An ensemble approach to diagnose breast cancer using fully complex-valued relaxation neural network classifier

This paper presents a new improved classification technique using Fully Complex-Valued Relaxation Networks (FCRN) based ensemble technique for classifying mammogram images. The system is developed based on three stages of breast cancer, namely normal, benign and malignant defined by the MIAS database. Features like binary object features, RST invariant features, histogram features, texture features and spectral features are extracted from the MIAS database. Extracted features are then given to the proposed FCRN-based ensemble classifier. FCRN networks are ensembled together for improving the classification rate. Receiver Operating Characteristic (ROC) analysis is used for evaluating the system. The results illustrate the superior classification performance of the ensembled FCRN. The resultant ensembled FCRN approximates the desired output more accurately with a lower computational effort.