Computer Aided Diagnosis of Microcalcification Clusters

Amr Badr1 and Mostafa Abdel Azim2
1Computer Science Department, Faculty of Computers and Information, Cairo University, ruaab@rusys.eg.net
2Computer Engineering Department, Faculty of Engineering and Technology
Arab Academy for Science and Technology, and Maritime Transport,
 melbakary@aast.edu
 

Abstract:
Classification of Microcalcification clusters in mammogram images was achieved as an early sign of breast cancer. Different image enhancement techniques were tested such as Histogram Equalization (HE), Adaptive Histogram Equalization (AHE) and Wavelet Techniques. A Feedforward Neural Network was then trained on malignant and benign mammograms, for the purpose of their future classification. It was realized that combined AHE and Wavelet enhancement techniques performed better as a prior step to Neural Network training and execution. The classification of microcalcification clusters was about 97% correct, which is a proof of its accuracy and validity in practical application.
Keywords: Breast Cancer, Microcalcification Clusters, Histogram Equalization, Adaptive Histogram Equalization,
Wavelet Enhancement, Feedforward Neural Networks, Classification.