Needle Detection in 3D Ultrasound Images Using Anisotropic Diffusion and Robust Fitting

Needle insertion is a minimally invasive medical procedure with a vast domain of applications. 3D localization of the needle is important in needle visual tracking by a physician and robotic automatic needle guidance. Ultrasound is a fast and non-invasive medical imaging modality that is suitable for intra-operative imaging. But unfortunately, ultrasound images suffer from speckle noise and other artifacts that degrade the image quality. This paper investigates detection of the needle position in 3D Ultrasound images. We combined the RANSAC robust fitting algorithm with a structure adopted denoising method called anisotropic diffusion. The results show more accuracy compared to the previous method and significant improvement in the processing time.