Dual-step iterative temperature estimation method for accurate and precise fat-referenced PRFS temperature imaging.

Abstract

The aim of this study was to propose dual-step iterative temperature estimation (DITE) of a fat-referenced proton resonance frequency shift (PRFS) method to improve both the accuracy and precision of temperature estimations in fat-containing tissues. A fat-water signal model with multiple fat peaks was used to simultaneously estimate the temperature, fat/water intensity and , and field offset. In DITE, model fitting was implemented with alternating 2-step minimizations. The estimated temperature map was smoothed between the 2-step minimizations, which is considered to be the most important step for improving the temperature precision. The performance of DITE was evaluated with a Monte Carlo simulation, fat/water phantoms, and ex vivo brown adipose tissue experiments and then compared with the performance of previous fat-referenced proton resonance frequency shift methods. In fat/water phantom experiment with a smooth temperature profile, the temperatures estimated by DITE are consistent with the thermometer results and present a better accuracy and precision than those of previous fat-referenced proton resonance frequency shift methods. In the brown adipose tissue heating experiment, the average mean error, SD, and RMS error were -0.08ºC, 0.46ºC, and 0.56ºC, respectively, over all of the measurements within the region of interest. Our proposed DITE method improves both the accuracy and precision of temperature measurements in tissues with fat fractions between 20% and 80% under smooth distribution of the temperature profile and represents a simple fat-referenced thermometry method.