An improved deep learning-based hybrid model with ensemble techniques for brain tumor detection from MRI image

Abstract

Recent developments in artificial intelligence and medical imaging technologies have significantly improved disease analysis and prediction, especially regarding identifying brain tumors (BTs). With the advancement of medical imaging technology, 3D brain scans can now be captured using a variety of modalities, offering a comprehensive perspective for tumor diagnosis. A key stage in the diagnosis tactic is extracting pertinent features from magnetic resonance imaging (MRI) scans, and several researchers have suggested various methods. This work aims to develop an accurate BTs detection and classification system using machine and deep learning techniques. The system is designed to classify BTs and healthy data using three merged datasets. Deep learning algorithms, including 2D convolutional neural network (CNN) with 13 layers, CNN long short-term memory (LSTM), and another 2D CNN with nine layers, are employed to improve the classification performance of the system. All three deep learning models achieve high accuracy, with 2D CNN LSTM yielding the highest accuracy of 98.47%, followed by another 2D CNN at 97.71% and 2D CNN at 92.36%. These models are then combined using ensemble learning to make a hybrid network, which further improves the system’s accuracy. The comparative analysis demonstrates that the ensemble deep learning models outperform all other classifiers, achieving an accuracy of 98.82%, precision of 99%, and recall of 99%. The work findings indicate that the developed brain tumor detection and classification system, which combines deep learning techniques, offers high accuracy and precision, making it a promising tool for accurately diagnosing brain tumors.

Keywords

• 2D convolutional neural network
• Long Short-Term Memory (LSTM)
• Ensemble learning
• Brain tumor classification