Transformers Effectiveness in Medical Image Segmentation: A Comparative Analysis of UNet-Based Architectures

Medical image segmentation is a crucial task in healthcare as it helps in the accurate diagnosis and treatment of various medical conditions. UNet-based architectures have been widely used for medical image segmentation due to their ability to produce high-quality segmentations. However, there is a need to improve the performance of these architectures to enhance their effectiveness in medical image segmentation further. One promising approach is using transformers, which have shown great potential in improving the performance of various deep learning models. This research compares four UNet-based architectures (UNet, UNetR, TransUNet, and Swin-UNet) with and without transformers to evaluate their effectiveness in medical imaging using four independent datasets. The findings of this study will be valuable in advancing the field of medical image segmentation and contributing to the optimization of Unet-based architectures.