Reconstruction of randomly under-sampled spectra for in vivo¹³C magnetic resonance spectroscopy

Over the past decade, many techniques have been developed to reduce radiofrequency (RF) power deposition associated with proton decoupling in in vivo Carbon-13 (¹³C) magnetic resonance spectroscopy (MRS). In this work we propose a new strategy that uses data under-sampling to achieve reduction in RF power deposition.

Essentially, proton decoupling is required only during randomly selected segments of data acquisition. By taking advantage of the sparse spectral pattern of the carboxylic/amide region of in vivo¹³C spectra of brain, we developed an iterative algorithm to reconstruct spectra from randomly under-sampled data. Fully sampled data were used as references. Reconstructed spectra were compared with the fully sampled references and evaluated using residuals and relative signal intensity errors.

Numerical simulations and in vivo experiments at 7Tesla demonstrated that this novel decoupling and data processing strategy can effectively reduce decoupling power deposition by greater than 30%.

This study proposes and evaluates a novel approach to acquire ¹³C data with reduced proton decoupling power deposition and reconstruct in vivo¹³C spectra of carboxylic/amide metabolite signals using randomly under-sampled data. Because proton decoupling is not needed over a significant portion of data acquisition, this novel approach can effectively reduce the required decoupling power and thus SAR. It opens the possibility of performing in vivo¹³C experiments of human brain at very high magnetic fields.