Automatische Quantifizierung von Metabolitenkonzentrationen in :i:in vivo:/i: Spektren

Abstract

:p:A number of dedicated quantification tools have been developed for the interpretation of :i:in vivo:/i: magnetic resonance spectroscopy (MRS) data. To be useful in a clinical setting these tools have to meet the requirements of both a high level of automation and a general applicability in a wide range of clinical examinations. Existing tools put their main emphasis on one or other of these two aspects, but not on both simultaneously. The quantification tool of this work, PRISMA, was therefore developed to meet both of these requirements and to deliver reproducible and consistent results with good reliability. :p:The design of PRISMA allows on the one hand an easy adaptation to different applications due to a very flexible model parameterization. On the other hand it guaranties a robust and automatic result generation by an assumption-free assignment of the metabolites. For the baseline a new model is introduced which is based on a finite time signal. :p:The robust behavior of the PRISMA quantification is demonstrated on diverse :sup:1:/sup:H MRS data acquired from different anatomical regions. Besides synthetic data signals from brain and prostate measurements of healthy and pathological tissue were analyzed. The measurements were done at varying field strength, echo times and signal to noise ratios using single voxel and multi voxel techniques (SVS, SI). In addition, typical characteristics of PRISMA such as reproducibility and reliability were investigated and found to compare well with the performance of other quantification tools.