GPU-Based Volume Segmentation

Volume segmentation is an important part of any medical image analysis framework used for diagnoses, treatment planning and biomedical modelling and visualisation. Recent advances in modern graphics hardware have made it possible to perform general purpose computing on the GPU. In this paper we survey and analyse the current state-of-the-art of GPU-based volume segmentation algorithms. Limitations of GPU-based implementations due to the architecture of graphics hardware are discussed and guidelines for developing new GPU-based algorithms are given. We conclude with the GPU-based implementation of a hybrid volume segmentation method which makes use of the latest GPU features such as frame buffer objects. We present a framework for the interactive control and visualisation of the segmentation process and we demonstrate the speed advantage achieved by using the GPU.