Brain Tumor Classification Utilizing Deep Learning with Long Short-Term Memory Techniques via Magnetic Resonance Images

Abd Elnabi, Amira Mahmoud; Torkey, Hanaa; El-Bahnasawy, Nirmeen A. Wahab

doi:10.21608/ijt.2025.363661.1086

Brain Tumor Classification Utilizing Deep Learning with Long Short-Term Memory Techniques via Magnetic Resonance Images

Document Type : Original Article

Authors

¹ Master student, Communication and Electronics Department, Delta Higher Institute of Engineering and Technology, 32952, Mansoura, Egypt

² Associate Professor at Computer Science and Engineering Department, Faculty of Electronic Engineering, Menoufia University, 32952, Menouf, Egypt

10.21608/ijt.2025.363661.1086

Abstract

Detection of the brain tumor is a major challenge in the field of medical imaging. Manually recognizing brain tumors constitutes a costly and arduous process for radiologists; hence it is necessary to implement a computerized technique. However, even experienced medical radiologists face difficulties in accurately and reliably analyzing MRI (Magnetic Resonance Imaging) images to diagnose brain tumors. Because precision matters most for classification, computer vision scientists have established a variety of techniques, however they still have trouble with low fidelity. The suggested technique can be separated into two distinct steps: preprocessing and classification. Initially, the data are pre-processed via two main steps: image enhancement, noise removal, and data augmentation. Next, images that have been processed are loaded to the ResNet50 model for extraction of features. Secondly, all of the extracted features are trained and categorized into four classes using a LSTM (Long short term memory) multilayer with dropout between its layers. The experiment’s results exhibits that the proposed technique scored an accuracy of 99.26%.

Keywords