Intelligent Brain Tumor Detection using Fine-Tuned EfficientNets

Improving patient prognosis and treatment success is contingent upon timely and precise detection of brain tumors. Because Magnetic Resonance Imaging (MRI) can create highly contrasted images of brain regions, it is a primary diagnostic technique. Recently, advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have significantly enhanced the capability to interpret these medical images. This paper proposes a novel approach employing a fine-tuning methodology with EfficientNets, specifically EfficientNetB3, for the classification of MRI brain images into four tumor categories: glioma, meningioma, pituitary, and no tumor. The EfficientNetB3 model was selected for its balance between efficiency and accuracy in computational performance. Comparative analysis was conducted between the fine-tuned EfficientNetB3 and other pre-trained CNN models to evaluate its effectiveness in tumor detection. Superior performance was shown by the proposed EfficientNetB3-based brain tumor detector, which achieved an accuracy of 99.34%, F1-score of 99%, recall of 99%, and precision of 99%.