GhostNet Model Based Brain Tumor Classification

Brain tumor is a severe and widely spread disease nowadays that must be classified accurately and promptly to treat it effectively. Hence, many researchers have shown their interest in it and devised different models for brain tumor classification today. Rule based models, statistical models, machine learning, deep learning, etc. are used for brain tumor classification. Deep learning based models usually have better performance compared to other models. In the proposed work, we inducted a new deep learning model i.e. the GhostNet model for brain tumor classification. The GhostNet model incorporates ‘Ghost modules’ to replace resource-intensive convolutional layers, significantly reducing computational demands while preserving robust performance in feature extraction and classification. Additionally, Feature Fusion Blocks (FFB) enhance feature extraction capabilities by merging information from different network layers. The GhostNet model is learned on a diverse number of MRI images, achieving an accuracy of 85.25% in differentiating between various tumor types and non-tumor images.