MRI temperature map reconstruction directly from k-space with compensation for heating-induced geometric distortions

Proton resonance frequency (PRF) change is used to measure tissue heating, but also distorts the image and causes geometric distortions in temperature estimates in the same manner as other chemical shift distortions if left uncompensated. We propose an algorithm that produces PRF temperature maps free of these distortions by fitting a signal model directly to acquired k-space data that accounts for PRF-induced phase both up to and during the readout. We also introduce a faster method compatible with Cartesian data that corrects distortions from image-domain temperature maps. Gel heating experiments show the proposed CS compensation algorithms correct magnitude image artifacts and hotspot distortions. Without CS compensation, thermal dose values are overestimated in spiral data, and are spatially offset in 2DFT and EPI data. Compensating for heat-induced CS distortions improves the accuracy of temperature change and thermal dose measurements, and can have a significant positive impact on clinical and resea...