Detecting Malignant Tumors Using a Network of Spiking Neurons Through a Population Based Stochastic Temporal Encoding
Agnivo Ghosh, Ayan Chakraborty, Aniruddha Chandra, Sumit Kundu, Saswat Chakrabarti · 2024
Spiking neural network (SNN) has gained attention of researchers for its energy efficient and brain-like cognitive abilities. SNNs are now being explored for medical diagnosis to leverage their benefit of sparse, event driven and energy-efficient information processing. In this work an SNN has been designed to classify malignant and benign tumors from cell tissue data obtained from the Wisconsin breast cancer dataset. A novel population-based stochastic temporal encoding scheme has been adopted to convert the cell tissue features into equivalent neuronal spike trains. A supervised learning algorithm has been developed in synergy with the encoding scheme. Our method has resulted in an average accuracy of 90.8% and a best accuracy of 94% for the classification task. The proposed method performs reasonably well (within 2%) in comparison to popular machine learning algorithms such as Support Vector Machine (SVM) with linear kernel and Multi Layered Perceptron (MLP). The present method has a promising potential for further exploration in medical and healthcare applications.