International Research Journal of Education and Technology

Classifying brain tumours is essential for both medical diagnosis and therapy planning. Using MRI pictures, this study uses a specially designed Convolutional Neural Network (CNN) to classify brain tumours into four groups: pituitary, meningioma, glioma, and no tumour. 7,022 MRI scans total, split into training, validation, and testing sets, make up the dataset. To improve model generalisation, the preprocessing pipeline applies data augmentation, normalises pixel values, and resizes images to 224x224 pixels. Convolutional, pooling, and dense layers are all part of the model, which was constructed with TensorFlow and Keras and optimised with the Adamax optimiser.After 12 epochs, the model showed robustness and dependability with a training accuracy of 98% and a validation accuracy of 95%. To assess the model's efficacy, important performance indicators like the classification report and confusion matrix were examined.A Flask backend and a React-based frontend are used for deployment, enabling real-time forecasts. This project demonstrates how deep learning can improve medical imaging and judgement.

Key Words: Brain Tumor Classification, Convolutional Neural Network (CNN), MRI Images, Deep Learning, Medical Imaging, TensorFlow, Keras, Adamax Optimizer, Flask, React, Data Preprocessing, Model Deployment, Healthcare AI, Real- time Prediction etc…