Accuracy of fat fraction estimation using Dixon: experimental phantom study

Abstract

Objective. Quantitative assessment of Dixon two-point and three-point technologies operation using phantom modeling in the range from 0 to 70%.Materials and methods. To simulate substances with different concentrations of the fat phase we chose direct oil-in-water emulsions. Tubes with ready-made emulsions were placed in a phantom. Emulsions based on vegetable oils were presented in the range from 0–70%. The phantom was scanned on an Optima MR450w MRI tomograph (GE, USA) in two Dixon modes: the accelerated two-point method “Lava-Flex” and the three-point method “IDEAL IQ”. A scan was performed on a GEM Flex LG Full RF coil. We calculated fat fraction (FF) using two formulas.Results. There is a linear relationship of the determined values when calculating the fat concentration in “IDEAL IQ” mode and using the formula based on Water and Fat. The accuracy of body fat percentage measurement in “IDEAL IQ” mode is higher than in “Lava-Flex” mode. According to the MR-sequence “Lava-Flex” draws attention to the overestimation of the measured values of the concentration of fat in relation to the specified values by an average of 57.6% over the entire range, with an average absolute difference of 17.2%.Conclusion. Using the “IDEAL IQ” sequence, the results of the quantitative determination of fractions by formulas were demonstrated, which are more consistent with the specified values in the phantom. In order to correctly quantify the fat fraction, it is preferable to calculate from the Water and Fat images using Equation 2. Calculations from the In-phase and Out-phase images provide ambiguous results. Phantom modeling with direct emulsions allowed us to detect the shift of the measured fat fraction.