| The utilization of ultrasound in diagnostic medicine has increased since the 1950's. Ultrasound is a non-invasive two dimensional (2D) imaging technique that provides useful information about underlying soft tissue anatomical structures. Developed recently, three dimensional (3D) reconstruction algorithms convert a series of sequential 2D ultrasound images into 3D data sets. 3D reconstruction is one phase in a pipeline for 3D ultrasound volume visualization that typically takes upwards of 2 minutes to complete. The objective of this research focuses on high speed 3D ultrasound reconstruction by comparing reconstruction speeds on parallel and sequential processors. To decrease reconstruction times, optimizations such as; conversion from floating point to fixed point, multithreading, in-line coding, and mathematical analysis were performed. Optimizations achieved a 15%--50% performance increase. Although parallel processors are not a requirement for 3D reconstruction, they will be necessary to achieve the ultimate goal of real-time 3D ultrasound volume visualization. |
