Refraction corrected transmission ultrasound computed tomography for application in breast imaging

Purpose: We present an iterative framework for CT reconstruction from transmission ultrasound data which accurately and efficiently models the strong refraction effects that occur in our target application: Imaging the female breast. Methods: Our refractive ray tracing framework has its foundation in the fast marching method (FNMM) and it allows an accurate as well as efficient modeling of curved rays. We also describe a novel regularization scheme that yields further significant reconstruction quality improvements. A final contribution is the development of a realistic anthropomorphic digital breast phantom based on the NIH Visible Female data set. Results: Our system is able to resolve very fine details even in the presence of significant noise, and it reconstructs both sound speed and attenuation data. Excellent correspondence with a traditional, but significantly more computationally expensive wave equation solver is achieved. Conclusions: Apart from the accurate modeling of curved rays, decisive factors have also been our regularization scheme and the high-quality interpolation filter we have used. An added benefit of our framework is that it accelerates well on GPUs where we have shown that clinical 3D reconstruction speeds on the order of minutes are possible.

A multi-GPU algorithm for large-scale neuronal networks

Large-scale simulations of parts of the brain using detailed neuronal models to improve our understanding of brain functions are becoming a reality with the usage of supercomputers and large clusters. However, the high acquisition and maintenance cost of these computers, including the physical space, air conditioning, and electrical power, limits the number of simulations of this kind that scientists can perform. Modern commodity graphical cards, based on the CUDA platform, contain graphical processing units (GPUs) composed of hundreds of processors that can simultaneously execute thousands of threads and thus constitute a low-cost solution for many high-performance computing applications. In this work, we present a CUDA algorithm that enables the execution, on multiple GPUs, of simulations of large-scale networks composed of biologically realistic Hodgkin-Huxley neurons. The algorithm represents each neuron as a CUDA thread, which solves the set of coupled differential equations that model each neuron. Communication among neurons located in different GPUs is coordinated by the CPU. We obtained speedups of 40 for the simulation of 200k neurons that received random external input and speedups of 9 for a network with 200k neurons and 20M neuronal connections, in a single computer with two graphic boards with two GPUs each, when compared with a modern quad-core CPU. Copyright (C) 2010 John Wiley & Sons, Ltd.