Brain Tumor Segmentation of MRI images with U-Net and DeepLabV3+

In recent years, notable advancements have been made in medical imaging technology, with Magnetic Resonance Imaging (MRI) assuming a pivotal role in the diagnosis of brain tumors. Despite these advancements, medical image segmentation continues to pose a formidable challenge, as highlighted by various factors documented in existing literature. This study delves into the cutting-edge developments in Deep Learning for semantic segmentation, specifically concentrating on the precise identification of brain tumor pixels in 2D images. Employing U-Net and DeepLabV3+ architectures, the research provides experimental evidence that underscores the unparalleled performance of DeepLabV3+ with the Binary Cross Entropy loss function, offering valuable insights for enhancing the accuracy of brain tumor segmentation in medical imaging.