Classification of Breast Lesions using Histopathology Images and Neural Network

Breast cancer occurs when a malignant tumor originates in the breast. As breast tumors mature, they may metastasize to other parts of the body. However, it is important to keep in mind that, if identified and properly treated while still in its early stages, breast cancer can be cured [1].To achieve the above target it is necessary to develop a computer-aided Diagnosis system which helps in better diagnosis of the condition. It can be achieved by using Digital Image Processing techniques to obtain the regions of interest which show extra growth in the breast. So, a system is developed to classify lesions into Benign (non-cancerous) and Malignant (cancerous) condition. To classify the lesions the stain-color is considered as the important criteria to remove the noise from the digital images. To achieve this, initially the region of interest is obtained using k-means clustering and shape features are extracted. The binary image obtained as the result is further given as an input to obtain the regions of interest using the marker-controlled watershed image segmentation approach. The result of the hybrid approach gives us texture features. Further, the combination of these features is considered for classification. The performance measures namely accuracy , sensitivity , specificity , precision of the system are calculated for Naïve Bayes , Support Vector Machine , Adaptive Boosting , Classification Tree, Random Forest and Feed-Forward Neural Network Classifier.