An Efficient Spiking Neural Network Approach based on Spike Response Model for Breast Cancer Diagnostic

Asmaa Ourdighi and Abdelkader Benyettou

Department of Computer Science, University of Sciences and Technology of Oran-Mohamed Boudiaf, Algeria

Abstract: This study investigates the efficiency of the one-layered Spiking Neural Network (SNN) on the enhancing of the breast cancer diagnostic results. The proposed network is based on Spike Response Model (SRM) with multiple delays per connection. Beside its simplicity, this model allows to modeling the production of a biologically realistic response to incoming synaptic events. By using a supervised learning, the training process was founded around of an error-backpropagation algorithm depending only on the time of the first spike emitted. In experimentation, our approach was exclusively tested on Wisconsin Breast Cancer Database (WBCD). The results were evaluated in accuracy classification and the area under Receiver Operating Characteristics (ROC) curve (AUC). In summary, we achieved 99.26% of accuracy classification with an AUC equal to 0.992.

Keywords: SNN, SRM, a gradient descent rule, WBCD.

Received February 2, 2013; accepted March 19, 2014