A Novel Enhanced Cuckoo Search-SVM Model for Robust Brain Tumor Segmentation

• Published in: 2024 9th International Conference on Communication and Electronics Systems (ICCES) pp. 1 - 7
• Main Authors: J, Alphonsa, Kumari, V Sheeja, Roslin Dayana, K
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.12.2024
• Subjects: Accuracy; Automated Tumor Identification; Brain modeling; Brain Tumor Segmentation; Brain tumors; Classification algorithms; Computational modeling; Cuckoo Search Algorithm; Feature extraction; Feature Selection; Health care; Hybrid Model; Image segmentation; Machine learning algorithms; Medical Image Processing; MRI Images; Sensitivity; Support vector machines; SVM Classification
• DOI: 10.1109/ICCES63552.2024.10859815

The accurate segmentation of brain tumors from medical imaging is a critical task that aids in early diagnosis and treatment planning. Existing techniques often need help with issues like high computational demands, sensitivity to image noise, and suboptimal performance in varied imaging conditions. Enhanced Cuckoo Search-SVM (ECS-SVM) model that integrates an optimization-driven approach with machine learning for robust and efficient segmentation. We utilize the ECS algorithm to fine-tune the hyperparameters of a Support Vector Machine (SVM), which classifies extracted features from MRI images. The proposed approach leverages the search capabilities of the enhanced Cuckoo Search algorithm to identify optimal parameters that enhance the SVM's performance. With the valid application on a comprehensive dataset, the proposed method obtains a high accuracy of 95.3%. In comparative experiments, it is shown that ECS-SVM outperforms traditional and existing hybrid models, making it a great option for automated and reliable brain tumor segmentation. The results indicate significant improvements in segmentation precision, robustness, and computational efficiency, marking a notable advancement in the field.