Hybrid approach towards feature selection for breast tumour classification from screening mammograms

A hybrid approach has been developed to extract the optimal features from the breast tumours using hybrid harmony search and presented in this paper. The texture feature, intensity histogram feature, radial distance feature and shape features have been extracted and the optimal feature set has been obtained using hybrid harmony search (HHS). The hybrid scheme for feature selection is obtained by combining cuckoo search and harmony search. The minimum distance classifier, k-NN classifier and SVM classifier are used for classification purpose and its produces 98.19%, 98.34% and 97.18% average classification accuracy respectively with minimum number of features. The performance of the new hybrid algorithm is compared with the genetic algorithm, particle swarm optimisation algorithm, cuckoo search and harmony search. The result shows that the hybrid of cuckoo and harmony search algorithm is more accurate than the other algorithm. The proposed system can provide valuable information to the physician in medical pathology.