AI auto-contouring in radiotherapy

Radiotherapy treatment planning requires accurate identification of the target volume and organs at risk (OARs), a task traditionally performed by experienced radiation oncologists. However, manual contouring of OARs is time-consuming and subject to inter-observer variability¹. In recent years, the development of deep learning models, particularly convolutional neural networks (CNNs) and U-Net architectures, has driven the introduction of artificial intelligence (AI)-based auto-contouring systems. 

Clinical studies have shown that auto-contouring can significantly reduce the time required for OAR delineation. Urago et al.² demonstrated a significant reduction in delineation time with auto-contouring, requiring six minutes compared with the three hours needed for manual delineation in head and neck cancer cases. The review by Zafar et al.³reported that auto-segmentation can reduce delineation time by up to 75% compared with manual contouring. A survey of French radiation oncologists found that 35% of users achieved time...