Abstract
Robotic needle steering with deflection model and 3D US feedback
The precision during needle involved procedures is very critical for patient safety and operation success. Clinicians know for a long time that bevel needle tends to bend during insertion and sometimes use this property to guide the needle tip toward their target, which is called needle steering. This technique suffers from Human limitation such as precision or rapidity, that is the reason why robotic devices are very promising in this fields. However, needle steering needs to have prediction of needle deflection. Simple kinematic models have been developed but deals with imprecision in certain case, as instance when the needle is relatively thick and exert large force on surrounding tissue. We therefore choose to consider a mechanical-based model to predict needle deflection. Another issue is the feedback used during the needle steering. 3D Ultrasound seems to be the best imaging modality regarding its cost and level of risk for both patient and clinician but it suffers from its poor quality due to low signal to noise rate and acoustic artifacts. We develop a needle segmentation program inspired by previous works using Random Sample Consensus algorithm and Bezier curve.