An Improved Neural Network for Mammogram Classification Using Genetic Optimization

Mammograms are digital images of breast used to diagnose masses. Micro calcifications seen in mammograms are indicators of tumour. Computer Aided Detection (CAD) systems play a very important role to detect mammographic abnormalities as they reduce the false negative rate. In this work, a novel weight optimized Multi-Layer Perceptron (MLP) based classifier with a Genetic Algorithm (GA) is proposed to select optimum input to the Neural Network and weight selection. The aim of weight selection is the reduction of the connections between the neurons without reducing the classification accuracy and hence speeding up the computation. Experimental results demonstrate that the classification accuracy is significantly improved with the proposed MLPNN GA weight selection. It is observed that proposed MLPNN GA weight selection with GA feature selection increases accuracy by 10.53% than MLP NN and by 8.69% when compared to MLPNN with GA feature selection.