Quantitative measurement of temperature by proton resonance frequency shift at low field: a general method to correct non-linear spatial and temporal phase deformations

MRI thermometry methods are usually based on the temperature dependence of the proton resonance frequency. Unfortunately, these methods are very sensitive to the phase drift induced by the instability of the scanner which prevents any temperature mapping over long periods of time. A general method based on 3D spatial modelling of the phase drift as a function of time is presented. The MRI temperature measurements were validated on gel samples with uniform and constant temperature and with a linear temperature gradient. In the case of uniform temperature conditions, correction of the phase drift proved to be essential when long periods of acquisition were required, as bias could reach values of up to 200 degrees C in its absence. The temperature uncertainty measured by MRI was 1.2 degrees C in average over 290 min. This accuracy is coherent with the requirements for food applications especially when thermocouples are useless.