Device-specific spatio-temporal drifts in proton resonance frequency shift thermometry at 3 T

Magnetic resonance imaging is the modality of choice for noninvasive temperature mapping to monitor and also to control hyperthermia. Temperature rise is observed through the proton resonance frequency shift (PRFS) reflecting in the phase of gradient echo acquisitions. Temporal and spatial magnetic field drift, generated under intense gradient use affects the phase and induces strong temperature misinterpretations. This work investigates the impact of hardware and imaging sequence on long-term temperature monitoring accuracy.

The behavior of field drift was recorded over time using image-based field mapping. Measurements were performed on a custom agar phantom with different imaging protocols and MRI scanners.

The spatio-temporal evolution of drift varies with MRI device, vendor, and site and can exceed 10² °C. Strongest changes of center frequency and spatial inhomogeneity arise over up to 70 min of high gradient intensity. Saturation is not reached even after almost 2 h. Observed field distortions could be associated with the employed gradient groups. Drift was even observed under minimal gradient use. Drift patterns were reproducible using the same device and sequence configuration and could successfully be modeled by multidimensional polynomials.

Spatio-temporal drift characterization is important for PRFS thermometry. It can be used to define 'warm up' phases for MRI scanners, after which further drift effects are reduced. A priori knowledge on the spatio-temporal drift evolution could also be applied to correct for background fields instead of blind fitting. These findings may pave new pathways to increase temperature mapping accuracy for various thermal intervention techniques.