Influence of water and fat heterogeneity on fat-referenced MR thermometry.

PURPOSE: To investigate the effect of the aqueous and fatty tissue magnetic susceptibility distribution on absolute and relative temperature measurements as obtained directly from the water/fat (w/f) frequency difference. METHODS: Absolute thermometry was investigated using spherical phantoms filled with pork and margarine, which were scanned in three orthogonal orientations. To evaluate relative fat referencing, multigradient echo scans were acquired before and after heating pork tissue via high-intensity focused ultrasound (HIFU). Simulations were performed to estimate the errors that can be expected in human breast tissue. RESULTS: The sphere experiment showed susceptibility-related errors of 8.4 °C and 0.2 °C for pork and margarine, respectively. For relative fat referencing measurements, fat showed pronounced phase changes of opposite polarity to aqueous tissue. The apparent mean temperature for a numerical breast model assumed to be 37 °C was 47.2 ± 21.6 °C. Simulations of relative fat referencing for a HIFU sonication (ΔT = 29.7 °C) yielded a maximum temperature error of 6.6 °C compared with 2.5 °C without fat referencing. CONCLUSION: Variations in the observed frequency difference between water and fat are largely due to variations in the w/f spatial distribution. This effect may lead to considerable errors in absolute MR thermometry. Additionally, fat referencing may exacerbate rather than correct for proton resonance frequency shift-temperature measurement errors.