Abstract: Implementation of a Real-time 3D-thermometry Pipeline in Gadgetron for Easy Clinical Integration.

MR thermometry as measure for guiding thermal ablation of tumor tissue makes realtime imaging and arbitrary slice positioning possible while assuring precise knowledge of the size of the ablation zone. However, difficult clinical integration currently hamper the success of clinical thermometry. Purpose of this work was to implement a real-time 3D MR thermometry pipeline for easy clinical integration. The pipeline starts with fast data acquisition by a multigradient echo 3D stack-of-stars sequence. The acquired data is then sent to an external server where images temperature maps as well as necrosis maps are calculated from it via a modular Gadgetron pipeline running in a Docker container. Using Docker guarantees multi-platformportability and reproducibility of our pipeline whereas the ISMRMRD format makes it vendor independent. Parallel imaging, compressed sensing (PICS) as well as GPU accelerated reconstruction are implemented. Apart from that, the pipeline contains correction algorithms for noise, gradient delay, phase drift and phase errors. To test the pipeline, microwave ablations on nine static bioprotein phantoms were performed on a 1.5 T scanner. Fiber optical temperature sensors were used for gaining reference temperatures. Retrospectively, the calculated necrosis zones were compared to a ground truth. The pipeline yielded temperature accuracies of 1.15 °C ± 0.69 °C compared to the temperature sensors. The comparison of the ablation zone with the ground truth yielded dice scores of 84.02 % ± 2.36 %. For a 3D volume with 24 slices and 8 echoes the processing time was about 26 s. The experiments show that the thermometry pipeline is already close to real-time 3D thermometry. Further research will focus on optimizing it for moving phantoms as, in the future, it is planned to be used for clinical applications in moving organs such as the liver. The modular and opensource character of the pipeline makes future improvements and extensions easy to integrate [1].