Assessment of higher-order singular value decomposition denoising methods on dynamic hyperpolarized [1-(13)C]pyruvate MRI data from patients with glioma

BACKGROUND: Real-time metabolic conversion of intravenously-injected hyperpolarized [1-(13)C]pyruvate to [1-(13)C]lactate and [(13)C]bicarbonate in the brain can be measured using dynamic hyperpolarized carbon-13 (HP-(13)C) MRI. However, voxel-wise evaluation of metabolism in patients with glioma is challenged by the limited signal-to-noise ratio (SNR) of downstream (13)C metabolites, especially within lesions. The purpose of this study was to evaluate the ability of higher-order singular value decomposition (HOSVD) denoising methods to enhance dynamic HP [1-(13)C]pyruvate MRI data acquired from patients with glioma. METHODS: Dynamic HP-(13)C MRI were acquired from 14 patients with glioma. The effects of two HOSVD denoising techniques, tensor rank truncation-image enhancement (TRI) and global–local HOSVD (GL-HOSVD), on the SNR and kinetic modeling were analyzed in [1-(13)C]lactate data with simulated noise that matched the levels of [(13)C]bicarbonate signals. Both methods were then evaluated in patient data based on their ability to improve [1-(13)C]pyruvate, [1-(13)C]lactate and [(13)C]bicarbonate SNR. The effects of denoising on voxel-wise kinetic modeling of k(PL) and k(PB) was also evaluated. The number of voxels with reliable kinetic modeling of pyruvate-to-lactate (k(PL)) and pyruvate-to-bicarbonate (k(PB)) conversion rates within regions of interest (ROIs) before and after denoising was then compared. RESULTS: Both denoising methods improved metabolite SNR and regional signal coverage. In patient data, the average increase in peak dynamic metabolite SNR was 2-fold using TRI and 4–5 folds using GL-HOSVD denoising compared to acquired data. Denoising reduced k(PL) modeling errors from a native average of 23% to 16% (TRI) and 15% (GL-HOSVD); and k(PB) error from 42% to 34% (TRI) and 37% (GL-HOSVD) (values were averaged voxelwise over all datasets). In contrast-enhancing lesions, the average number of voxels demonstrating within-tolerance k(PL) modeling error relative to the total voxels increased from 48% in the original data to 84% (TRI) and 90% (GL-HOSVD), while the number of voxels showing within-tolerance k(PB) modeling error increased from 0% to 15% (TRI) and 8% (GL-HOSVD). CONCLUSION: Post-processing denoising methods significantly improved the SNR of dynamic HP-(13)C imaging data, resulting in a greater number of voxels satisfying minimum SNR criteria and maximum kinetic modeling errors in tumor lesions. This enhancement can aid in the voxel-wise analysis of HP-(13)C data and thereby improve monitoring of metabolic changes in patients with glioma following treatment.