Mammographic Image Classification System via Active Learning
- 106 Downloads
Training an accurate prediction model for mammographic image classification is usually necessary to require a large number of labeled images. However, the manually acquiring rich and reliable annotations is known to be tedious and time-consuming process, especially for medical image. The advances in machine learning yielded a branch of technique, termed active learning (AL), which has been proposed for solving the problem of the limited training samples and expensive labeling cost, and has resulted in highly successful applications in many pattern recognition tasks such as image processing and speech recognition. In this article, a comparison is provided among the mammographic image classification systems, relying on traditional supervised learning, un-supervised learning and AL, aiming to obtain a system with low labeling cost. The experiments based on digital database for screening mammography demonstrate that the AL is able to minimize the labeling cost of mammographic image without sacrificing the accuracy of final classification system. In addition, some specific characteristics of mammographic image: file information and spatial feature, which are not available to the traditional AL methods, have been found to further decrease the labeling cost. In conclusion, we suggest that the AL is a reasonable alternative to supervised learning for the researchers in the field of medical image classification with limited experimental conditions.
KeywordsImage classification Active learning Mammography Labeling cost
This research is partially supported by the National Key Research Program of China (2016YFC0106200), the 863 national research fund (2015AA043203) of China, the National Natural Science Foundation of China (81301283, 61190124 and 61271318), and the special funding of capital health research and development with No. 2016-1-4011. The authors are grateful to the Massachusetts General Hospital, the University of South Florida, and Sandia National Laboratories, which provides DDSM as a resource for our experimental data. We also express our sincere gratitude towards Department of Computer Science in University of North Carolina at Charlotte for their free tech support.
Compliance with Ethical Standards
Conflict of interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
The study doesn’t involve human or animal subjects.
- 2.World Health Organization. (2012). International Agency for Research on Cancer GLOBOCAN 2012: Estimated cancer incidence, mortality and prevalence worldwide in 2012. Geneva: WHO.Google Scholar
- 6.Bekker, A. J., Shalhon, M., Greenspan, H., & Goldberger, J. (2015). Learning to combine decisions from multiple mammography views. In 2015 IEEE 12th international symposium on biomedical imaging (ISBI) (pp. 97–100). IEEE. https://doi.org/10.1109/isbi.2015.7163825.
- 12.Junior, G. B., da Rocha, S. V., Gattass, M., Silva, A. C., & de Paiva, A. C. (2013). A mass classification using spatial diversity approaches in mammography images for false positive reduction. Expert Systems with Applications, 40, 7534–7543. https://doi.org/10.1016/j.eswa.2013.07.034.CrossRefGoogle Scholar
- 13.de Oliveira, F. S. S., de Carvalho Filho, A. O., Silva, A. C., de Paiva, A. C., & Gattass, M. (2015). Classification of breast regions as mass and non-mass based on digital mammograms using taxonomic indexes and SVM. Computers in Biology and Medicine, 57, 42–53. https://doi.org/10.1016/j.compbiomed.2014.11.016.CrossRefGoogle Scholar
- 14.Kashyap, K. L., Bajpai, M. K., & Khanna, P. (2017). Globally supported radial basis function based collocation method for evolution of level set in mass segmentation using mammograms. Computers in Biology and Medicine, 87, 22–37. https://doi.org/10.1016/j.compbiomed.2017.05.015.CrossRefGoogle Scholar
- 18.Oliver, A., Marti, J., Marti, R., Bosch, A., & Freixenet, J. (2006). A new approach to the classification of mammographic masses and normal breast tissue. In 18th International conference on pattern recognition, 2006. ICPR 2006 (pp. 707–710). IEEE. https://doi.org/10.1109/icpr.2006.113.
- 20.Raghavendra, U., Acharya, U. R., Fujita, H., Gudigar, A., Tan, J. H., & Chokkadi, S. (2016). Application of Gabor wavelet and Locality Sensitive Discriminant Analysis for automated identification of breast cancer using digitized mammogram images. Applied Soft Computing, 46, 151–161. https://doi.org/10.1016/j.asoc.2016.04.036.CrossRefGoogle Scholar
- 21.Jiang, F., Liu, H., Yu, S., & Xie, Y. (2017). Breast mass lesion classification in mammograms by transfer learning. In Proceedings of the 5th international conference on bioinformatics and computational biology, 2017 (pp. 59–62). ACM. https://doi.org/10.1145/3035012.3035022.
- 25.Settles, B. (2010). Active learning literature survey 52-11. Madison, WI: University of Wisconsin.Google Scholar
- 28.Shannon, C. E. (2001). A mathematical theory of communication. ACM SIGMOBILE Mobile Computing and Communications Review, 5, 3–55. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x.CrossRefGoogle Scholar
- 33.Lewis, D. D., & Catlett, J. (1994). Heterogeneous uncertainty sampling for supervised learning. In Machine learning proceedings 1994 (pp. 148–156). Elsevier. https://doi.org/10.1016/b978-1-55860-335-6.50026-x.
- 34.Settles, B., Craven, M., & Ray, S. (2008). Multiple-instance active learning. In Advances in neural information processing systems, 2008 (pp. 1289–1296).Google Scholar
- 35.Olsson, F. (2009). A literature survey of active machine learning in the context of natural language processing. Swedish Institute of Computer Science.Google Scholar
- 36.Hoi, S. C., Jin, R., Zhu, J., & Lyu, M. R. (2006). Batch mode active learning and its application to medical image classification. In Proceedings of the 23rd international conference on machine learning, 2006 (pp. 417–424). ACM. https://doi.org/10.1145/1143844.1143897.
- 37.Rubens, N., Elahi, M., Sugiyama, M., & Kaplan, D. (2015). Active learning in recommender systems. In Recommender systems handbook (pp. 809–846). Springer. https://doi.org/10.1007/978-0-387-85820-3_23.
- 39.Heath, M., Bowyer, K., Kopans, D., Kegelmeyer, P., Moore, R., Chang, K., & Munishkumaran, S. (1998). Current status of the digital database for screening mammography. In Digital mammography (pp. 457–460). Springer. https://doi.org/10.1007/978-94-011-5318-8_75.
- 40.USF digital mammography home page (2007). http://marathon.csee.usf.edu/Mammography/Database.html.
- 41.Rose, C., Turi, D., Williams, A., Wolstencroft, K., & Taylor, C. (2006). Web services for the DDSM and digital mammography research. In International workshop on digital mammography, 2006 (pp. 376–383). Springer. https://doi.org/10.1007/11783237_51.
- 43.Dalal, N., & Triggs, B. (2005). Histograms of oriented gradients for human detection. In IEEE Computer Society conference on computer vision and pattern recognition, 2005. CVPR 2005 (pp. 886–893). IEEE. https://doi.org/10.1109/cvpr.2005.177.
- 51.Huang, H., Zhang, C., Hu, Q., & Zhu, P. (2016). Multi-view representative and informative induced active learning. In Pacific Rim international conference on artificial intelligence, 2016 (pp. 139–151). Springer. https://doi.org/10.1007/978-3-319-42911-3_12.