Pixel-Based Texture Classification of Tissues in Computed Tomography

Previous research has been done to classify different tissues/organs of interest present in medical images, in particular in Computed Tomography (CT) images. Most of the research used the anatomical structure present in the images in order to classify the tissues. In this paper, instead of using the anatomical structure, we propose a pixel-based texture approach for the representation and classification of the regions of interest. The approach incorporates various texture features and decision trees to accomplish tissue classification in normal Computed Tomography (CT) images of the chest and abdomen. First, we introduce a new “direction vs. displacement pairs” (DDP) approach to calculate a co-occurrence matrix for capturing all possible combination between directions and displacements necessary in calculating the texture features at the pixel-level. Second, we evaluate various different neighborhood sizes for the pixel-based texture representation in order to find the optimal window size for differentiating among 8 organs/tissues of interest: aorta, fat, kidney, liver, lung, muscle, spleen, and trabecular bone. For all organs/tissues (except for aorta), the optimal window was 13-by-13 allowing the classification sensitivity metric to be at least 96% for all organs/tissues. For aorta, the optimal window size was 9-by-9 with the classification sensitivity being 81%.

[1]  Robert M. Haralick,et al.  Textural Features for Image Classification , 1973, IEEE Trans. Syst. Man Cybern..

[2]  King-Sun Fu,et al.  Shape Discrimination Using Fourier Descriptors , 1977, IEEE Trans. Syst. Man Cybern..

[3]  Masaharu Kobashi,et al.  Knowledge-based organ identification from CT images , 1992, Medical Imaging.

[4]  Rajesh N. Dave,et al.  Robust shape detection using fuzzy clustering: practical applications , 1994, CVPR 1994.

[5]  Rangachar Kasturi,et al.  Machine vision , 1995 .

[6]  D. L. Kirch,et al.  Abdominal organ segmentation using texture transforms and a Hopfield neural network , 1999, IEEE Transactions on Medical Imaging.

[7]  George D. Magoulas,et al.  Detecting Abnormalities in Colonoscopic Images by Textural Description and Neural Networks , 1999 .

[8]  Nasser Kehtarnavaz,et al.  Texture based classification of mass abnormalities in mammograms , 2000, Proceedings 13th IEEE Symposium on Computer-Based Medical Systems. CBMS 2000.

[9]  Shigeo Abe DrEng.  Pattern Classification , 2001, Springer London.

[10]  Osmar R. Zaïane,et al.  Mammography Classification By an Association Rule-based Classifier , 2002, MDM/KDD.

[11]  Pau-Choo Chung,et al.  Identifying multiple abdominal organs from CT image series using a multimodule contextual neural network and spatial fuzzy rules , 2003, IEEE Transactions on Information Technology in Biomedicine.

[12]  Artur Przelaskowski,et al.  Segmentation and feature extraction for reliable classification of microcalcifications in digital mammograms , 2003 .

[13]  Demetri Terzopoulos,et al.  Snakes: Active contour models , 2004, International Journal of Computer Vision.

[14]  Jacob D. Furst,et al.  Classification of Tissues in Computed Tomography using Decision Trees , 2004 .

[15]  Jacob D. Furst,et al.  A classification approach for anatomical regions segmentation , 2005, IEEE International Conference on Image Processing 2005.

[16]  Jacob D. Furst,et al.  Contrast enhancement of soft tissues in computed tomography images , 2006, SPIE Medical Imaging.