Automated diagnosis of brain tumors from MRI scans using U-Net segmentation

Abstract

This work uses U-Net delineation on MRI scans to demonstrate a reliable automated approach for brain tumor diagnosis. A diversified dataset was collected for machine validation and training using a deductive methodology and an interpretive philosophy. The coefficients of Dice and sensitivity values for the U-Net topology were both 0.92 and 91%, respectively, demonstrating its remarkable performance. The system's correctness and dependability were proven by comparison analysis with conventional ground truth segments. The algorithm also shows proficiency in grading and categorizing tumor subtypes, which has important implications for individualized treatment plans. Segmentation time was 60% shorter after integration into clinical procedures, increasing workflow effectiveness. These results highlight the system's capacity to transform brain tumor diagnostics by providing neuro-oncology practitioners with a dependable and time-saving method. Future studies should concentrate on improving the system's flexibility to various imaging techniques and investigating the integration of numerous modalities for thorough tumor characterization.