[2508.17128] CE-RS-SBCIT A Novel Channel Enhanced Hybrid CNN Transformer with Residual, Spatial, and Boundary-Aware Learning for Brain Tumor MRI Analysis

View the PDF file for the paper entitled CE-SS-SBCIT

PDF view

a summary:Brain tumors remain among the most deadly human diseases, as early detection and accurate classification are very important to effective diagnosis and treatment for treatment. Although learning -based learning -backed diagnostic systems showed remarkable progress. However, traditional tagine nervous networks (CNNS) and transformers face continuous challenges, including high calculations, sensitivity to simple contrast changes, structural inceptions, and texture contradictions in MRI data. Therefore, a new hybrid frame, CE-RS-SBCIT, is presented, the remaining CNNS and the learning-based spaqu with transformer units. The proposal framework for local and international granules with four basic innovations takes advantage of the CNN (SBCIT), (2) remaining the remaining and spacious CNNS, (III), the CEO enhancement strategy (CE), and (IV) a new spatial mechanism. Advanced SBCIT employed STEM’s destruction and contextual reaction adapter with homogeneity and systematic bodies, allowing modeling effective global features. Moreover, the remaining CNNS and spatial CNNS, which are strengthened by the additional features that were transferred, raises the acting space, while the CE unit enlarges discriminatory channels and relieves repetition. Moreover, the spatial attention mechanism selectively emphasizes hidden contrast and textual changes across the tumor. An intense assessment of the Kaggle MRI and Figshare databases, includes a diet, mening tumor, pituitary tumors, health controls, superior performance, 98.30 % accuracy, 98.08 % allergies, 98.25 % F1, and 98.43 %.