Robust Segmentation and Volumetric Registration in a Multi-view 3D Freehand Ultrasound Reconstruction System

In this paper, we present a new reconstruction method with volumetric registration and semi-automatic segmentation in a multi-view 3D freehand ultrasound imaging system. The new volumetric registration approach is performed on binarized walls using non-linear least squares. It can provide accurate multi-view reconstructions despite significant rigid target motion between different acoustic window acquisitions. A hybrid adaptive gradient vector flow (GVF) geometric active contour (GAC) model is used for image sequence segmentation. It allows for relatively simple initialization of the deformable model, while avoiding edge leaking at poor edges and small boundary gaps. The algorithms were validated on four ultrasound phantom data sets (eight sequences of a total of 336 images) and two echocardiography data sets (four sequences for a total of 75 images). Quantitative evaluation shows that automatic segmentation is comparable with manual segmentation. Using breath-holding and cardiac gating, volume estimates from multi- view reconstructions of the left-ventricle were found to be in better agreement with clinical estimates than volumes estimated from single view reconstructions.