UNet-VT: Integrating U-Net and Vision Transformers for Enhancing Brain Tumor Segmentation in MRI scans

The earlier detection of brain tumour with enhanced accuracy has revolutionized medical image analysis leading to efficient diagnosis and real time patient treatment. Several DL-based methods have been proposed when it comes to extracting important features from an MRI image. Hence, in the present study, the investigators develop a new architecture, termed as UNet-VT using Brain Tumor Image Segmentation Benchmark (BRATS) 2020 to segment the tumor more automatically. To achieve the exact type of segmentation for the comparative performance analysis, a few other recent models have also been trained. Hence, the findings offer enhanced clinical and research use by providing accurate tumor segmentation. This makes it easier to plan for each patient since it assures that the course of treatment can be altered depending on features of each tumor. All in all, UNet-VT can be considered one of the best models for brain tumor segmentation with the highest scores of Dice coefficient 80.45% and mean IOU 89.5%.Thus, the higher segmentation accuracy and reliability of the proposed method can facilitate the development of the custom treatment plans and thereby constitutes a valuable tool for researchers and clinicians.