Affiliations: [a] Department of Computer Science and Engineering, PPG Institute of Technology, Coimbatore, India | [b] Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, India | [c] Department of Computer Science, Nawroz University, Duhok, Kurdistan Region, Iraq | [d] Department of Artificial Intelligence and Data Science, Tagore Institute of Engineering and Technology, Salem, India | [e] Department of Artificial Intelligence and Machine Learning, Tagore Institute of Engineering and Technology, Salem, India
Correspondence:
[*]
Corresponding author: ThamilSelvi C P, Department of Computer Science and Engineering, PPG Institute of Technology, Coimbatore, India. E-mails: [email protected], [email protected].
Abstract: Technological developments in medical image processing have created a state-of-the-art framework for accurately identifying and classifying brain tumors. To improve the accuracy of brain tumor segmentation, this study introduced VisioFlow FusionNet, a robust neural network architecture that combines the best features of DeepVisioSeg and SegFlowNet. The proposed system uses deep learning to identify the cancer site from medical images and provides doctors with valuable information for diagnosis and treatment planning. This combination provides a synergistic effect that improves segmentation performance and addresses challenges encountered across various tumor shapes and sizes. In parallel, robust brain tumor classification is achieved using NeuraClassNet, a classification component optimized with a dedicated catfish optimizer. NeuraClassNet’s convergence and generalization capabilities are powered by the Cat Fish optimizer, which draws inspiration from the adaptive properties of aquatic predators. By complementing a comprehensive diagnostic pipeline, this classification module helps clinicians accurately classify brain tumors based on various morphological and histological features. The proposed framework outperforms current approaches regarding segmentation accuracy (99.2%) and loss (2%) without overfitting.
Keywords: VisioFlow FusionNet, brain tumor segmentation, NeuraClassNet, cat fish optimizer, medical image analysis, deep learning
