Percutaneous kidney cryoablation has emerged as an option for renal tumor treatment that avoids surgical morbidity and complications [1]. MRI is an ideal tool to monitor ablation margins in cryoablation, because it can visualize both the tumor and the ice ball with greater accuracy than CT and ultrasound [2,3]. However, the ice ball seen as a signal void on ordinary MRI sequences (with TE>1ms) does not represent the ablation volume accurately; studies have shown that the critical temperature to induce malignant cell necrosis is -20°C [4], while MRI can only delineate the boundary of frozen tissue. To address this issue, MRI-based temperature mapping using ultrashort TE (UTE) imaging was proposed [5-7]. In this study, we tested the feasibility of R2*-based temperature mapping using a 3D UTE Point-wise Encoding Time Reduction with radial Acquisition (PETRA) sequence [8] to determine the volume of tissue ablated or “kill zone” created by cryoablation in the kidneys, using a post-ablation contrast-enhanced MR (CE-MRI) as the “gold standard”.

Ex Vivo Swine Kidney Experiment for Calibration. The relationship between the R2* change and temperature was determined using two ex vivo swine kidneys. Two 17-gauge stainless-steel MRI-compatible cryoablation probes (Galil Medical) were inserted into each kidney with two thermocouples (Omega) embedded in carbon-fiber needles for reference temperature measurement. Cryoablation was performed over 30 minutes in a 3T MRI Scanner (Verio 3T, Siemens) using an MRI-compatible cryoablation system (Galil Medical), while the temperatures at the thermocouple tips were recorded. MR images were continuously acquired using a dual-echo PETRA sequence (TR/TE1/TE2=4.5/0.07/2.0ms; matrix=160×160×160; pixel size=2mm³; flip angle=8°; FOV=350mm³; 11000 spokes, TA=1min/vol). R2* maps were computed by fitting the first and second echoes to a decaying exponential. In order to cancel inhomogeneous R2* close to the cryoablation probes, a ΔR2* map was calculated by subtracting a baseline (25 °C) R2* map from each R2* maps. The mean ΔR2* within a region of interest of ~300 mm³ at each thermocouple tip was correlated with the reference temperature. A linear least-square fit between ΔR2* and temperature was performed.

Data Acquisition in MRI-guided Renal Cryoablation. The Institutional Review Board approved this study. MR images were acquired in six patients during MRI-guided renal focal tumor cryoablation with two freeze-thaw cycles. Patients were treated under monitored anesthesia care (MAC) (n=5) or general anesthesia (GA) (n=1). PETRA images were acquired at baseline and between regular T2-weighted scans with a 2D half-Fourier acquisition single-shot turbo spin echo sequence (TR/TE=1000/200ms; matrix=320×190; FOV=289×340mm²; slice=3.7mm) during a 15-minute freezing cycle. During the scan, the patient was instructed to breath-hold (MAC), or was under controlled apnea (GA).

Data Analysis. After computing ΔR2* maps, a temperature map within the ice ball at the end of the second freezing cycle (15-min) was estimated using the calibrated parameters. The ablation volume was estimated using: (a) the -20°C boundary on the temperature map (R2*Temp volume); (b) the signal void on the intraoperative T2-weighted image (T2w volume); and (c) the hypo-intense, non-perfused area on a CE-MRI (Post CE volume) acquired 24 hours after the procedure.

The ΔR2*-to-temperature calibration provided the equation ΔR2*=0.00633×T+0.0628, where T is the temperature in °C. Fig. 1 shows representative temperature maps, T2-weighted MRI, and follow-up CE-MRI at corresponding slices. The estimated ablation volumes using (a)-(c) are shown in Table 1. Overall, the temperature map showed smaller ablation volumes relative to the T2-weighted MRI (p<0.05). The temperature map underestimated the ablation volume in 5 cases, whereas T2-weighted MRI overestimated in 4 cases, but the differences were not significant (p=0.16 and 0.06). The R2*-based temperature maps appear to provide a reliable lower limit on the extent of the kill zone, but there are inconsistent results when comparing R2*Temp and T2w volumes. It is likely that the kill zone is larger than the thermal estimate, since the mechanisms of cell death involve various factors such as microvasculature thrombosis, the number of freeze-thaw cycles, and the rate of thawing. R2*-based temperature maps contained artifact due to large streaking artifacts (case 3) and mis-registration due to breathing or gross motion (cases 1 and 6; both under MAC), which might have contributed to under-/overestimation of the ablation volume. The streaking artifact resulted from the use of fewer radial spokes (11000 vs. 25000 for minimal-streaking) as required to maintain reasonable breath-hold durations. Accurate image registration may improve these results. The current analysis was limited to a single time point, and ignored time-increment freeze-thaw cycles and voxels that became the critical temperature before 15-min. We are currently analyzing the other time points in an effort to tighten the lower bound.