Researchers in interventional MRI and physicians with interest in temperature imaging and MR guided cardiovascular catheterizationAlthough catheter-based radiofrequency (RF) ablation is increasingly used to treat cardiac arrhythmia, insufficient temperature increase may lead to incomplete treatment whereas excessive local energy deposition may result in severe adverse effects such as myocardial tissue disruption (e.g. esophageal fistulas or tamponade due to steam pop). Since MRI can monitor in real-time the temperature distribution in the cardiac muscle1, automatic regulation of RF energy deposition from temperature images may improve safety and efficiency of the therapeutic procedure. In this study, we demonstrate the potential of such an automatic regulation using well established proportional, integral controller algorithms, adapted to monopolar and bi-polar RF ablation configurations. Experiments were performed ex-vivo on a sample of beef muscle.MR thermometry: temperature images were acquired at 1.5 T (Avanto, Siemens Healthcare) using a multi-slice, single shot echo planar imaging (TE=20ms, TR=85ms, FA=60°, Bandwidth=1568 Hz/Pixel, acceleration factor = 2, Partial Fourier=6/8) sequence with 110x110 voxels, leading to a 1.6x1.6x3mm3 spatial resolution. RF-ablation device: Two cylindrical (diameter 1mm, length 5 mm) home-made MR-compatible RF electrodes were inserted into a piece of beef muscle and connected to a programmable RF generator (Image Guided Therapy, Pessac France) located outside the Faraday cage. Image reconstruction and temperature calculation was performed using the Gadgetron reconstruction framework². Control algorithm: A dedicated regulation gadget was implemented to adjust energy deposition from online temperature images and a predefined temperature-time target profile. The control algorithm used 3x3 pixels ROI in the heated region to calculate the mean temperature value. The new output power was computed using a proportional-integral algorithm integrating the last available mean temperature data and next target temperature value to reach. Processing time for the regulation gadget was lower than 50 ms, ensuring real-time operation with the acquisition update rate of 5 temperature slices/s of the thermometry sequence. This regulation loop was tested for two configurations, namely bipolar RF ablation and monopolar ablation, the 3x3 pixel was positioned, respectively, in the middle of both and centered on one electrode.Figure 1 (top row) displays magnitude and temperature data for a bipolar (1 cm inter-electrode distance) ablation. The graph on the right shows the target temperature profile (blue curve), the mean (green curve) and the maximal temperature increases (red curve) over the 3x3 ROI. Residual oscillation of 2°C can be observed on the green curve for a targeted temperature increase of 20°C, with a maximal experimental temperature increase of 28°C. Figure 1 (bottom row) displays similar data for the experiment performed in the monopolar configuration. A latency of the green curve as compared to the target curve was observed, due to maximal output power set on the generator by the regulation algorithm. The maximal temperature increase was also higher (33°C) than for the bipolar configuration. This can be explained by stronger temperature gradients in the heated area as compared to the bipolar configuration, leading to a higher difference between the mean and maximal temperature values within the ROI. Since the regulation was performed on the average value, higher temperature increase were expected. However, for both monopolar and bipolar configurations, assuming an initial body temperature of 37°C for a clinical application, the maximal temperature increase would have remained below 70°C, far below the boiling temperature.The implemented regulation gadget was fast enough to ensure real-time update of RF power at a realistic update rate of 5 images/cardiac cycle. The proposed algorithm is simple to implement and allows controlling the mean temperature increase in a restricted region whose dimensions correspond to lesion sizes observed in common RF ablation procedures in the heart (~8 mm in diameter). Temperature control on the mean value over a ROI increases the safety of the procedure as compared to single pixel approach³, particularly in presence of a mobile organ such as the heart where the catheter position may vary during the RF energy deposition for successive acquisitions of MR temperature images. More sophisticated algorithms may be implemented in future work to avoid exceeding a predefined maximal value, with the aim of reducing the risks of creating tissue disruption. Also, the regulation quality on the mean temperature value may be improved by adding a derivative term to the control algorithm, as suggested in ⁴ to cancel temperature oscillations observed in the present work. Such temperature regulation are expected to improve safety and efficiency of the MR-guided catheter-based RF ablation, after in vivo validation on large animal models.