Proton resonance frequency-shift MR thermometry has been widely applied to monitor temperature changes during thermal ablation. However, in RF, microwave and laser ablations, the ablation probe may induce signal loss in tissue surrounding it due to susceptibility-induced B₀ field gradients, which prevents accurate temperature monitoring where the thermal dose is highest. To address this problem, we present a dual echo sequence with z-shimming¹ that recovers signal around metallic ablation probes, and an associated penalized likelihood approach to estimate a single temperature map from both echoes.

Sequence Figure 1 shows the dual echo z-shimming sequence. To recover signal around the probe, the first echo is partially refocused in the slice dimension by reducing the slice rephasing lobe immediately preceeding it to p% of its full-refocusing area. Signals away from the probe are fully refocused by a second slice refocusing gradient placed between the two echoes with q% = 100% - p% of the full refocusing gradient area, so the second echo images the rest of the tissue away from the probe, and has a typical PRF thermometry echo time.

Temperature estimation The multi-echo hybrid algorithm ^2,3 was adapted to jointly estimate temperature changes from the two echoes. The algorithm works by jointly fitting the following model to the echo images:$${\widetilde{y}_j}=(\sum\limits^{N_{b}}_{l=1}b_{j,l}w_{l})e^{i(\left\{Ac\right\}_{j}+f_{j})T_{E}}+\epsilon_{j},$$where $$$j$$$ indexes pixels, $$$N_{b}$$$ is the number of baseline images, $$$w_{l}$$$ is the weighting of baseline image $$$b_{l}$$$. $$$A$$$ is a polynomial matrix with coefficient vector $$$c$$$ to model background field drifts, $$$f$$$ is the temperature-induced frequency shift, $$$T_{E}$$$ is echo time, and $$$\epsilon$$$ is complex Gaussian noise. The model is fit using an iterative gradient-based algorithm, while regularizing for a sparse temperature map since heating is focal.

Phantom experiments The sequence was implemented on a 3T scanner (Philips Achieva, Philips Healthcare). Two sets of axial images of an agar phantom with a nitinol wire (diameter=3mm) inserted in it parallel to the B₀ field were acquired using an 8-channel head coil (z-shim p=50%, TE1=10ms, TE2=16ms, TR=30ms, FOV=150mm×150mm, slice thickness=3mm). The p value used in these experiments was equivalent to 0.81 cycles of phase through the slice. The first data set comprised 100 dynamic images without heating to compare temperature standard deviation (STD) maps of dual-echo z-shim and single-echo temperature maps. In the second data set, heating was performed with a microwave generator to validate recovery of temperature mapping precision near the probe. An additional spin echo image of the same slice was also acquired (TE=10ms, TR=600ms) to illustrate the size of the wire and the extent of the signal loss it created.

Fig. 3 shows the magnitude of the first and second echo images, and complex combined images from the dual echo z-shim sequence. The diameter of the signal loss region in the second echo GRE image was approximately 12.8 mm. The diameter of the signal loss region combined dual echo image was more similar to the spin echo image, approximately 3.8 mm. Compared to single (second) echo temperature maps, the temperature maps estimated from both echoes using the dual-echo hybrid algorithm showed significant temperature closer to the wire. Fig.4 shows that the dual echo z-shim temperature maps have lower STD than single-echo maps around the metallic wire. Temperature curves of a pixel near the wire also reflect a large reduction in uncertainty over time with dual-echo z-shimmed thermometry.A dual echo sequence with z-shimming and an associated temperature reconstruction algorithm were proposed, which can recover signal for precise temperature mapping near metallic wires. The z-shim value can be selected manually before heating or automatically chosen using a self-adaptive method as in [4].