Abstract
Recent advances in transformer-based deep learning architectures have significantly enhanced medical image segmentation by enabling precise delineation of anatomical structures across various imaging modalities and clinical contexts. Transformers leverage self-attention mechanisms to model long-range dependencies and global context, overcoming the limitations of convolutional neural networks with localized receptive fields. This chapter reviews the application of transformers in segmentation tasks such as brain tumors, retinal vessels, and skin lesions, highlighting innovations in architecture, performance benchmarks, and evaluation metrics. Comparative analyses reveal trends in accuracy, robustness, and efficiency, while addressing challenges like computational demands, limited annotated data, and clinical interpretability. Future directions include lightweight architectures, multimodal integration, and clinical validation. These developments illustrate the transformative role of transformers in precision medicine and AI-driven healthcare.