Publikace na 1. lékařská fakulta |
2019
Abstrakt
In this paper, we describe a fully automatic CAD system for spine detection in CT data followed by vertebra identification and segmentation. There are several basic problems: spine detection including the determination of spinal axis in spinal CT data, a localisation of individual vertebrae and identification of their types (order in spine) in case of incomplete scans of spine and also the final vertebra segmentation.
By a subjective strict expert validation, the algorithm provides 82.6% of fully correct vertebra segmentations. Based on that, it seems to be routinely usable and fully applicable in preparation for the following automatic spine bone lesion analysis.