Impact of time sample selection and model function design on the quantification of fatty acid composition: in vitro and in vivo studies.
Résumé
SynopsisInterest in the follow-up of fatty acid composition(saturated, polyunsaturated and monounsaturated fatty acid) in the body is growing. Quantitative MR spectroscopy can give access to this fat composition. Today, severalquantification methods are used (e.g LCModel, AMARES). However the statistical outcome issued from a quantitative analysis of the lipid signal can be greatly influenced by the used quantification method. We analyze 1) the impactof the time sample selection and design of the model function on the parameter identifiability 2) the quantification results obtained with different quantification models on acquisitions performed in vitro (oils) and in vivo(subcutaneous adipose tissue).PurposeInterest in the fat quantification has grown in recent years. In particular, the nutrition studies are interested in the variation of fatty acid composition in the body: saturated fatty acid (SFA), polyunsaturated (PUFA) and monounsaturated(MUFA). Magnetic resonance spectroscopy (MRS) is a quantitative technique that gives access to this fat composition. At this time, several quantification methods, primarily developed for brain MRS analysis, are used (e.g LCModel, AMARES).However, it has been recently described that, in the case of lipid signal, the statistical outcome issued from a quantitative analysis can be greatly influenced by the used quantification method .In the present work, we investigate the particular case of lipid quantification in adipose tissue; for which the lipid components are predominant and we study the impact of the time sample selection and design of the model function on the parameter identifiability (given by the condition number of the Jacobian Matrix ) and on the parameter relative Cramer Rao Lower Bound (CRLB). These two criteria give insight about the ability of the parameter estimation to provideprecise inference. We then analyze the quantification results obtained with different quantification models- adding results from the reference LCModel method- on in vitro oil and in vivo subcutaneous adipose tissue (AT) acquisitions