In vivo comparison of MRI‐based and MRS‐based quantification of adipose tissue fatty acid composition against gas chromatography

Abstract

Purpose To compare MR‐based fatty acid composition (FAC) quantification methods against the gold standard technique, gas chromatography (GC), with comparison of a free and a constrained signal model. The FAC was measured in the healthy and edematous legs of lymphedema patients. Methods In vivo MRS and MRI data were acquired from 19 patients at 3 T. Biopsies were collected from subcutaneous adipose tissue of both thighs during liposuction. The saturated, monounsaturated, and polyunsaturated fatty acid fractions (fSFA, fMUFA and fPUFA, respectively) were estimated with the MR‐based methods using two signal models: free and constrained (number of methylene‐interrupted double bonds expressed in number of double bonds, based on GC data). Linear regression, Bland–Altman plots, and correlation coefficients were used to evaluate the MR methods against the GC of the biopsies. Paired t‐test was used to compare the FAC difference between edematous and healthy legs. Results The estimated parameters correlated well with the GC data (rSFA, rMUFA, and rPUFA = 0.82, 0.81 and 0.89, respectively) using the free model MRI‐based approach. In comparison, the MRS‐based method resulted in weaker correlations and larger biases compared with MRI. In both cases, correct estimation of fMUFA and fPUFA fractions were not possible using the constrained model. The difference in FAC of healthy and edematous legs were estimated to 0.008 (P = .01), −0.009 (P = .005), and 0.002 (P = .03) for fSFA, fMUFA, and fPUFA. Conclusion In this study, MRI‐based FAC quantification was highly correlated, although slightly biased, compared with GC, whereas the MRS‐based approach resulted in weaker correlations. Small but significant differences could be found between the healthy and edematous legs of lymphedema patients using GC analysis.

Quantitative 1H MRI and MRS of fatty acid composition

“…58 Similar to the GC results for the same subject group, the MRI approach was able to detect a significantly higher proportion of SFA and lower fraction of MUFA in the edematous tissue, whereas the results for PUFA were contradictive between the two methods. 58…”

“…Thus, models based on human subcutaneous adipose tissue may be more suitable. 58 However, a recent study concludes that a free estimation of the ndb and nmidb may be necessary for accurate distinction between MUFA and PUFA. 58 Due to the low inter-subject variability of the cl this parameter is often fixed or restricted, 21,58,59 likely without any substantial impact on estimation accuracy.…”

“…Several publications have investigated the performance of imaging-based FAC estimation using both phantoms 20,22,59,60 and in vivo experiments. 20,51,58,60,66,67 High in vitro reproducibility has been demonstrated, 20 with accuracy comparable to that of MRS. 22 Although the method, in theory, allows for simultaneous estimation of fat content and FAC also in the presence of water signal, the accuracy and robustness of the technique was found to be lower in water/fat mixtures in vitro. 21,22 Example phantom images of a range of vegetable oils are shown as Figure 4.…”

“…The correlation between the GC and MRI methods was strong for all three parameters. 58 Out of the three descriptive parameters, ndb appears to be the most reliant as the estimates of nmidb and especially cl have been reported to be less accurate and robust. 20,22,59 Similarly, the test-retest reliability of f SFA was found "good" to "excellent", but only "moderate" to "good" of the f MUFA and f PUFA estimates, 60 which can be explained by f MUFA and f PUFA being derived from nmidb (cf.…”

“…21 However, as the nmidb is expressed as a quadratic function of ndb, this approach is not compatible with the iterative least-squares method most often used for fat/water separation, but requires an alternative curve-fitting technique. Alternatively, linear expressions may be valid in limited ranges of FAC parameters, 57 as those recently derived from GC data of human subcutaneous adipose tissue by Trinh et al 58 Using these models to constrain the reconstruction to estimation of ndb only, an improved image quality of the resulting f SFA , f MUFA , and f PUFA maps was obtained, at the cost of a substantially lower accuracy of the f MUFA parameter. 58 The results of this study indicate that a free estimation of ndb and nmidb may be necessary for a reliable distinction between f MUFA and f PUFA .…”