Abstract
Radiation therapy remains a cornerstone in the treatment of cancer, with the delineation of Organs at Risk (OARs), tumors, and malignant lymph nodes playing a critical role in the planning process. However, the manual segmentation of these anatomical structures is both time-consuming and costly, with inter-observer and intra-observer variability often leading to delineation errors. In recent years, deep learning-based automatic segmentation has gained increasing attention, leading to a proliferation of scholarly works on OAR and tumor segmentation algorithms utilizing deep learning techniques. Nevertheless, similar comprehensive reviews focusing solely on malignant lymph nodes are scarce. This paper provides an in-depth review of the advancements in deep learning for malignant lymph node segmentation and detection. After a brief overview of deep learning methodologies, the review examines specific models and their outcomes for malignant lymph node segmentation and detection across five clinical sites: head and neck, upper extremity, chest, abdomen, and pelvis. The discussion section extensively covers the current challenges and future trends in this field, analyzing how they might impact clinical applications. This review aims to bridge the gap in literature by providing a focused overview on deep learning applications in the context of malignant lymph node challenges, offering insights into their potential to enhance the precision and efficiency of cancer treatment planning.