Parallel computation for rapid reconstruction of volumetric ultrasonic images

A massively parallel computer, the Princeton Engine (PE), is applied to the task of volumetric ultrasound image reconstruction. Results are compared with those obtained on a conventional, sequential computer (a Sun SparcStation 1). The PE reduces reconstruction times by a factor of approximately 500 over the Sun, making possible the creation of a flexible and efficient tool for experimenting with image-reconstruction parameters and algorithms. The image reconstruction of volumetric data is reviewed, and the approach used to map the reconstruction algorithm onto the parallel processor, the implementation problems resolved, and the acceleration achieved are discussed