Non-iterative guidewire reconstruction from multiple projective views

We present a method to reconstruct a 3D curve, a guidewire or a needle, using an arbitrary number of multiple projective views. In the first step, an unstructured set of 3D points representing the curve is found through a ray-ray intersection search. In the second step, a partial ordering is imposed on the points to recover a parametric curve representation. The proposed method does not require a parametrization of the 2D projections or a solution to the correspondence problem between 2D curve projections. The proposed method yields a high precision reconstruction of the curve with the average error of the reconstruction of about 0.95±0.47mm when compared to the cone beam reconstruction. Applications of the reconstruction method include both guidewire guidance during endovascular procedures and needle positioning during neurosurgery.