The magnetic properties of venous blood allow veins to be directly imaged with MRI, without an extrinsic contrast agent, using susceptibility-weighted imaging¹ (SWI) and, more recently, quantitative susceptibility mapping² (QSM). These techniques are sensitive to both small and large veins, and are minimally invasive, making them suitable for studies in healthy and diseased populations. Automated venography methods for processing these scans are advancing, reducing the resources required to analyse these images in large populations. With large populations, sophisticated imaging, and automated processing methods, it is possible to characterise the venous physiology and, with longitudinal data, identify biomarkers of prognostic and diagnostic value. In this study we take the first steps towards these goals by reporting vein density and venous blood deoxygenation in a healthy elderly population.

150 healthy elderly subjects (age > 70 years) were recruited, as part of the ASPREE NEURO sub-study, and scanned on a 3T Siemens Skyra with a 32-channel head and neck coil. A single echo, fully flow compensated, GRE sequence was used (TE=20ms, TR=30ms, Voxel=0.9x0.9x1.8mm³, Matrix=256x232x72, FA=15). Standard SWI images were obtained directly from the Siemens console (IDEA version VD13A), and raw k-space data was saved.

Laplacian unwrapping and V-SHARP were used for background field removal and phase unwrapping on individual coil images³. Coil images were combined using a sensitivity-weighted sum. QSM was calculated using STI Suite. Whole brain vein segmentation (example in Figure 1) was performed using a shape-based Markov Random Field (ShMRF) technique⁴, which combines both SWI and QSM using an anatomic prior.

Brain regions were extracted by running the freesurfer recon_all pipeline and combining sub-regions into lobes⁵. Venous density was calculated for each region as the sum of venous voxels divided by the sum of all voxels in the region. Oxygen-extraction fraction (OEF) was calculated using Equation [1] where $$$\Delta\chi_{tissue-vein}$$$ is the maximum venous QSM in the region, $$$Hct$$$ is hematocrit and $$$\chi_{do}$$$ is the magnetic susceptibility of fully deoxygenated haemoglobin. Published values⁶ were used for $$$Hct$$$ (0.4) and $$$\chi_{do}$$$ (0.264ppm).

$$OEF = \Delta\chi_{tissue-vein} \cdot 4 \pi \cdot \chi_{do} \cdot Hct \>\>\>\>\>\>\>\>\>[1]$$

The venous density (Figure 2, Table 1) was found to be different across many anatomic areas. In the cortical region, the frontal and parietal lobes were found to have consistently lower venous density (mean 4.25% and 4.06% respectively) than the temporal lobe, occipital lobe and cingulate cortex (9.06%, 9.02% and 7.39% respectively). There was no significant difference found between the left and right hemispheres. In the sub-cortical region the values were far less orderly, with lower volume resulting in higher standard deviations.

OEF was found to be significantly higher in the caudate, thalamus and occipital cortex, and lower in the frontal lobe, parietal lobe and putamen (Figure 3, Table 1). OEF was higher for brain regions, which are known to be active (occipital lobe, thalamus) when the brain is at rest or the subject is observing passively.

Venous density measurements and oxygen extraction fraction derived from magnetic susceptibility imaging have been recently proposed as imaging biomarkers. Our results indicate that while the venous density can be variable across the brain and across subjects, it may potentially be useful as a biomarker for neurovascular disease.

TheOEF results are also consistent with what is expected in regards to oxygen consumption in different parts of the brain when it is at rest. Additionally, OEF measurements normalised for red blood cell volume to account for physiological differences may provide increased sensitivity to detect neurovascular changes in disease.