Skin removal techniques for breast cancer radar detection based on artificial neural networks

The paper presents new real skin artifact removal techniques when applied to a novel ANN based radar data processing approach for breast cancer detection. Realistic 2D breast geometries derived by the UWCEM database. When the ANN approach was trained and tested using an ideal cleaning technique the accuracy on the tumor detection was 91%. Using the best real cleaning technique proposed in this paper the ANN based approach is able to achieve a sensitivity of 90%, a specificity of 84% and an accuracy of 87%.