SWAN - Enabling Fast and Mobile Histopathology Image Annotation through Swipeable Interfaces

The introduction of SWipeable ANnotations (SWAN), an open-source web application, aims to enhance the efficiency of histopathology image annotation. Traditional folder-based methods are often slow and cumbersome, hindering the development of deep learning models for tasks like mitotic figure classification. In a pilot study involving four pathologists who annotated 600 mitotic figure image patches, SWAN demonstrated a pairwise percent agreement of 86.52% to 93.68%, comparable to the 86.98% to 91.32% agreement of the traditional method. The study highlights SWAN's usability and its mobile-friendly interface, making it a promising tool for scalable and efficient image annotation in clinical settings.