Unlike ionizing imaging modalities, the SNR in MRI is proportional to voxel volume, but downsampling or voxel averaging after acquisition only improves SNR by the square root of the voxel volume. To take advantage of this distinction, we use a frequency selective imaging approach to independently excite the hyperpolarized 13C substrate (pyruvate) and downstream metabolites (lactate, alanine, and bicarbonate). This allows us to tailor the spatial resolution for each metabolic product, yielding high-resolution images for pyruvate as well as quantification at a coarser resolution for the lower SNR metabolites, such as bicarbonate, which would be undetectable at the higher resolution.
Hyperpolarized (HP) [1-13C]pyruvate and other substrates have been used to non-invasively image metabolism in cancer [1], diabetes [2] and heart disease [3]. Despite a 10,000-fold increase in polarization, SNR can often be the limiting factor when imaging metabolic products – such as 13C bicarbonate for C1-pyruvate, or TCA metabolites for C2-pyruvate - due to slow transport across the cell membrane, low rates of conversion, and potential T1 differences. This results in a 10 to 100-fold difference in SNR between pyruvate and metabolites [4], yielding a mismatch in optimal spatial resolutions. In particular, high resolution is especially important for pyruvate in order to minimize partial-volume effects between the vasculature and extravascular/extracellular spaces, but is unsupportable for lower SNR metabolites.
Unlike ionizing imaging modalities, the SNR in MRI is proportional to voxel volume but downsampling or voxel averaging after acquisition only improves SNR by $$$\sqrt{voxel \ volume}$$$. There is therefore an SNR benefit to acquiring data at a coarser resolution rather than downsampling after acquisition. To explore this potential, in this work we use a frequency selective imaging approach to independently excite each metabolite of interest. This allows us to tailor the spatial resolution for each metabolic product, yielding high-resolution images for pyruvate and coarser resolution for downstream metabolites while maintaining adequate SNR for each.
The advantages of the variable resolution approach can be seen in the dynamic timecourse of pyruvate metabolism in Fig. 1 and 2. At the constant 2.5×2.5mm in-plane resolution (Fig. 1), bicarbonate was not visible on the dynamic images and only slightly above the noise floor (SNR = 4.8) in the area under the curve image (Fig. 3). In contrast, the frequency selective imaging approach allowed for independent resolutions for each metabolite. At 3-fold larger voxel size (7.5×7.5mm in-plane resolution), conversion to bicarbonate was clearly visible on both the dynamic timecourse (Fig. 2; peak SNR = 8.8) and AUC maps (Fig. 3; SNR = 30.4). Overall, the peak SNR improved by 3.6-fold for lactate, 7.8-fold for alanine, and 7.2-fold for bicarbonate (Table 1), in good agreement with the expected SNR increase of 4 and 9, respectively.
By using the same readout waveform and bandwidth, each metabolite map had the same TE, noise statistics, and echo-spacing (i.e. sensitivity to off-resonance artifacts). Quantification was not impacted, as signal differences from larger voxels are readily accounted for by a scale factor of 1/(voxel size)2. This was borne out by the lactate-to-pyruvate AUC ratio, which was not statistically different (0.12 ± 0.06 and 0.12 ± 0.03, respectively; p = 0.93) for the constant (2.5×2.5mm) and variable (5×5mm) acquisition.