Abstract
Brain tumor (BT) detection from magnetic resonance imaging (MRI) images plays a critical role in early diagnosis and effective treatment planning. This research proposes a novel deep learning framework that integrates a VGG16-based Convolutional Autoencoder (CAE) with a Bi-Directional Long Short-Term Memory (BiLSTM) network for accurate and robust BT classification. The VGG16-CAE is employed for hierarchical spatial feature extraction, capturing both low-level and high-level representations of MRI images, while the BiLSTM module models long-range dependencies within the extracted features, enhancing classification performance. The model was evaluated using the publicly available Kaggle Brain MRI Dataset with systematic hyperparameter tuning and fivefold cross-validation to ensure reproducibility and generalization. The proposed framework achieved superior performance, with an average accuracy of 96.8%, precision of 97.1%, recall of 96.5%, F1-score of 96.8%, and an ROC-AUC of 0.985, with narrow confidence intervals (CIs) indicating statistical robustness. Comparative analysis with baseline convolutional neural network (CNN) architectures (Standard VGG16, ResNet50, DenseNet121, and VGG16-CAE without BiLSTM) and recent literature demonstrated that the proposed method significantly outperforms existing approaches by effectively capturing sequential dependencies in MRI data. The modular design of the framework allows for future extensions, including tumor localization and segmentation tasks, making it a promising solution for clinical applications. This work highlights the advantage of combining hierarchical feature learning with sequential modeling for precise BT detection, contributing to improved decision support in neuro-oncology.