Investigation On Surgical Needle Localisation Algorithms Based On Modified Parallel Integral Projection Transform

Image-guided surgical procedures require the precise knowledge of the position andorientation of the surgical needle with respect to the target at all times during the procedure.This task is commonly performed by a needle navigation system that determines thepositions and orientation of surgical needles in ultra-sound images in real-time. Hence, thenavigation system is an essential component of any image-guided intervention platform.However, a basic problem of the clinical application of such systems is to find a method thataccurately and reliably localizes such tools in2D/3D ultra-sound images real time. Aultra-sound based navigation system combined with a3D visualization system couldenhance the surgeon’s3D view of the position and orientation of the surgical needle. Suchinteractive visual feedback may provide the exact position of surgical needles relative tocritical anatomical structures, such as vessels, nerves and therefore, improve the outcome ofsurgery. The impetus for the work is derived from an interest in real time localisation of longstraight needles in2D/3D ultra-sound images during surgical interventions.This thesis introduces four state of art surgical needle localisation algorithms based on themodified parallel integral projection transform (MPIP). We particularly focus on optimizingthe modified parallel integral projection method by using four state of art optimizationmethods. All the four algorithms namely: the Multi-Resolution MPIP method (MR-MPIP);the Genetic Algorithm MPIP method (GA-MPIP); the Generalized Pattern Search method(GPS), have been successfully and sufficiently tested for accuracy and speed. The resultsindicate that all the four algorithms could be used for real-time application in IGSinterventions.