Implementation of Wavelet Transform and Back Propagation Neural Network for Identification of Microcalcification in Breast

Wavelet decomposition has been applied to mammogram image to obtain four different coefficients for 5 levels of decompositions. The coefficients are: low frequency coefficients (A), vertical high frequency coefficients (V), horizontal high frequency coefficients (H), and diagonal high frequency coefficients (D). The features of the mammography image are obtained using the wavelet transform selecting the different levels of decompositions. The proposed method presents a new classification approach to microcalcification (MC) detection in mammograms using wavelet and back propagation algorithm (BPA) Neural Network. These features obtained from wavelet are representation of MC as well as other information of the image. Daubauchi wavelet has been used to decompose image to 5 levels. Statistical features are extracted from the wavelet coefficients. Training the BPA with features and testing the BPA to identify the presence of MC has been done. The percentage identification is above 96.2%. The performance of the proposed method based on the quality of the mammogram image.