An Efficient Neural Network Based System for Diagnosis of Breast Cancer

Breast Cancer is one of the fatal diseases causing more number of deaths in women. Constant efforts are being made to develop more efficient techniques for early and accurate diagnosis of breast cancer. Classical methods required cytopathologists or oncologists to examine the breast lesions for detection and classification of various stages of the cancer. Such manual attempts have proven to be time consuming and inefficient in many cases. Hence there is a need for efficient methods that diagnoses the cancerous cells without human involvement with high accuracies. This paper proposes an automated technique using artificial neural networks as decision making tools in the field of breast cancer. The features extracted from biopsy slide images are used to train the neural network. Both supervised and unsupervised methods of neural networks are tested to develop the most efficient alternative for breast cancer diagnosis. Self-organization map (SOM) method under unsupervised techniques is used to classify the WDBC dataset into benign and malignant. Under supervised method, a variant of back propagation algorithm, scaled conjugate gradient is investigated for the same. The generalization capability of the network is improved using the Bayesian regularization technique. Index Terms Artificial Neural Network, Breast cancer diagnosis, SOM, LVQ, Back propagation, Generalization, Bayesian regularization.