Multi-objective Archimedes Optimization Algorithm with Fusion-based Deep Learning model for brain tumor diagnosis and classification

• Published in: Multimedia tools and applications Vol. 82; no. 11; pp. 16985 - 17007
• Main Authors: Devanathan, B., Kamarasan, M.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.05.2023; Springer Nature B.V
• Subjects: Abnormalities; Algorithms; Brain; Brain cancer; Classification; Computer Communication Networks; Computer Science; Computer vision; Data Structures and Information Theory; Deep learning; Diagnosis; Entropy (Information theory); Feature extraction; Image contrast; Image enhancement; Image processing; Image segmentation; Machine learning; Magnetic resonance imaging; Medical imaging; Multimedia Information Systems; Multiple objective analysis; Optimization; Optimization algorithms; Special Purpose and Application-Based Systems; Tumors; Brain tumor; Deep learning; Magnetic resonance imaging; Multiobjective optimization; Fusion models; Computer aided diagnosis; Image classification
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-14164-5

Earlier identification of brain tumors (BT) is essential to increase the survival rate. Magnetic Resonance Imaging (MRI) is a commonly employed method that records brain abnormalities by the use of several modalities for clinical study. The recently developed computer vision and image processing schemes can be used for the detection and localization of tumor regions in the brain, which can be utilized for further treatment. In this regard, the study presents a novel Multiobjective Archimedes Optimization Algorithm with Fusion based Deep Learning (MOAOA-FDL) technique for brain tumor diagnosis and classification. In addition, the MOAOA-FDL technique preprocesses the MRI images via contrast enhancement and skull stripping. Moreover, AOA with Shannon entropy based multi-level thresholding approach is developed for medical image segmentation. Furthermore, the fusion of two deep learning models namely MobileNet and EfficientNet models is employed for feature extraction process. Finally, the AOA with long short term memory (LSTM) method is applied for classification model and thus allocates proper class label to it. The AOA is used to properly choose the hyper parameters like batch size, learning rate, and epoch count. The design of fusion process and MOAOA for BT diagnosis demonstrates the innovation of this study. For showcasing the better performance of the MOAOA-FDL method, a series of simulations have been executed utilizing benchmark dataset. The experimental outcome shows that the MOAOA-FDL method has outperformed the other recent approaches in terms of several performance measures.