Classification of colorectal polyp regions in optical projection tomography

The potential of optical projection tomography (OPT) to enhance colorectal polyp diagnosis is beginning to be explored. This paper presents, to the best of our knowledge, the first study on automatic image analysis of OPT images of colorectal polyps. 3D regions are classified using the bag of visual words framework and support vector machines. Independent subspace analysis is used to learn a domain-specific feature dictionary. This is compared to the use of raw patches (after random projection) and local binary patterns. Classification experiments (across patients) at the patch level and at the region level are presented using a set of 30 expert-annotated OPT images. Results show that accurate classification of 3D OPT image regions is feasible using this approach; regions of low-grade dysplasia and invasive cancer were discriminated with approximately 90% accuracy.

[1]  Heikki Mannila,et al.  Random projection in dimensionality reduction: applications to image and text data , 2001, KDD '01.

[2]  Matti Pietikäinen,et al.  Multiresolution Gray-Scale and Rotation Invariant Texture Classification with Local Binary Patterns , 2002, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Guido Gerig,et al.  User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability , 2006, NeuroImage.

[4]  Quoc V. Le,et al.  Learning hierarchical invariant spatio-temporal features for action recognition with independent subspace analysis , 2011, CVPR 2011.

[5]  A. Ercil,et al.  Robustness of Local Binary Patterns in Brain MR Image Analysis , 2007, 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[6]  Matti Pietikäinen,et al.  Dynamic Texture Recognition Using Local Binary Patterns with an Application to Facial Expressions , 2007, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7]  Maria Coats,et al.  1028 Optical Projection Tomography: Can it Help Diagnose the Colorectal Polypoid Cancer? , 2012 .

[8]  Ron Kohavi,et al.  A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection , 1995, IJCAI.

[9]  Quoc V. Le,et al.  Learning invariant features of tumor signatures , 2012, 2012 9th IEEE International Symposium on Biomedical Imaging (ISBI).

[10]  J. Vane,et al.  Optical Projection Tomography as a Tool for 3D Microscopy and Gene Expression Studies , 2002 .

[11]  Loris Nanni,et al.  Local binary patterns variants as texture descriptors for medical image analysis , 2010, Artif. Intell. Medicine.

[12]  Cordelia Schmid,et al.  Local Features and Kernels for Classification of Texture and Object Categories: A Comprehensive Study , 2006, 2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'06).

[13]  Paul W. Fieguth,et al.  Texture Classification from Random Features , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[14]  Chih-Jen Lin,et al.  LIBLINEAR: A Library for Large Linear Classification , 2008, J. Mach. Learn. Res..

[15]  S. Raab,et al.  Inter-observer variation in the histological diagnosis of polyps in colorectal cancer screening , 2012 .

[16]  Aapo Hyvärinen,et al.  Emergence of Phase- and Shift-Invariant Features by Decomposition of Natural Images into Independent Feature Subspaces , 2000, Neural Computation.

[17]  C. Mathers,et al.  Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008 , 2010, International journal of cancer.