Analysis of Pre-trained CNN Models in MRI-Based Brain Tumor Detection

Deep learning has become a cornerstone of modern healthcare technology. With its rapid advancements, deep learning has enabled the early prediction and diagnosis of various diseases, including critical conditions like brain tumors. Brain tumors are among the most severe health challenges, requiring accurate and timely detection for effective treatment. Traditionally, brain tumor diagnosis has heavily relied on MRI or CT imaging for straightforward classification. This study aims to enhance brain tumor prediction by employing deep learning techniques on CT or MRI images. This research provides a comprehensive comparative analysis between five pre-trained models—InceptionV3, ResNet-50, VGG-16, MobileNetV2, and DenseNet121—to classify brain tumors in both binary (tumor vs. no tumor) and multi-class settings (glioma, meningioma, no tumor, and pituitary). This study not only emphasizes the modes strengths and limitations but also their applicability to both binary and multi-class brain tumor classification. The findings pave the way for more accurate and efficient early detection and treatment planning, displaying the potential for these models to be incorporated into real-world medical applications.