Unified approach for accurate brain tumor Multi-Classification and segmentation through fusion of advanced methodologies

• Published in: Biomedical signal processing and control 2025-02, Vol.100, p.106872, Article 106872
• Main Authors: Karthik, A., Kumar Sahoo, Santosh, Kumar, Ajay, Patel, Nileshkumar, Chinnaraj, P., Phaneendra Maguluri, Lakshmana, shuaib, Mohammed, Rajaram, A.
• Format: Article
• Language: English
• Subjects: Attention-Augmented Convolutional Neural Network (CNN); Brain tumor; Magnetic Resonance Imaging (MRI); Random Forest; U-Net
• ISSN: 1746-8094
• DOI: 10.1016/j.bspc.2024.106872

•The process involves analyzing MRI data to categorize brain tumors based on various characteristics. The proposed model integrates an Attention-Augmented Convolutional Neural Network (CNN), Random Forest (RF), and U-Net to leverage the strengths of attention mechanisms, ensemble learning, and semantic segmentation. The Attention-Augmented CNN captures intricate features with attention focus, the RF refines classification decisions through ensemble learning, and the U-Net ensures precise tumor segmentation.•The fusion of these components results in a comprehensive framework that not only achieves high accuracy in classification but also provides detailed and accurate tumor segmentation in MRI brain images. Experimental results on a diverse dataset of medical MRI brain images demonstrate the effectiveness of the proposed unified model, showcasing its potential for advancing brain tumor analysis in medical imaging.•The fusion of Attention-Augmented CNN and Random Forest contributed to a robust ensemble-based classification, while U-Net enhanced segmentation precision. The performance metrics like recall, precision, accuracy, loss, and F1-score are verified in Python. This research introduces a unified approach for accurate brain tumor Multi-classification and segmentation through the fusion of diverse yet complementary methodologies. Brain tumor classification using magnetic resonance imaging (MRI) scans is crucial in non-invasive assessment of brain tumors, providing anatomical information for accurate classification. Classifying brain tumors involves several challenges, inherent complexity of tumor heterogeneity, imaging protocols variations, and the need for distinguishing between benign and malignant lesions. Traditional methods often face limitations in handling the intricate patterns present in medical images, prompting the adoption of sophisticated computational approaches. The process involves analyzing MRI data to categorize brain tumors based on various characteristics. The proposed model integrates an Attention-Augmented Convolutional Neural Network (CNN), Random Forest (RF), and U-Net to leverage the strengths of attention mechanisms, ensemble learning, and semantic segmentation. The Attention-Augmented CNN captures intricate features with attention focus, the RF refines classification decisions through ensemble learning, and the U-Net ensures precise tumor segmentation. The fusion of these components results in a comprehensive framework that not