Digital mammogram segmentation and tumor detection using artificial neural network

Abstract

Breast cancer is one of the leading causes of cancer related death among women. Women over 40 years of age and those who have family history are recommended to take mammograms regularly for screening. Manual examination of mammograms may result in misdiagnosis due to human errors caused by visual fatigue. The main aim of this research is to improve the accuracy of tumor detection from mammograms. In this research, four approaches have been designed and implemented for tumor detection. The general structure of all these approaches of tumor detection consists of enhancement, segmentation, feature extraction and classification stages. The proposed approach1 for tumor detection relies on contour searching method that uses an iterative procedure. The approach 2 of tumor detection in mammogram follows the scheme of enhancement techniques using wavelet decomposition and morphological operations. In the approach 3, the digitized mammogram is complemented using image negative and dimensionally reduced using Discrete Wavelet Transform (DWT). In the approach 4, smoothing filtering is used to remove noise from the complemented image and DWT to reduce the dimension of the image. The roughness value is measured using Fractal dimension analysis. Adaptive Thresholding with Fast Segmentation is used for segmentation process, Back Propagation Neural network (BPN), Radial Basis Function network (RBF) and Support Vector Machine (SVM) classifiers for classification. Mini mammographic data base from Mammographic Image Analysis Society (MIAS) has been used for the implementation of these approaches. From the analysis of the obtained results it is found that BPN provides more accurate results comparing to RBF and SVM. The performance accuracy is increased in the approach 4. This approach provides a sensitivity of 99.1% which is higher comparing to the other approaches and the existing approaches. The result shows that the proposed approach 4 is efficient. newline newline newline