Background: Accurate and early diagnosis of brain tumors from Magnetic Resonance Imaging (MRI) is crucial to improve patient outcomes. Existing methods rely on radiologist manual interpretation, which may be time-consuming and prone to human error.
Objective: The present study proposes an automated system for brain tumor detection using a fine-tuned EfficientNet-B4 Convolutional Neural Network (CNN) to enhance the accuracy of classification and reduce the time for diagnosis.
Objective: A proposed automatic brain tumor detection system with an efficient fine-tuned EfficientNet-B4 Convolutional Neural Network (CNN) to improve classification accuracy and minimize diagnostic time.
Methods: The system utilizes preprocessing methods (resizing, normalization, noise removal, and contrast enhancement), data augmentation, and a hybrid method that combines CNN-based feature extraction and optimization methods. The model is trained and validated on an MRI dataset.
Results: The system reaches high accuracy of tumor detection, classification (malignant/benign), and localization, while it brings the computational expense dramatically down when compared to traditional hybrid models.

Keywords: Brain tumor detection, MRI, Deep learning, EfficientNet-B4, Convolutional Neural Network, Medical imaging1.