ANRAN: AN ADVANCED NEURAL RADIOMICS ATTENTION NETWORK FOR EARLY DETECTION OF ASYMPTOMATIC BRAIN TUMORS

Early detection of asymptomatic brain tumors is crucial for improving patient survival rates; however, conventional magnetic resonance imaging (MRI) analysis often fails to identify subtle abnormalities at early stages. This study presents an Artificial Intelligence-driven radiomics framework for automated and non-invasive detection of early brain tumors. Radiomic features-including statistical, shape, and texture descriptors-were extracted from MRI datasets such as BraTS and analyzed using deep learning architectures to identify subtle tumor-related biomarkers that are invisible to the human eye. The results demonstrate that integrating radiomic feature engineering with deep neural models significantly enhances sensitivity and specificity in early tumor detection. Key radiomic features such as entropy, contrast, and homogeneity showed strong discriminative power between normal and early pathological tissues. Moreover, Vision Transformer-based hybrid architectures and self-supervised learning approaches improved the model's generalizability across heterogeneous MRI acquisition protocols. Overall, this research highlights the transformative potential of Artificial Intelligence in neuro-oncology, providing a foundation for multimodal and radiogenomic integration and supporting early, personalized, and non-invasive brain tumor diagnosis toward the advancement of precision neuro-oncology.

کلیدواژه ها:

Brain Tumor ، Radiomics ، Deep Learning ، Brain Tumor Detection ، Vision Transformers ، Early Diagnosis