Enhanced Brain Tumor Classification with Convolutional Spiking Neural Networks

Medical image analysis has brought a particular interest amongst researchers over the last few years as it exposes challenges to improve the way we diagnose and treat illnesses and diseases. The wide variety of shapes and locations brain tumors can take make diagnoses difficult to establish and automate. Continuously offering more accuracy, Artificial Neural Networks have been widely studied and successfully adapted to solve medical image processing tasks. However, they suffer from their significant memory and energy cost as well as a high dependency to powerful devices, preventing their realistic deployment in the medical field. A recent growth of interest for Spiking Neural Networks (SNN) is leading to the development of faster, lighter and power efficient models proven to tackle the aforementioned challenges in cognitive tasks such as object recognition, speech recognition, image classification and segmentation. In this work, we present the implementation of a single trainable layer Convolutional Spiking Neural Network (C-SNN) for brain tumor classification in order to prove the efficiency of spiking models in medical image analysis.