Parallel moduli space sampling: Robust and fast surgery planning for image guided steerable needles

Steerable needles can improve many medical procedures through their ability to reach targets behind critical or impenetrable anatomical structures, such as blood vessels and bones. However, path planning in real-time is challenging due to complicated kinodynamic constraints and tissue deformation. This paper introduces a parallel path planning algorithm that finds paths that avoid complex obstacles in real-time. This algorithm is based on moduli space planning algorithm. It transforms non-linear path planning problems into linear formulations, which can be efficiently processed in parallel, resulting in low computation time and high precision in spite of obstacle structure and tissue deformation. Our method works directly with raw image data in the voxel format, which greatly simplifies image processing and boosts system performance as a whole. The efficacy of the method has been demonstrated on simulated anatomy phantoms. For a simulated liver surgery planning task, 70 safe and accurate paths were generated after sampling more than 10,000 paths in 0.01 seconds. This allows the surgical path to be updated in real-time at 100Hz.