Improved approach towards classification of histopathology images using bag-of-features

The paper addresses the problem of detecting malignancy in histopathological sample of breast tissue. The proposed method uses bag-of-features approach along with color normalization. The proposed method uses color normalization as pre-processing step prior to feature extraction. SIFT/DCT is used for feature extraction. The images are described using SIFT/DCT descriptor each of 128/192 dimension. Further these image descriptors are quantized using bag-of-feature into predefined codebook of sizes 150 and 500. In classification, image is classified using multi-class SVM classifier which groups image into suitable class namely normal, in-situ, and invasive. The proposed method reports classification accuracy as 100% for ductal carcinoma in situ, 98.88% for invasive carcinoma, and 100% for normal class images. Experimentally, we have found that color normalization as pre-processing step improves the classification accuracy.

[1]  B. Yener,et al.  Automated cancer diagnosis based on histopathological images : a systematic survey , 2005 .

[2]  Konstantinos N. Plataniotis,et al.  A Complete Color Normalization Approach to Histopathology Images Using Color Cues Computed From Saturation-Weighted Statistics , 2015, IEEE Transactions on Biomedical Engineering.

[3]  Fabio A. González,et al.  Visual pattern mining in histology image collections using bag of features , 2011, Artif. Intell. Medicine.

[4]  Yousef Al-Kofahi,et al.  Improved Automatic Detection and Segmentation of Cell Nuclei in Histopathology Images , 2010, IEEE Transactions on Biomedical Engineering.

[5]  Anant Madabhushi,et al.  Automated grading of breast cancer histopathology using spectral clustering with textural and architectural image features , 2008, 2008 5th IEEE International Symposium on Biomedical Imaging: From Nano to Macro.

[6]  A. Madabhushi,et al.  Histopathological Image Analysis: A Review , 2009, IEEE Reviews in Biomedical Engineering.

[7]  Lin Yang,et al.  Robust Segmentation of Overlapping Cells in Histopathology Specimens Using Parallel Seed Detection and Repulsive Level Set , 2012, IEEE Transactions on Biomedical Engineering.

[8]  Angel Cruz-Roa,et al.  Automatic annotation of histopathological images using a latent topic model based on non-negative matrix factorization , 2011, Journal of pathology informatics.

[9]  Jing Li,et al.  A comprehensive review of current local features for computer vision , 2008, Neurocomputing.

[10]  G LoweDavid,et al.  Distinctive Image Features from Scale-Invariant Keypoints , 2004 .

[11]  Smriti H. Bhandari.  A bag-of-features approach for malignancy detection in breast histopathology images , 2015, 2015 IEEE International Conference on Image Processing (ICIP).