Real-time analysis of ultrasound images using GPU

Analysing the increasing number of medical images stemming from various imaging modalities is a time-consuming task for the clinicians. Presently, medical images are mainly hand-segmented and manually analysed. Automatic tools are needed to make medical image analysis easier, and to create better presentations of information needed prior to and during surgery. In this article, we focus on real-time analysis of ultrasound images, and on how inexpensive graphics hardware can be utilised to accelerate current image processing techniques. Algorithms evaluated on the GPU were mathematical morphology, gradient vector flow, and the snake model. The results are promising and show that a large number of iterations can be run in 1/24 of a second.