Home | Issues | Profile | History | Submission | Review
Vol: 58(72) No: 1 / March 2013      

Simple Texture Descriptors for Classification of Mitotic Cells Figures of Histopathological Images from Breast Cancer
Mircea-Sebastian Serbanescu
Department of Medical Informatics, University of Medicine and Pharmacy of Craiova, Faculty of Medicine, Petru Rares St. , No. 2, 200349, Craiova, Romania, phone: (0040) 351-443-561, e-mail: mircea_serbanescu@yahoo.com


Keywords: mitosis detection, computer aided diagnosis, automated mitotic index

Abstract
Confirmation of clinical breast cancer diagnosis is done histopathologically on microscopic slides taking into consideration cell modifications, architecture modifications and mitotic cell index. Mitotic cells count (multiplying cells) is a separate index represented by counting the number of multiplying cells figures in at least 10 high power magnification microscopic fields (20x, 40x). It is a time consuming and demanding method, with low reproducibility, so it is suitable for an automated (computer aided) method. A total number of 9099 cell nucleus images, with 184 mitotic figures, were obtained from 10 high power microscope fields (40x, 0.2456 µm/pixel) of breast cancer images. All mitotic figures were manually annotated by several pathologists. Our study aimed to see if simple texture descriptors (Contrast, Correlation, Energy, Homogeneity, Entropy) are suitable for an automated diagnosis of mitotic cells. We applied a k-means clustering algorithm to see the aggregation of mitotic and non-mitotic cells image descriptors. No clear cluster of mitotic figures was obtained. There were some non-mitotic, smaller, compact clusters but we found them irrelevant for our study. Furthermore we have trained a feed foreword neural network with two hidden layers for the classification task. The overall prediction of the network was poor. We conclude that the simple texture descriptors taken into consideration in our study are insufficient for detecting mitotic figures, probably because of the heterogeneity of the mitotic figure.

References
[1] V. Roullier, V.-T Ta, O. Lézoray and A. Elmoataz, “Graph-based multi-resolution segmentation of histological whole slide images”, 2010 IEEE International Symposium on Biomedical Imaging: From Nano to Macro, pp. 153-156.
[10] D. Diaconescu, S. Diaconescu, A. Chesca and S. Toma, “The Value of Mitotic Counting in Prostate Carcinomas”, International Journal of Mathematical Models and Methods in Applied Sciences, Issue 2, Volume 5, 2011, pp. 379-386
[11] M. Sarbia, F. Bittinger, R. Porschen, P. Dutkowski, M. Torzewski, R. Willers, and H. E. Gabbert, “The prognostic significance of tumour cell proliferation in squamous cell carcinomas of the oesophagus”, Br J Cancer. 1996 October; 74(7), pp. 1012–1016.
[12] P J van Diest, E van der Wall, and J P A Baak, “Prognostic value of proliferation in invasive breast cancer: a review”, J Clin Pathol. 2004 July; 57(7), pp. 675–681.
[13] J. S. Meyer, C. Alvarez, C. Milikowski, N. Olson, I. Russo, J. Russo, A. Glass, B.A. Zehnbauer, K. Lister and R. Parwaresch, “Breast carcinoma malignancy grading by Bloom-Richardson system vs proliferation index: Reproducibility of grade and advantages of proliferation index”, Modern Pathology 18, 2005, pp. 1067–1078.
[14] P.M.L Drezet, R.F Harrison and S.S. Cross, Intelligent Methods, “Automatic mitotic index estimation for the prognostication of breast cancer from histology images”, IEE Colloquium on Healthcare and Medical Applications (Digest No. 1998/514), pp. 14/1-14/3.
[15] C.-H. Huang, A. Veillard, D. Racoceanu, N. Lomenie, and L. Roux. “Time-efficient sparse analysis of histopathological whole slide images”, accepted by Computerized Medical Imaging and Graphics, Elsevier, 2010.
[16] C. Somme, L. Fiaschi, F. A. Hamprecht and D. W. Gerlich, “Learning-based mitotic cell detection in histopathological images”, Available from http://hci.iwr.uni- heidelberg.de/publications/mip/techrep/sommer_12_learning-based.pdf
[17] J. A. Belien, J. P. Baak, P. J. van Diest, and A. H. van Ginkel, “Counting mitoses by image processing in Feulgen stained breast cancer sections: The influence of resolution”, Cytometry, vol. 28, pp. 135–40, Jun. 1997.
[18] Latson L, Sebek B and Powell KA., “Automated cell nuclear segmentation in color images of hematoxylin and eosin-stained breast biopsy”, Anal Quant Cytol Histol. 2003 Dec;25(6):321-31.
[19] M.N. Gurcan, L.E. Boucheron,, A/ Can, A. Madabhushi, N.M. Rajpoot, B. Yener, “Histopathological Image Analysis: A Review”, Biomedical Engineering, IEEE Reviews in , vol.2, no., pp.147-171, 2009.
[20] D. L. Weaver, D. N. Krag, E. A. Manna, T. Ashikaga, S. P. Harlow, and K. D. Bauer, “Comparison of pathologist-detected and automated computer-assisted image analysis detected sentinel lymph node micrometastases in breast cancer”, Mod. Pathol., vol. 16, pp. 1159–63, Nov. 2003.
[21] M.S. Serbanescu, “Clustering of Mitotic Cells Figures of Histopathological Images from Breast Cancer”, Proceedings Conference of the Romanian Society of Medical Informatics ROMEDINF2012, Timisoara, Romania, pp. 98 – 102, 2012