Conference Paper

An Approach Based on Immune Algorithm and SVM for Detection and Classification of Microcalcifications

Coll. of Comput. Sci. & Eng., South China Univ. of Technol., Guangzhou
DOI: 10.1109/ICBBE.2007.154 Conference: Bioinformatics and Biomedical Engineering, 2007. ICBBE 2007. The 1st International Conference on
Source: IEEE Xplore

ABSTRACT As the feature based detection and classification of microcalcifications (MCs) in digital mammograms is considered here as a machine-leaning problem, we investigate an approach using immune algorithm (IA) and support vector machine (SVM), called IA-SVM, to solve it. Firstly, because only support vectors (SVs) are needed to build the classification hyperplane, we compress the training set according to their intra-class and inter-class Euclidean distances without losing any SVs. Meanwhile, an IA based MCs' features selector is provided to select an optimal feature subset, which can construct the input vectors for the latter SVM training; Secondly, the compressed and optimized training samples are fed to a SVM based classifier to make the optimal classification hyperplane more efficiently and more effectively. Experiments demonstrate that our method has better computing performance than other traditional classifiers (training samples were compressed by about 15%) and yields a satisfying Az value (about 0.83).

0 Bookmarks
 · 
46 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we propose an efficient algorithm based on principal component analysis (PCA) and fuzzy support vector machine (SVM) for the diagnosis of breast cancer tumor. First, PCA algorithm is implemented to project high-dimensional breast tumor data into much lower dimensional space, then the processed data are classified by a fuzzy SVM classifier. Experimental and analytical results show that in the diagnosis of breast cancer tumor the proposed method can greatly speed up the training and testing of the classifier, get high testing correct rate and pick out untypical cases to be reexamined by experienced doctors, superior to the traditional rigid margin SVM classifier.
    Natural Computation, 2008. ICNC '08. Fourth International Conference on; 11/2008
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes an efficient algorithm based on manifold learning and support vector machine (SVM) for the diagnosis of breast cancer tumor. First, Isomap algorithm is implemented to project high-dimensional breast tumor data to much lower dimensional space, then the processed data are classified by the SVM. Experimental and analytical results show that in the diagnosis of breast cancer tumor the proposed method can greatly speed up the training and testing of the classifier and get high testing correct rate, superior to the classical principal component analysis (PCA) algorithm.
    Information and Automation, 2008. ICIA 2008. International Conference on; 07/2008