Other Open Access
Anjany Sekuboyina; Bjoern Menze; Jan Kirschke
This is the challenge design document for the "Large Scale Vertebrae Segmentation Challenge", accepted for MICCAI 2020.
A primary step in automated quantification of spinal morphology and pathology is vertebral labelling and segmentation. Aimed at these tasks, the first iteration of the 'Large Scale Vertebrae Segmentation Challenge' (VerSe'19) was held at MICCAI 2019 and received considerable participation from the community (>250 registrations and data downloads, 20 participating teams). With its first iteration, VerSe addressed a severe shortage of publicly-available, large, accurately annotated CT spine data in the community by releasing 160 CT scans and their voxel-level annotations comprised of a large variety in fields of view, spatial resolutions, spinal and vertebral pathologies, collected over several scanners from two major vendors.
Building on the data, experience, and learning from VerSe'19, we propose to organise a second iteration for of the Large Scale Vertebrae Segmentation Challenge (VerSe'20) at MICCAI 2020. With VerSe'20, we aim to work with 300 CT scans (~100% increase over its previous iteration). While retaining the richness of its predecessor, the data will now be multi-centre with five different institutions and and all four major scanner manufacturers. Additionally, challenging the learning algorithms, focus is given to include atypical anatomies such as transitional vertebrae and additional vertebrae such as L6. With clinical-translation being the primary objective of data preparation, we believe the algorithms using this data would be more robust and generalisable.