Abstract—This paper presents an efficient feature extraction technique, i.e., Bidimensional Empirical Mode Decomposition (BEMD) for mammogram images. The EMD approach is fully adaptive and data driven, proved reliable for monodimensional signals. BEMD is used to extract features at numerous scales or spatial frequencies. Five statistical textural features had been extracted from preprocessed digital mammograms by using BEMD. These features are mean, standard deviation, kurtosis, skewness and entropy. Artificial Neural Network (ANN) was employed to distinguish mass and non mass tissue based on these features. An accuracy of 92.5%, sensitivity of 87.5% and specificity of 96.55% was obtained by proposed method which is better than other methods.

Index Terms—BEMD, ANN, Region of Interest (ROI), Classification, Feature extraction

Cite: Srishti Sondele and Indu Saini, "DIGITAL MAMMOGRAMS CLASSIFICATION USING ARTIFICIAL NEURAL NETWORK BASED ON BIDIMENSIONAL EMPIRICAL MODE DECOMPOSITION FEATURES," International Journal of Electrical and Electronic Engineering & Telecommunications, Vol. 2, No. 3, pp. 44-48, July 2013.