Breast Cancer Detection Using Polynomial Fitting Applied on Contrast Enhanced Spectral Mammography

This paper presents a polynomial fitting technique on pixels inside the Region Of Interest (ROI) to obtain a new feature set that can be used to differentiate between normal and cancerous cases. In this paper few features have been used if it compared to other previous works that have been used many feature sets, and then used different techniques to reduce the dimensionality, which might complicate the system also require a great computational cost. These few features are used without any need to reduce the dimensionality then it is applied on two different classifiers, the K-nearest neighbor (KNN) and the artificial neural network (ANN). The proposed techniques have been applied on 50 real cases from `BAHEYA Foundation for Early Detection & Treatment of Breast Cancer' with the help of physicians and radiologists in the hospital. The used images are Contrast Enhanced Spectral Mammograms (CESM) that has clearer and more contrast images than the conventional mammograms. Applying the proposed feature set on these CESM mammograms using the two different classifiers-the K-nearest neighbor (KNN) and the artificial neural network (ANN) - give accuracies of 96% and 92% respectively.