Chest CT automatic analysis for lung nodules detection implemented on a GPU computing system

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

[1]  Carl-Fredrik Westin,et al.  Tissue Classification Based on 3D Local Intensity Structures for Volume Rendering , 2000, IEEE Trans. Vis. Comput. Graph..

[2]  Ilaria Gori,et al.  A voxel-based neural approach (VBNA) to identify lung nodules in the ANODE09 study , 2009, Medical Imaging.

[3]  Bram van Ginneken,et al.  Multi-scale Nodule Detection in Chest Radiographs , 2003, MICCAI.

[4]  Maria Evelina Fantacci,et al.  Methodology for automated detection of internal and juxtapleural lung nodules in computed tomography images. , 2009 .

[5]  Joachim Kopp.  EFFICIENT NUMERICAL DIAGONALIZATION OF HERMITIAN 3 × 3 MATRICES , 2006 .

[6]  Qiang Li,et al.  Selective enhancement filters for nodules, vessels, and airway walls in two- and three-dimensional CT scans. , 2003, Medical physics.

[7]  Ilaria Gori,et al.  Combination of computer-aided detection algorithms for automatic lung nodule identification , 2011, International Journal of Computer Assisted Radiology and Surgery.

[8]  M. Abramowitz,et al.  Handbook of Mathematical Functions With Formulas, Graphs and Mathematical Tables (National Bureau of Standards Applied Mathematics Series No. 55) , 1965 .

[9]  Ilaria Gori,et al.  Comparing and combining algorithms for computer-aided detection of pulmonary nodules in computed tomography scans: The ANODE09 study , 2010, Medical Image Anal..

[10]  K. Doi,et al.  Computerized detection of lung nodules in thin-section CT images by use of selective enhancement filters and an automated rule-based classifier. , 2008, Academic Radiology.

[11]  Ilaria Gori,et al.  Algorithms for automatic detection of lung nodules in CT scans , 2011, 2011 IEEE International Symposium on Medical Measurements and Applications.

[12]  Guido Gerig,et al.  3D Multi-scale line filter for segmentation and visualization of curvilinear structures in medical images , 1997, CVRMed.

[13]  Rachid Deriche,et al.  Fast algorithms for low-level vision , 1988, [1988 Proceedings] 9th International Conference on Pattern Recognition.