Hybrid model for multi-class brain tumor classification

This project focuses on designing and implementing a hybrid deep learning model for multi-class classification of brain tumor images. The classification process aims to distinguish between multiple types of brain tumors using advanced convolutional neural network (CNN) architectures. The hybrid model integrates multiple deep learning components, including Multi-Branch CNN, EfficientNetB2, MobileNetV2, DenseNet121, and SE (Squeeze-and-Excitation) Blocks. These models are strategically combined to extract deep hierarchical features, enhance model generalization, and improve classification accuracy. The input dataset consists of brain tumor images obtained from a structured image dataset available on a data-sharing platform. Preprocessing techniques are applied to normalize and resize the images for better training performance. Feature extraction and fusion are carried out through combined model architectures. The model is trained using supervised learning techniques and validated using evaluation metrics such as accuracy, precision, recall, and F1-score. The front-end of the system is built using HTML, CSS, and JavaScript, while the back-end is powered by Python and Flask. The system allows users to register, log in, classify tumor images, and log out. The classification module uses the hybrid model to predict the type of brain tumor from the input image. This approach leverages both the robustness of CNN architectures and the efficiency of modern neural networks to produce reliable classification outputs. The fusion of various models aims to reduce overfitting, handle image variations effectively, and ensure faster convergence. The modular implementation also ensures better scalability and maintainability. Overall, the proposed hybrid model provides a systematic approach for solving multi-class image classification tasks with improved efficiency and higher accuracy.

Keywords: Brain Tumor, Multi-Class Classification, Hybrid Model, CNN, EfficientNetB2, MobileNetV2, DenseNet121, SE Blocks, Deep Learning, Flask