A BIO-INSPIRED SPIKING NEURAL NETWORK FOR AUTOMATED BRAIN TUMOR CLASSIFICATION USING MRI IMAGES

Abstract

Brain tumors are among the most serious neurological disorders and can pose a significant risk to human life if not diagnosed at an early stage. Magnetic Resonance Imaging (MRI) is widely used for brain tumor diagnosis because it provides detailed images of brain tissues. However, accurately classifying brain tumors from MRI images remains a challenging task due to variations in tumor size, shape, location, texture, and intensity. To overcome these challenges, this study presents a machine learning-based approach for brain tumor classification using a Spiking Neural Network (SNN). Inspired by the functioning of biological neurons, SNNs effectively learn complex patterns from medical images while offering improved computational efficiency. In the proposed approach, MRI images are first segmented using Sharp No New U-Net (SnnU-Net), which accurately identifies tumor regions with different shapes and intensity levels. Next, texture, shape, and intensity features are extracted to capture the distinctive characteristics of tumor tissues. These features are then used to train the Spiking Neural Network for accurate brain tumor classification. The proposed method is evaluated using the BraTS2019, BraTS2020, and BraTS2021 benchmark datasets, achieving classification accuracies of 99.67%, 99.89%, and 99.93%, respectively. The experimental results demonstrate that the proposed SNN-based machine learning model provides highly accurate and reliable brain tumor classification, outperforming conventional neural network-based approaches and showing strong potential for supporting computer-aided diagnosis in clinical practice.