GPU-based Medical Visualization for Large Datasets

Hua Zou,Fu Lin,Jie Han,Wen Zhang
DOI: https://doi.org/10.1166/jmihi.2015.1560
2015-01-01
Journal of Medical Imaging and Health Informatics
Abstract:With improving medical equipment quality, the scale of acquired medical volume data continues to grow explosively. When the size of medical dataset exceeds the capacity of texture memory, GPU-based volume rendering algorithm is no longer able to process data directly. To solve this problem, a dynamic block rendering method is proposed to implement GPU-based real-time visualization of large-scale medical volume dataset. First, volume data is divided into equal-size sub-blocks. Second, according to the distance from the view plane, the sub-blocks are sorted in order and loaded into texture memory in turn dynamically. Then, the sub-blocks are rendered, and the rendering speed is accelerated using dataset layout and space skipping. Last, the rendering image of the whole volume dataset is produced by mixing rendering results of all sub-blocks. The experiment demonstrates that the proposed method can realize real-time interaction when the size of volume data is larger than the capacity of texture memory.
What problem does this paper attempt to address?