<p>Brain cancers are the second most prevalent cause of tumor development, posing significant challenges to healthcare systems worldwide. Enhancing patient outcomes and creating efficient treatment regimens depend heavily on the early and precise analysis of brain tumors. Conventional diagnostic methods, including manually reviewing Magnetic Resonance Imaging (MRI) scans, obtain a lot of time, are prone to human error, and may not be able to differentiate between subtle tumor kinds. to overcome these limitation, the hybrid deep learning strategy that combines dense layers for classification with Long Short-Term Memory (LSTM) layers for binary classification of X-ray images into tumor and normal classes. By using Sobel detection of edges and Local Binary Patterns (LBP) to extract features from X-ray images, the model is improve able to identify complex patterns that may be signs of brain malignancies. The X-ray-based tumor versus normal classification was used to train and evaluate the suggested hybrid deep learning model. The dataset, which included annotated X-ray pictures, was acquired via Kaggle for this investigation. To ensure uniform input to the network, images were preprocessed by scaling to 224 × 224 pixels, turning to grayscale, and normalized pixel values. To prevent over fitting and ensure robust performance, we introduce dropout and pooling layers. Extensive training and validation of our model on a divided dataset resulted in a test accuracy of 99.50% and a test loss of 0.0279, outperforming existing methods in terms of precision (98.91%), accuracy (99.50%), f1-score (99.45%), and specificity (99.08%). This research shows that using cutting-edge computational methods, Deep learning in particular is both practical and advantageous for automatic and accurate brain tumor identification and categorization.</p>