Brain Tumor Detection Using Deep Learning: ResNet+LSTM

Brain tumors are among the most severe medical conditions, necessitating early and accurate detection to improve patient outcomes. This research paper explores a hybrid deep learning model that combines. Combining Long Short-Term Memory (LSTM) models with Residual Networks (ResNet-50) improves the ability to identify brain cancers from MRI data. The Brain MRI Images for Tumor Detection dataset, which is accessible on FigShare, contains annotated MRI scans of a variety of tumor forms, including pituitary tumors, meningiomas, and gliomas. This study makes use of this dataset. For the purpose of comprehending the evolution of brain tumors, the model uses ResNet-50 to extract spatial characteristics from individual MRI slices, while LSTM records temporal relationships across several slices. The dataset comprises images that were preprocessed by resizing to 128×128 pixels and normalizing pixel values for con-sistent input. The proposed ResNet-LSTM model classifies MRI slices into three categories: gliomas, meningiomas and pituitary. Performance metrics such as accuracy, precision, and confusion matrix were used to evaluate the model, demonstrating significant improvements over traditional methods. The findings suggest that the integration of ResNet and LSTM can effectively capture both spatial and sequential information, enhancing diagnostic accuracy and reliability. The potential of hybrid deep learning models to improve medical imaging diagnoses is highlighted in this study.