Alignment of Freehand 2-D Ultrasound Images for 3-D Volumetric Reconstruction

Abstract

Ultrasound is an imaging modality used in many medical procedures including the diagnosis of diseases in Neonates. We aim to develop a Computer simulation using 3-D ultrasound to teach clinicians how to perform Neonatal head Ultrasound. To achieve this, we need to align a series of 2-D ultrasound data to reconstruct the 3-D volume. The source data was acquired by freehand movement of the ultrasound probe which results in misalignment and incorrect spacing/angle of the 3-D data. Previous techniques used to solve this problem rely on speckles in the ultrasound images. We devised an algorithm to automatically correct the misalignment using 2-D registration and estimate the elevational spacing/angle by optimizing a cost function against a perpendicular reference Image. After running the algorithm on the dataset, we observed significant improvement in the alignment and 3-D structure. This algorithm will facilitate generating 3-D volume data needed to develop a high quality Simulator.