Francisco Xavier

GPU Power for Medical Imaging

A Practical Approach to General-Purpose Computing, with CUDA, Machine Learning and Medical Imaging

• Broschiertes Buch

Produktbeschreibung

The Book addresses the use of Graphic Processor Units (GPUs) to optimize an existing classification algorithm based on Support Vector Machines, through parallelization with the NVIDIA CUDA framework. This algorithm is used for automatic topographic segmentation of gastrointestinal tract videos, obtained through capsule endoscopy used in real Hospital environments. After optimization of the sequential version of the algorithm, two approaches were compared: 1) the simple use of CUDA libraries, and 2) the development of full code, both Host and device. Both approaches granted substantial speedups up to ~7 times with a single GPU, and up to ~27 times with a cluster of four GPUs. With the biggest video, the classification time went down from 1h25min to less than two minutes, suggesting not only that this was a successful integration, but also that GPUs are very well suited for (and even be the answer to) several current issues that Medicine stumbles upon.

Produktdetails

• Verlag: LAP Lambert Academic Publishing
• Seitenzahl: 92
• Erscheinungstermin: 10. März 2014
• Englisch
• Abmessung: 220mm x 150mm x 6mm
• Gewicht: 155g
• ISBN-13: 9783659251894
• ISBN-10: 3659251895
• Artikelnr.: 40539816

Autorenporträt

High-Performance Engineer, Graduated with Masters from the University of Aveiro, Portugal. His research addresses the use of General-Purpose computation in several Topics, which range from Capsule Endoscopy to Mass Spectrometry. His current experience abroad covers Portugal, India and Germany, and a Ph.D research follows along.