A Preliminary Evaluation of a Flexible Needle Steering Algorithm Using Magnetic Resonance Images as Feedback Pedro Moreira, Gert van de Steeg, Ferdi van der Heijden, Jurgen J. Fütterer and Sarthak Misra Abstract— Needle-based procedures are commonly per- formed for cancer diagnosis and treatment. Imaging modalities are used to visualize the needle tip and the target during these needle insertion procedures. Among the available imaging techniques, magnetic resonance (MR) offers the best tissue contrast, where detection of an early stage cancer is possible. MR-guided needle insertions are currently performed with rigid needles, which have limited steerability. Flexible needles have been introduced to increase the steerability during the insertion. In this paper, we present a preliminary evaluation of a steering method for flexible bevel-tipped needles using MR as an imaging modality. The steering algorithm uses a needle deflection model to predict the tip motion and calculate the optimal rotation to reach the target. The best sequence of rotations are defined by an optimization algorithm based on the Nelder-Mead technique. The needle tip and the target are manually tracked through a graphical user interface. The needle is inserted by a device fabricated with MR-compatible material. The MR-guided flexible needle steering is evaluated by a series of insertions in two phantoms with real obstacles and targets. The average targeting error with flexible needles is 4.3mm, which is 28% lower than the values reported in the literature with rigid needles. The results indicate the feasibility of MR-guided flexible needle insertions. I. I NTRODUCTION Early detection and treatment are of major importance to reduce cancer mortality rate. Several diagnostic and treat- ment techniques require needle insertion procedures, such as biopsy and brachytherapy. The success of such techniques are closely dependent on the needle placement accuracy. During needle insertion procedures, medical imaging modalities are used by clinicians to identify the needle tip and the target locations. Ultrasound (US) imaging is the commonly used technique to guide the needle towards a target due to its simplicity and real-time imaging [1]. However, US images present a poor tissue contrast and an early stage cancer lesion might not be visible [2]. Computed tomography (CT) offers a better tissue contrast than US, but the lack of real-time imaging and the toxic radiation delivered to the patient are important shortcomings [3]. On the other hand, magnetic resonance (MR) imaging offers high resolution images of tissues, in which the detection of early stage cancers is possible, without toxic radiations. Nonetheless, MR-guided The authors are affiliated with MIRA - Institute for Biomedical Technol- ogy and Technical Medicine (Robotics and Mechatronics Group), University of Twente, The Netherlands. Jurgen J. Fütterer is also affiliated with the Department of Radiology, Radboud University Nijmegen Medical Centre, The Netherlands. This research is supported by funds from the Dutch Ministry of Economic Affairs and the Province of Overijssel, within the Pieken in de Delta (PIDON) Initiative, Project: MIRIAM (Minimally Invasive Robotics In An MRI environment). Insertion device Phantom MR image Target Needle Obstacles Bevel tip Fig. 1. A bevel-tipped flexible needle is steered towards a real target. The user inserts and axially rotates the needle into a phantom with obstacles using an insertion device. A magnetic resonance scanner is used as imaging modality to track the needle and the target. needle insertion faces many challenges due to the space constraints of the MR bore, concerns about MR compatibility and the difficulty of acquiring real-time images. Several robotic devices have been developed to assist MR- guided needle insertions. Fully automated MR-safe robotic systems have been developed to insert rigid and flexible needles [4], [5]. These systems are built with MR-compatible materials and the needle is driven by pneumatic or piezoelec- tric actuators. Semi-automated systems have been designed to robotically position the needle guide, allowing a manual needle insertion in the direction of the target [6], [7], [8]. Manual insertions are performed using thick and rigid nee- dles while assuming that they follow a straight path. Those needles have limited steerability and may induce target mo- tion by tissue deformation. In the last decade, the use of thin and flexible needles have been extensively studied in order to reduce patient discomfort, tissue deformation and also increase needle steerability [9], [10], [11]. However, those needles have not yet been used on needle-based procedures guided by MR images. Several models and algorithms have been proposed to steer bevel-tipped flexible needles. Webster et al. [9] and Misra et al. [12] presented a kinematic and mechanics- based models to steer a bevel-tipped flexible needle, re- spectively. Two-dimensional (2D) steering algorithms are 2014 5th IEEE RAS & EMBS International Conference on Biomedical Robotics and Biomechatronics (BioRob) August 12-15, 2014. São Paulo, Brazil 978-1-4799-3127-9/6/14/$31.00 ©2014 IEEE 314