Eye Torsion Determination Using Optical Flow

摘要

Torsional image registration is of crucial importance for a wide variety of medical applications especially in the field of ophthalmology and refractive surgery. We would like to underline particularly the relevance for refractive surgery where, beside the translational, also the torsional motion of the eye has to be known exactly in order to guide the tissue ablating laser beam towards the intended location on the anterior corneal surface to achieve the desired shape. So-called eye trackers, nowadays mostly video based, are used to determine the eye position accurately. Those systems compare the current image with a reference image (e.g. taken at the beginning of the operation or from a diagnostic device) to derive the torsional component of the eye motion relative to that reference. A further array of application with growing significance are implantations of toric Intraocular lenses (IOL) after cataract surgery. In such an operation an artificial plastic lens is used to replace the crystalline lens. Due to the fact that toric IOLs are not rotationally symmetrical they have to be properly aligned with respect to the visual axis of the eye in order to achieve the optimum visual outcome. A video based system that during implantation keeps track of the torsional orientation of the eye can be used to assist the surgeon to place the toric lens into the right angular orientation. In this paper we propose a novel approach for finding the torsional displacement between two eye images - mostly this pair will consist of a reference image and the current camera image. Our technique makes use of optical flow methods, which are well known and widely used.