A fast referenceless PRFS-based MR thermometry by phase finite difference

Abstract

Proton resonance frequency shift-based MR thermometry is a promising temperature monitoring approach for thermotherapy but its accuracy is vulnerable to inter-scan motion. Model-based referenceless thermometry has been proposed to address this problem but phase unwrapping is usually needed before the model fitting process. In this paper, a referenceless MR thermometry method using phase finite difference that avoids the time consuming phase unwrapping procedure is proposed. Unlike the previously proposed phase gradient technique, the use of finite difference in the new method reduces the fitting error resulting from the ringing artifacts associated with phase discontinuity in the calculation of the phase gradient image. The new method takes into account the values at the perimeter of the region of interest because of their direct relevance to the extrapolated baseline phase of the region of interest (where temperature increase takes place). In simulation study, in vivo and ex vivo experiments, the new method has a root-mean-square temperature error of 0.35 °C, 1.02 °C and 1.73 °C compared to 0.83 °C, 2.81 °C, and 3.76 °C from the phase gradient method, respectively. The method also demonstrated a slightly higher, albeit small, temperature accuracy than the original referenceless MR thermometry method. The proposed method is computationally efficient (~0.1 s per image), making it very suitable for the real time temperature monitoring.

Quantitative MR thermometry based on phase-drift correction PRF shift method at 0.35 T

“…Temperature uncertainty for ex vivo porcine liver experiments in this study (1.49 °C) maintained an accuracy similar to the results reported by Harth et al on a 1.5 T MRI scanner (1.50 °C) [ 41 ]. Human brain studies showed that the temperature error obtained by the phase-drift correction PRFS thermometry was larger than that calculated by the phase gradient method, and smaller than that obtained by the phase finite difference model from a previous study [ 40 ]. The distinction of results could be acceptable and reasonable due to the obvious difference in field strengths.…”

“…Images were scanned by GRE sequence in the transverse plane using the following scanning parameters: TR/TE = 50/30 ms, FA = 30°, Matrix size = 128 × 128, FOV = 256 mm, number of slices = 3, ST/SG = 8/0 mm, acquisition time = 6.4 s/slice. In order to compare with the results obtained by Zou et al at 3.0 T MR scanner [ 40 ], the width of the selected inner and outer ROI (see as Fig. 3 b) were both 10 pixels (2 cm), which was the same as the above published literature.…”

MRI for Noninvasive Thermometry

“…Further improvements are offered by reference-less methods [126,153], whereby a background phase is derived from the n th image during ablation. This is achieved by fitting a polynomial to the surrounding unheated region and extrapolating to the heated region, thus avoiding subtraction with a separate image [21,154]. This approach is only successful if sufficient SNR is available, and if water-fat interfaces which can lead to inaccurate polynomial fitting are absent.…”

Magnetic resonance-based thermometry during laser ablation on ex-vivo swine pancreas and liver

“…The main drawback of PRF, especially in in vivo applications [14], is its high sensitivity to motion. Although recently many authors are investigating several promising strategies to reduce artifact due to patient movements [17,18], the effective spread of PRF is still limited in clinical applications. For the above mentioned reasons, our experiments employed the thermal dependence of T 1 of tissue for the estimation of temperature distribution within pancreas and liver during LA.…”