Reference skin lesion images suitable for inclusion in the Public Skin Lesion Image Archive are being annotated and marked up using standards growing out of the work of the ISIC project teams.
The National Cancer Institute (NCI) has created a free and open source National Biomedical Imaging Archive (NBIA) that serves as a model for the public archive of skin lesion images. Many existing sets of skin lesion images have been thoroughly annotated in a consensus process involving multiple participants. These sets have been created for varied academic and commercial research projects. Contributions of such data sets from multiple academic and commercial entities provide the backbone of the database. These images are archived and annotated using web-based open source software predicated on lessons from the NBIA initiative, the Radiological Society of North American (RSNA) Medical Imaging Resource Center (MIRC), and the AIM standard developed by NCI. The software is being developed by a team led by David Gutman, MD, PhD of Emory University. The director of the archive is Rainer Hoffman-Wellenhof.
The creation of an archive of skin lesion images in the public domain benefits from collaborations with Kitware and IBM. Kitware, Inc. is a leader in the creation and support of open-source software and state of the art technology. By fostering extended, collaborative communities, Kitware provides flexible, cost-effective visualization, computer vision, medical imaging, and quality software process solutions to a variety of academic and government institutions worldwide. IBM, an international leader in computer technology, is developing a set of image analysis tools (e.g., lesion segmentation) and semi-automated annotation/mark-up tools. These tools will not only be useful in the acquisition and annotation of images for the public image archive, but the tools will serve as benchmarks for others interested in developing image analysis programs for skin lesions.
Crowd sourcing techniques supervised by pigmented lesion experts will be used to refine the semi-automated annotation and mark-up tools to be developed by IBM and to provide preliminary annotation and mark up of the numerous images to be gathered into the archive. Strict standards will be established for inclusion of images based on image quality and the validity of the associated clinical and pathology diagnoses.