For single-voxel spectroscopy, Hamilton et al.¹ proposed to model the triglyceride ¹H MR spectrum using three free parameters: fatty acid chain length (CL), number of double bonds per triglyceride (ndb), and number of methylene-interrupted double bonds per triglyceride (nmidb). The purpose of this work was to develop an imaging method for mapping Fatty Acid Composition using only ndb as a free parameter (FAC1) and evaluate the performance compared to Fatty Acid Composition using all three parameters (FAC3).

Chemical shift based water/fat separation methods with simultaneous estimation of B₀ inhomogeneity and R₂* can be extended by modeling four spectral fat components rather than one^2,3. From these four components, CL, ndb, and nmidb can be estimated directly. We modified the FAC3 method of Berglund et al.² into a FAC1 method by simply assigning CL=17.4 and nmidb=0.2ndb (coefficients calculated based on previously reported gas-liquid chromatography analysis of adipose tissue samples⁴). This enabled modeling only two spectral fat components, from which ndb could be estimated.

A Siemens Verio 3T system and an 18-channel receive body coil was used to acquire 3D spoiled gradient multi-echo images with 12 echoes of a patient with an intramuscular lipoma enrolled in a study approved by the local ethics committee after giving informed consent. Imaging parameters were: TE1/ΔTE/TR=1.8/1.55/21 msec; FA=20°; matrix 96×96×48; voxel 2.7×2.7×5 mm³; NSA=12. The total acquisition time was 19 min 21 sec.

Images were reconstructed by both FAC3 and FAC1 using in-house software written in Python and C++. Reconstruction times were ∼30 sec/slice using a standard laptop computer.

ROI measurements (∅=19mm) were made at three locations within the subcutaneous adipose tissue and two locations within the lipoma, as indicated in Fig. 1a. The ndb measurements by the FAC3 method and the FAC1 method are given in Table 1. The FAC3 measurements vary significantly between the adipose tissue sites. This is also evident in Fig. 1c, where some large-scale variation from left to right is seen in the FAC3 ndb map. In comparison, the FAC1 ndb map (Fig. 1d) has a more uniform appearance and less noise. This is reflected in Table 1, where the FAC1 measurements have higher precision and are consistent between the adipose tissue sites, all in agreement with literature values. Taken together, measurements from the FAC1 ndb map seem more reliable. Interestingly, using FAC1 the measured ndb appears reduced in one of the lipoma compartments (ROI 4). Using FAC3, this difference cannot be appreciated with the same level of confidence.

The possibility of FAC mapping opens up some interesting medical applications, such as differentiation of fatty tumors. There is also an association between prostate cancer aggressiveness and FAC in the periprostatic adipose tissue⁵. These applications require high measurement precision since quite small differences can be expected. Another application is the study of non-alcoholic steatohepatitis⁶, which also requires high precision due to low fat content in the liver.

Reducing the degrees of freedom in the triglyceride model is a straightforward way to increase measurement precision. CL is not expected to vary much in humans, which motivates a reduction of this parameter. Reduction of nmidb significantly improves measurement precision since the resonances associated with nmidb have relatively small amplitudes. Parameter reduction also results in smaller demands on SNR. This allows more flexibility in imaging protocols, such as reducing the NSA, improving the spatial resolution, using fewer echoes, etc.

FAC parameter reduction in the spectrum model was done previously by Leporq et al.⁶ (reduction of CL) and by Bydder et al.⁷ (reduction of CL and nmidb). An advantage of our approach is that it is a simple extension of well-established chemical shift based water/fat separation methods, meaning that efficient available methods for the challenging determination of the B₀-map can be used, offering robustness to B₀ inhomogeneity and practical reconstruction times.

It is straightforward to reduce the number of free FAC parameters in order to increase measurement precision or allow more flexible imaging protocols. In particular, mapping only ndb appears less vulnerable to artifacts and offers greater confidence when comparing ndb between anatomical sites.