Peripheral Nerves Segmentation in Ultrasound Images Using Non-linear Wavelets and Gaussian Processes

Abstract

Regional anesthesia is carried out using a technique called peripheral nerve blocking (PNB), which involves the administration of an anesthetic nearby the nerve. Ultrasound images have been widely used for PNB procedure due to their low cost and because they are non-invasive. However, the segmentation of nerve structures in ultrasound images is a challenging task for the specialists since the images are affected by echo perturbations and speckle noise. Automatic or semi-automatic segmentation systems can be developed in order to aid the specialist for locating nerves structures accurately. In this paper we propose a methodology for the semi-automatic segmentation of nerve structures in ultrasound images. We use non-linear Wavelets transform in the feature extraction step and for the classification stage we use a Gaussian Processes classifier. Experimental results show that the implemented methodology can segment nerve structures accurately.