Deuterium Metabolic Imaging Denoising Using a Linear Tangent Space Alignment (LTSA) Model and Performance Analysis

Deuterium metabolic imaging (DMI) is an emerging technique to map metabolism of human body non-invasively. However, conventional DMI acquisition methods are limited by the signal-to-noise ratio (SNR) due to the small gyromagnetic ratio of deuterium and low concentration of deuteriumlabeled metabolites. This work presents a manifold learningbased method that leverages the intrinsic low-dimensional manifold structure of the underlying DMI signals for denoising via a linear tangent space alignment (LTSA) model. Cramer-Rao Lower Bound (CRLB) analysis was performed to characterize the reduction in noise and in the variance of the metabolite concentration estimation achieved by the proposed method, which was validated using numerical simulation studies.