Early detection of brain tumors is very important for effective treatment. Traditional methods like biopsy are invasive and risky. This study explores the use of artificial intelligence to automatically detect and classify brain tumors from MRI images. Two deep learning models—a new 2D Convolutional Neural Network (2D-CNN) and an autoencoder-based model—were developed and compared with six traditional machine learning methods. The models classified MRI images into four categories: glioma, meningioma, pituitary tumor, and healthy brain. A dataset of 3,264 MRI images was used, along with image preprocessing and data augmentation. The proposed 2D-CNN achieved the best performance, with high accuracy, recall, and excellent ROC-AUC values close to 1.0. Among traditional methods, KNN performed best, while MLP showed poor results. Statistical analysis showed that deep learning models performed significantly better than traditional methods. Overall, the proposed 2D-CNN is accurate, simple, and fast, making it suitable for clinical use in early brain tumor diagnosis.

Keywords—Brain tumor detection, Magnetic resonance imaging (MRI), Deep learning, Convolutional neural network (CNN), Autoencoder, Machine learning, Medical image classification, Computer-aided diagnosis: