In this paper, we design a GPU-based algorithm for determining the minimum distance from the tip of CUSA scalpel to the closest blood vessel. The calculation time of GPU-based algorithm is to be O(1), which does not depend on shape complexity of blood vessels, i.e., their patch number. On the other hand, calculation time of CPU-based algorithm is to be O(n). Even if we use several kinds of hierarchical structures in positioning such as sorting in XYZ axes, calculation complexity of CPU-based algorithms is to be O(log(n)) at the best.
As contrasted with this, when each STL is always converted into rectangular parallelepiped group in Z-buffer of GPU, we should accept quantization errors in XY-plane and Z-axis. Furthermore, so as to design parallel processing in GPU, we should omit all exclusive controls. As a result, distance precision and scanning range calculated by the GPU-based algorithm become worse than those done by the CPU-based algorithm. However, the precision (0.5mm) and range (0〜50mm) are fully accepted by the doctor’s request.
View full abstract