VerSe: A vertebrae labelling and segmentation benchmark for multi-detector CT images
Reliable and accurate automated processing of spine images is expected to benefit clinical decisions on diagnosis and surgery planning, as well as population-based analysis of spine and bone health. The 2019 and 2020 Large Scale Vertebrae Segmentation Challenge (VerSe), organised in conjunction with the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), therefore called for algorithms tackling the labelling and segmentation of vertebrae. This work presents the benchmarking of 25 algorithms on two datasets containing 374 multi-detector CT scans from 355 patients, from which 4505 vertebrae were individually annotated at voxel level by a human-machine hybrid algorithm. Performance variation was investigated at the vertebra- and scan level, as well as for different fields of view. Furthermore, the generalisability of approaches was assessed by evaluating the top-performing algorithms of one challenge iteration on data from the other iteration.
See Sekuboyina et al., Medical Image Analysis
Access the VerSe content and code via: https://github.com/anjany/verse.