Background: Post processing for brain spectra has a great influence on the fit quality of individual spectra, as well as on the reproducibility of results from comparable spectra. This investigation used pairs of spectra, identical in system parameters, position and time assumed to differ only in noise. The metabolite amplitudes of fitted time domain spectroscopic data were tested on reproducibility for the main brain metabolites.
Methods: Proton spectra of white matter brain tissue were acquired with a short spin echo time of 30 ms and a moderate repetition time of 1500 ms at 1.5 T. The pairs were investigated with one time domain post-processing algorithm using different parameters. The number of metabolites, the use of prior knowledge, base line parameters and common or individual damping were varied to evaluate the best reproducibility.
Results: The protocols with most reproducible amplitudes for N-acetylaspartate, creatine, choline, myo-inositol and the combined Glx line of glutamate and glutamine in lesion free white matter have the following common features: common damping of the main metabolites, a baseline using only the points of the first 10 ms, no additional lipid/macromolecule lines and Glx is taken as the sum of separately fitted glutamate and glutamine. This parameter set is different to the one delivering the best individual fit results.
Discussion: All spectra were acquired in "lesion free" (no lesion signs found in MR imaging) white matter. Spectra of brain lesions, for example tumors, can be drastically different. Thus the results are limited to lesion free brain tissue. Nevertheless the application to studies is broad, because small alterations in brain biochemistry of lesion free areas had been detected nearby tumors, in patients with multiple sclerosis, drug abuse or psychiatric disorders.
Conclusion: Main metabolite amplitudes inside healthy brain can be quantified with a normalized root mean square deviation around 5 % using CH3 of creatine as reference. Only the reproducibility of myo-inositol is roughly twice as bad. The reproducibility should be similar using other references like internal or external water for an absolute concentration evaluation and are not influenced by relaxation corrections with literature values.