Glioma Detection Using Deep and Transfer Learning

Glioma is one of the most aggressive and dangerous types of brain tumors, making early and accurate diagnosis essential for better treatment outcomes. Traditionally, diagnosing gliomas depends on expert interpretation of MRI scans, which can be both time-consuming and subjective. In recent years, deep learning has proven to be a powerful tool for medical image analysis, offering faster and more accurate results. This study introduces a deep learning-based system to detect, segment, and classify gliomas using various types of MRI images. We worked with a large dataset of 33,400 MRI images and tested advanced models: EfficientNetB2 and MaxViT. These models were trained to distinguish between different glioma grades, helping to identify whether a tumor is low-grade or high-grade. Among the models, EfficientNetB2 performed better than MaxViT, showing higher accuracy and consistency. The results suggest that combining modern deep learning techniques with MRI analysis can lead to more reliable and scalable tools to support doctors in diagnosing brain tumors. Future improvements will focus on including patient-specific information and making the models more adaptable for real-world clinical use. Index Terms: Glioma Detection, Brain Tumor Classification, Medical Imaging, Magnetic Resonance Imaging (MRI), Deep Learning, Transfer Learning, EfficientNetB2, MaxViT (Vision Transformer)