Considerations for measuring the fractional anisotropy of metabolites with diffusion tensor spectroscopy

Diffusion tensor spectroscopy of metabolites in brain is challenging because of their lower diffusivity (i.e. less signal attenuation for a given b value) and much poorer signal-to-noise ratio relative to water. Although diffusion tensor acquisition protocols have been studied in detail for water, they have not been evaluated systematically for the measurement of the fractional anisotropy of metabolites such as N-acetylaspartate, creatine and choline in the white and gray matter of human brain. Diffusion tensor spectroscopy was performed in vivo with variable maximal b values (1815 or 5018  s/mm(2)). Experiments were also performed on simulated spectra and isotropic alcohol phantoms of various diffusivities, ranging from approximately 0.54 × 10(-3) to 0.13 × 10(-3) mm(2)/s, to assess the sensitivity of diffusion tensor spectroscopic parameters to low diffusivity, noise and b value. The low maximum b value of 1815  s/mm(2) yielded elevated fractional anisotropy (0.53-0.60) of N-acetylaspartate in cortical gray matter relative to the more isotropic value (0.25-0.30) obtained with a higher b value of 5018  s/mm(2); in contrast, the fractional anisotropy of white matter was consistently anisotropic with the different maximal b values (i.e. 0.43-0.54 for b = 1815  s/mm(2) and 0.47-0.51 for b = 5018  s/mm(2)). Simulations, phantoms and in vivo data indicate that greater signal attenuation, to a degree, is desirable for the accurate quantification of diffusion-weighted spectra for slowly diffusing metabolites.