Original Research (Published On: 06-Aug-2026 )
DOI : https://doi.org/10.54364/AAIML.2026.64326Milia Habib, Azzam El Haffar, Abdel Rahman Hamzeh, Rabih Rammal, Zaher Merhi and Tony Karam
Adv. Artif. Intell. Mach. Learn., 6 (4):5891-5903
1. Milia Habib: Department of Computer & Communications Engineering, Lebanese International University
2. Azzam El Haffar: Department of Computer & Communications Engineering Lebanese International University Beirut, Lebanon
3. Abdel Rahman Hamzeh: Department of Computer & Communications Engineering Lebanese International University Beirut, Lebanon
4. Rabih Rammal: Department of Electrical Engineering Lebanese International University Beirut, Lebanon
5. Zaher Merhi: Department of Computer & Communications Engineering Lebanese International University Beirut, Lebanon
6. Tony Karam: Department of Electrical Engineering Lebanese International University Beirut, Lebanon
DOI: 10.54364/AAIML.2026.64326
Article History: Received on: 05-May-26, Accepted on: 31-Jul-26, Published on: 06-Aug-26
Corresponding Author: Milia Habib
Email: milia.habib@liu.edu.lb
Citation: Azzam El Haffar, et al. Dual-Task ResUNet++ with Genetic Algorithm Hyperparameter Optimization for Brain Tumor Segmentation and Classification Advances in Artificial Intelligence and Machine Learning. 2026;6(4):326. https://dx.doi.org/10.54364/AAIML.2026.64326
Abstract
Brain tumor segmentation and classification from Magnetic Resonance Imaging (MRI) scans remain a challenging task in clinical procedures that require high accuracy and efficiency. This study proposed a dual-task ResUNet++ architecture that conducts tumor segmentation and multi-class classification in a single forward pass. To automate hyper-parameter selection, a Genetic Algorithm (GA) was used to optimize and increase segmentation and classification performance. The model was evaluated on a brain MRI dataset comprising four categories: glioma, meningioma, pituitary tumor, and no tumor. The proposed technique achieved a Dice score of 88.03% and an IoU score of 79.61% on the segmentation task, while maintaining a classification accuracy of 98.80% and an F1-score of 98.70%. These results confirm that integrating multi-task models with automated hyperparameter optimization leads to better performance in both segmentation and classification.
Statistics
Article Views: 392
PDF Downloads: 8