Neural networks for the detection and localization of breast cancer

Abstract

This paper investigates the use of neural networks to detect and locate early breast cancer using a simple feed-forward back-propagation neural network. In order to test the proposed algorithm, an electromagnetic simulator is used to build a three-dimensional breast model. Spherical tumors of radii 1 mm, 2 mm, 4 mm, and 5 mm are assumed to be at different locations in the breast model. An ultra-wideband pulse is transmitted towards the breast model and four probes are located around the breast to capture the scattered signals. The collected signals are then analyzed using the neural networks to get useful information concerning the presence or otherwise of the tumor and its location if it does exist. The obtained results from using the proposed method are promising with 100% success in the detection and 95% success in the localization.