars) based on intravascular phase after spatial filtering with each of the three different complex Hanning filter widths. Results based on the infinite cylinder approximation (Method I) are shown in black and the forward field calculation (Method II) is shown in grey.

Fig. 5

Results for the hyperoxia phase contrast method (Method III). (a) Single subject map of Y_v, overlaid on the filtered (D = 64) normoxia phase data. Sagittal and axial slices are shown, whilst the green line shows the location of the slice position of the other orientation. (b) Histogram showing Y_v values for each vein across all subjects, where each vein is weighted by the number of voxels in that vein (blue histogram, left axis). Monte Carlo simulation results of this method are also shown (white histogram, right axis).