3D Reconstruction for 2D Spinal Ultrasound Imaging: An Overview

摘要

Common imaging techniques of the spine include radiography, CT and MRI. Spinal ultrasound has not yet gained popular acceptance in clinical environment. However, it has great potential for clinical imaging of the spine. Acquisition of two dimensional images is more common than three dimensional acquisition of images. In spinal imaging, besides cross sectional images, longitudinal images would also be of interest. Reconstruction of the two dimensional images into a three dimensional image is required for the longitudinal visualization of the spine. The available ultrasound reconstruction techniques have various advantages and disadvantages. An overview of these techniques is presented in this paper. They include VNN (Voxel Nearest Neighbour), Volume intersection techniques, Distance Weighted reconstruction; Statistics based reconstruction, PNN (Pixel Nearest Neighbour), DFI (Direct Frame Interpolation) and RBF (Radial Basis Function) based reconstruction. Both, forward and reverse approaches to the reconstruction problem are included. Most methods use data removal and prediction for validation. This review focuses on the concept used by each method, providing an overview of the various approaches to the problem of three dimensional reconstruction from ultrasound images. Further, the current scenario of three dimensional segmentation is presented and lastly, clinical applications of spinal ultrasound have also been described.