作者: Niccolo Camarlinghi , Francesco Bagagli , Piergiorgio Cerello , Alessandra Retico , Maria Evelina Fantacci
DOI: 10.1109/NSSMIC.2012.6551464
关键词: Artificial intelligence 、 Computational science 、 CUDA 、 Voxel 、 Filter (video) 、 Computer vision 、 Tomography 、 Hessian matrix 、 General-purpose computing on graphics processing units 、 CAD 、 Computer science 、 Image processing
摘要: The aim of this work is the efficient implementation Hessian based filters. These filters are commonly used in medical image analysis and employed Voxel Based Neural Approach (VBNA) lung CAD (Computer Aided Detection) system for nodule detection. This mainly focuses on optimization filter devoted to detection internal candidates, called Multi Scale Dot Enhancement (MSDE) algorithm. Two fast variants MSDE algorithm here proposed compared: first relies an analytical it implemented a standard CPU, whereas second consists implementing CUDA Graphical Processing Unit (GPU) framework. algorithms were tested with computed tomography images belonging Lung Image Database Consortium (LIDC) public research database using Intel Core i7 950 @ 3.07GHz NVIDIA GeForce GTX 580. Both approaches lead improvement performance respect original implementation, without any loss precision. initial realized Insight ToolKit open source framework (ITK), had average execution time 69 sec per CT five scales enhancement. analyticallyoptimized CPU leads computational speed gain 2.5× (28 CT), parallel speed-up 38x (1.8 CT) 15x approach. has been developed INFN-funded MAGIC-5 project.