Purpose
To perform multi‐echo water/fat separated proton resonance frequency (PRF)‐shift temperature mapping.
Methods
State‐of‐the‐art, iterative multi‐echo water/fat separation algorithms produce high‐quality water and fat images in the absence of heating but are not suitable for real‐time imaging due to their long compute times and potential errors in heated regions. Existing fat‐referenced PRF‐shift temperature reconstruction methods partially address these limitations but do not address motion or large time‐varying and spatially inhomogeneous B0 shifts. We describe a model‐based temperature reconstruction method that overcomes these limitations by fitting a library of separated water and fat images measured before heating directly to multi‐echo data measured during heating, while accounting for the PRF shift with temperature.
Results
Simulations in a mixed water/fat phantom with focal heating showed that the proposed algorithm reconstructed more accurate temperature maps in mixed tissues compared to a fat‐referenced thermometry method. In a porcine phantom experiment with focused ultrasound heating at 1.5 Tesla, temperature maps were accurate to within 1∘C of fiber optic probe temperature measurements and were calculated in 0.47 s per time point. Free‐breathing breast and liver imaging experiments demonstrated motion and off‐resonance compensation. The algorithm can also accurately reconstruct water/fat separated temperature maps from a single echo during heating.
Conclusions
The proposed model‐based water/fat separated algorithm produces accurate PRF‐shift temperature maps in mixed water and fat tissues in the presence of spatiotemporally varying off‐resonance and motion.