Hybrid Deep Learning and Machine Learning for Brain Tumor Detection: CNN Meets Random Forest

Medical imaging diagnosis heavily depends on precise and explainable AI models because brain tumor classification stands as an essential clinical task in the field. The research implements a deep learning framework which unifies Convolutional Neural Networks (CNNs) for feature extraction with Random Forest (RF) classification and Local Interpretable Model-Agnostic Explanations (LIME) interpretability. The proposed model received training and evaluation on 4,220 MRI images from the Kaggle dataset which displayed Glioma and Meningioma and Pituitary tumor contents. The proposed CNN-RF model reached 98.88% accuracy which surpassed traditional CNN network architectures. The LIME heatmaps validated how the model applied its decisions through meaningful clinical features which improved both transparency and reliability of its decisions. ROC curve analysis reported that the method achieved an AUC value of 0.988 which exceeded standard deep learning technology performance levels. A review of all tumor types showed that the confusion matrix contained only minor errors to validate broad usage across each type. The examination shows that the CNN-RF hybrid model presents itself as a dependable alternative that provides interpretability over black-box AI systems. This study provides essential contributions to AI trustworthiness development in healthcare through its accurate and explainable diagnostic framework that will help medical practice adopt AI-driven diagnostics at a higher level.