Brain Tumor Classification in MRI Using Hybrid ASA-Based Deep Learning and Masi-Entropy Multilayer Thresholding Segmentation with Sunflower Optimization

• Published in: Traitement du signal Vol. 42; no. 1; pp. 223 - 241
• Main Authors: Balasuburamani, Kannan, Shanmugavel, Karthigai Lakshmi
• Format: Journal Article
• Language: English; French
• Published: Edmonton International Information and Engineering Technology Association (IIETA) 01.02.2025
• Subjects: Accuracy; Algorithms; Brain; Brain cancer; Brain research; Classification; Datasets; Deep learning; Entropy; Feature extraction; Glioma; Machine learning; Magnetic resonance imaging; Medical imaging; Methods; Multilayers; Neuroimaging; Optimization; Optimization techniques; Search algorithms; Sunflowers; Tomography; Tumors
• ISSN: 0765-0019; 1958-5608
• DOI: 10.18280/ts.420120

Accurately diagnosing brain tumors at an early stage is critical for successful therapy and saves the lives of many people worldwide. Magnetic resonance imaging (MRI) scans are frequently employed for tumor detection because of their noninvasive nature, sparing patients with the discomfort of undergoing a biopsy. The process of identifying tumors is arduous and time-consuming because of the extensive array of three-dimensional (3D) images generated by an MRI scan of a patient's brain from various angles. Moreover, the diverse sizes, positions, and shapes of brain tumors pose challenges for their identification and classification. Consequently, computer-aided diagnostic (CAD) systems have been suggested as solutions for detecting brain tumors. This paper presents a new unified deep learning (DL) model called enhanced AlexNet for brain tumor detection and classification. Initially, a fast nonlocal means (FNLM) filter was used for preprocessing. The tumor nodule was segmented from the MR images using Masi-entropy-based multilevel thresholding with the sunflower optimization algorithm (MasiEMT-SFO). BMC-19 was used to extract various features in the feature extraction process. The extracted features were then classified using a hybrid classifier called the enhanced AlexNet classifier algorithm with the Anopheles search algorithm (ASA). The proposed hybrid classifier accurately detected brain tumors. The proposed model was implemented in MATLAB using two Kaggle datasets. The experimental results show that the proposed enhanced AlexNet algorithm outperforms the existing methods, providing compelling evidence for its application in other diseases. The proposed model outperformed existing methods in distinguishing abnormal and healthy brain tissue from MRI images, with an F1-score of 97.21%, specificity of 98.76%, precision of 97.61%, sensitivity of 96.73%, and accuracy of 99.76%. These findings confirm the efficacy of the proposed approach.