GPU-ASSISTED RENDERING OF LARGE TREE-SHAPED DATA SETS

by Mangalvedkar, Pallavi Ramachandra

Due to their complexity, medical data sets can easily comprise of several gigabytes in size. For example, geometric representations of a coronary arterial tree spanning vessels from the large proximal coronary arteries down to the capillary level consist of 6 gigabytes or more of geometry data, depending on the accuracy of the geometric representation. Visualization of such large data sets using the CPU alone can be inefficient and time consuming since they require special out-of-core techniques due to the size of the data set. The usage of the GPU can help render such data sets faster compared to CPU-based implementations by computing geometric information on-the-fly, thus eliminating the necessity of out-of-core methods. Here, we present a GPU-based algorithm for rendering large-scale tree-shaped data sets which takes advantage of the programmability of modern graphics hardware. The proposed approach makes use of a fragment shader to compute geometric information on-the-fly, thereby allowing faster visualization. Also to make use of the CPU’s programmability along with the GPU, we further extend the algorithm such that the computation of the geometry is executed in parallel by the CPU and the GPU, which makes it more efficient. We illustrate this with a vascular tree data set which resembles the entire arterial coronary vasculature.