Catheter-based radiofrequency ablation (RFA) has become a reference curative therapy for the treatment of cardiac arrhythmias. Magnetic Resonance (MR) Thermometry based on the water proton resonance frequency shift (PRFS) may provide real-time visualization of temperature distribution in the myocardium during RFA¹ for improved safety and online assessment of the therapy outcome. However, to allow accurate temperature mapping for any slice orientation, real-time slice tracking must be employed. Moreover, rapid and robust motion correction algorithms are mandatory to compensate susceptibility related phase (and therefore temperature) variations. In this study, we propose to combine an echo-navigated EPI sequence with slice tracking to allow arbitrary positioning of the images ^2,3 and optical flow algorithms to correct residual in-plane motion and susceptibility artefacts in the heart. This approach was evaluated on the left ventricle (LV) in five healthy volunteers without heating and in vivo in a large animal model during RFA, with a spatial resolution of 1.6x1.6x3 mm³ and 3 to 5 temperature slices acquired at each cardiac cycle.

MR protocol: MR Temperature imaging was performed on a 1.5T MR system (Avanto, Siemens Healthcare, Germany) using two 16-channel cardiac coils. A fat-saturated, single-shot Echo Planar Imaging (EPI) sequence was combined with GRAPPA (acceleration factor of 2) to achieve a 1.6x1.6x3 mm³ spatial resolution, TE/TR/FA=16-18ms/100ms/60°, bandwidth = 1576 Hz/px, FOV = 180x180x3 mm³. Up to five slices (slice gap = 0.3 mm) were acquired at each heart beat during 250 consecutive cardiac cycles with ECG triggering. A crossed-pair navigator was located on the diaphragm to adjust the slice position in real-time. Saturation slabs were positioned along the FOV in the phase encoding direction to avoid aliasing and two additional saturation bands were set parallel to the image slices to reduce signal of blood (Fig.1-A).

MR Thermometry: Residual in-plane motion were compensated online using a Principal Component Analysis (PCA) based optical flow algorithm⁴. The overall magnetic field variations with respiration were approximated as the sum of linear phase changes of each motion displacement on a pixel-by-pixel basis giving a parameterized magnetic field model⁵. During the interventional procedure, the current flow field was used to reconstruct magnetic field distribution from the parameterized model. Spatio-temporal phase drift was corrected and a temporal low-pass Butterworth filter was added at the end of the thermometry pipeline. Image reconstruction and all correction algorithms were implemented in C++/GPU and integrated into the Gadgetron reconstruction framework⁶, with a total processing time below 120 ms/image.

Volunteer study: MR-guided thermometry was evaluated in short axis orientation on volunteers (N=5) under free-breathing conditions. Pixel-wise temporal standard deviation of temperature σT was calculated to assess the precision of the thermometry.

Radiofrequency ablation on a sheep: Two MR-compatible catheters were inserted into the sheep heart (Fig.2-A) under fluoroscopic guidance (Toshiba InfiniX, Toshiba Medical, Japan). The first catheter (MRI Intervention, U.S.A.) was located into the left ventricle (LV) and connected to a RF generator outside the Faraday cage. The second catheter (BiosenseWebster, Israel) was positioned into the right ventricle (RV) for heart pacing (140 bpm). RFA was run for 85s at 40 W simultaneously to MR thermometry. The temperature evolution was visualized in real-time using Thermoguide software (Image Guided Therapy, France).

On each volunteer, σ_T remained below 2°C over the whole LV (Fig.1-A), except for a limited number of pixels located in area where residual phase wrapping were not properly corrected (Fig.1-C). On the animal, σT was found below 1°C due to a lower and more regular respiratory amplitude (1.25 cm, see Fig.2-B, mechanically controlled ventilation) and after excluding 2% of the pixels in the LV corresponding to the uncorrected phase jump. During the ablation, a temperature increase of approximately 30°C was observed in real-time in the vicinity of the catheter tip (Fig.2-C,D).Real-time PRFS-based cardiac MR Thermometry using an echo-navigated GRE-EPI sequence can achieve average temperature stability of 2°C or better in the myocardium, combining cardiac triggering, parallel imaging and PCA-based motion and susceptibility artefacts correction algorithms. The spatial (1.6x1.6x3 mm³) and temporal resolution allow direct visualization of temperature evolution in the myocardium during RFA at each heart beat, with full flexibility on image orientation provided by online, navigator-based, slice tracking.