Ferumoxytol is an FDA approved therapeutic iron-oxide agent for treatment of iron deficiency. Off-label use as an MR contrast agent for both vascular and parenchymal enhancement is being explored. This agent has a much longer vascular half-life than do gadolinium-based contrast agents, allowing the opportunity for detailed visualization or quantification of the vasculature. A recent report quantified the relative cerebral blood volume (rCBV) by comparing T₂* weighted images before and after administration of this⁽¹⁾ This prior work employed a single echo acquisition and was most sensitive to contrast agent concentrations (and thus rCBV values) resulting in moderate signal decay. Higher rCBV (too much signal decay) or lower rCBV (too little signal decay) were difficult to detect. Here, we present a multi-echo acquisition spanning a wide range of echo times and an optimized quantification strategy to effectively estimate rCBV from multi-echo data.

The rCBV is linearly related to the change in transverse relaxation rate (ΔR₂*) following administration of Ferumoxytol. This latter parameter can be quantified directly from pre- and post-contrast signal intensities (SIpre and SIpost, respectively) without data fitting:

$$\Delta R_{2}^\star = \frac{1}{\text{TE}} \cdot \ln \left(\frac{\text{SI}^\text{pre}}{\text{SI}^\text{post}} \right)$$

Assuming uncorrelated noise with power σ² in both scans, noise power in the quantitative map is given by:

$$\sigma^2_{\left(\Delta R_2^\star\right)} = \frac{1}{\text{TE}^2} \left[\left(\frac{1}{\text{SI}^\text{pre}}\right)^2 + \left(\frac{1}{\text{SI}^\text{post}}\right)^2 \right]$$

A multi-echo acquisition with N echoes provides N estimates of ΔR₂* , each with a different noise power based on the echo time and signal intensities. These maps provide the same information and can be combined for reduced variance. We propose a minimal-noise image combination that involves normalizing by noise power then averaging maps:

$$\Delta R_2^\star = \sum_i^\text{N} \; \frac{\Delta R_{2_i}^\star}{\sigma_i^2} \; / \; \sum_i^\text{N} \; \frac{1}{\sigma_i^2}$$

This weighted average, when applied on a voxel-by-voxel basis, is expected to provide an effective means of accurately estimating a wide range of rCBV values. Additionally, this strategy should enable measurements in high susceptibility areas, where a single echo acquisition or non-optimized combination would fail.

2D multi-echo spoiled gradient echo images were acquired before and after administration of Ferumoxytol in a patient with suspected Glioblastoma multiforme. Five echo times between 6.9 and 29 ms were acquired. Quantitative maps were computed at each echo time and were combined with the proposed method.Quantitative ΔR₂* maps computed at each echo time show differential noise variance among structures and echo times (Figure 1), as predicted by the second equation. The highly vascular tumor is most effectively characterized at short echo times. At long echo times, the complete signal decay prevents characterization of the tumor (note that signal gets masked out when approaching the noise floor). Normal appearing white and gray matter are best quantified at longer echo times due to the slower signal attenuation. Optimal image combination favors short echo times in the tumor (blue lines) while emphasizing longer echo times in gray matter (red lines). A composite image, optimized on a voxel-by-voxel basis, takes the best parts of individual echo times, making it sensitive to a high dynamic range of rCBV while minimizing noise and artifact.A multi-echo acquisition is a promising approach for high resolution CBV mapping with ferumoxytol. Optimized reconstruction is expected to improve the linearity, accuracy, and precision of quantitative maps relative to standard reconstruction and relative to single echo acquisitions. This approach is being investigated for high resolution quantification of tumor vascularity and outcome prediction.