Silver Corpus Label Fusion Framework

The VISCERAL silver corpus merging framework is available on GitHub. It provides the functionality to generate silver corpus annotations of anatomical structures in medical images by fusing data that is typically available from challenges or benchmarks. This is (1) a small set of manually performed expert annotations of high quality (gold corpus) and (2) large sets of possibly less accurate segmentation estimates obtained from algorithms participanting in a benchmark.

Label fusion approaches are used to fuse both segmentation sources to compute segmentation estimates of higher quality (than relying on one of both sources) in a large set of medical images resulting in a so called silver corpus.