Abstract: Brain tumor detection using magnetic resonance imaging (MRI) plays a critical role in early diagnosis and treatment planning. However, manual analysis of MRI images can be time-consuming and prone to human error. This study proposes a deep learning framework for brain MRI classification that integrates Contrast Limited Adaptive Histogram Equalization (CLAHE) preprocessing with a Multi-Stage Transfer Learning (MSTL) strategy. The proposed framework evaluates three convolutional neural network architectures, MobileNetV2, ResNet50, and EfficientNet-B0, to analyze their performance in classifying 3D-rendered brain MRI images into tumor categories. CLAHE was applied to enhance image contrast and improve the visibility of structural patterns before training. The MSTL framework progressively fine-tunes pretrained models through multiple stages, enabling better adaptation of learned features to the MRI dataset. Experimental results demonstrate that all three models achieved high classification performance. Among the evaluated architectures, ResNet50 achieved the highest accuracy of 99.06%, followed by EfficientNet-B0 at 98.90% and MobileNetV2 at 98.12%. Training curves and confusion matrix analysis further confirmed stable convergence and strong classification capability across the models. The novelty of this study lies in combining CLAHE-based MRI enhancement with a progressive transfer learning framework to improve deep learning performance in medical image classification. The proposed approach may support AI-assisted diagnostic systems for automated brain tumor detection and improve the efficiency of clinical decision-making processes.