Massively parallel computing for Biomedical Applications

April 10 2013, 3:00 PM

302- 309



Many scientific researches involve time-consuming numerical computations over extremely large, high-resolution datasets. One example is connectomics, a growing research field that studies the nano-scale neuronal connectivity of the mammalian brains.  Using high-resolution, large-scale optical and electron microscope images plays a central role in connectomics research,  but it also poses very challenging computational problems for 3D segmentation and visualization in terms of developing suitable algorithms,  coping with the ever-increasing data sizes, and maintaining interactive performance. Emerging massively parallel computing systems, such as graphics processors (GPUs), can be a solution for such computation-demanding tasks due to its scalable and parallel architecture. In this talk, I will introduce my research in GPU-accelerated biomedical image analysis. First, I will talk about the Fast Iterative Method, a parallel algorithm to solve a class of Hamilton-Jacobi equations for weighted distance computation and its application in brain connectivity analysis. Second, I will introduce a semi-automatic segmentation method for elongated neuronal structures using minimum cost path and variational methods. Last, I will briefly introduce our recent development in compressive sensing MRI reconstruction on a multi-GPU system.