Extracting surfaces from fuzzy 3D-ultrasound data

Abstract

Rendering 3D models from 3D-ultrasonic data is a complicated task due to the noisy, fuzzy nature of ultrasound imaging containing a lot of artifacts, speckle etc. In the method presented in this paper we first apply several filtering techniques (low-pass, mathematical morphology, multi-resolution analysis) to separate the areas of low coherency containing mostly noise and speckle from those of useful information. Our novel BLTP filtering can be applied at interactive times on-the-fly under user control & feed-back. Goal of this processing is to create a 'region-of-interest' (ROI) mask, whereas the data itself remains unaltered. Secondly, we examine several alternatives to the original Levoy contouring method. Finally we introduce an improved surface-extraction volume rendering procedure, applied on the original data within the ROI areas for visualizing high quality images within a few seconds on a normal workstation, or even on a PC, thus making the complete system suitable for routine clinical applications.