Pixel-level annotations of context region for trustworthy diagnosis of Breast Lymph Node Metastasis from Histopathological WSI
Year: 2025
Type: data-paper
Abstract: There is a dearth of annotated images in digital pathology, and annotations are pivotal for supervised automated diagnosis. This work aims to create a set of data on breast lymph node metastatic whole slide images (WSI) that is truly valuable as it is annotated by the domain experts in a very precise and intensive manner. The annotations provided here are of rare kind that has pixel-level labels divided into three categories; first is based on the diagnostically significant tissue regions, second is background tissue and the third is artefacts present in breast lymph node metastatic WSI. High level annotations, as is provided in this work, for areas apart from the metastatic region is crucial for diagnosis as this is the context region which is critical for cancer diagnosis. Breast lymph node metastasis is a severe medical condition that requires significant efforts by pathologists to examine cancer under microscope using glass slides. Automated cancer diagnosis using WSI, the digitized versions of glass slides, needs to be trustworthy and responsible, which is ensured by involving domain experts in the process that begins with identifying diagnostically significant regions on a WSI. In this dataset, a two-stage high-value pixel-level annotation protocol is designed where selected 73 training images of Camelyon16 dataset are annotated by doctors and verified by pathologists for their accuracy. These annotations would help researchers to reliably prepare patches for further processing that can be used as training samples for extracting more such significant regions for data preparation from WSI for diagnosing metastasis. These reliable and trustworthy annotations would help to take it to the clinic from the research lab much quickly.
Source: Data in Brief
Language: English
Topic: AI in cancer detection
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,220
OpenAlex ID: W7116830963