Flattening Yields Faster CT
2014 R&D 100 Winner
In 2012, more than 85 million computed tomography (CT) scans were performed in the U.S. Of these, 16% were thoracic scans. Up to now, this has been done manually and sequentially in what is a tedious, lengthy and error-prone process. Engineers at Siemens Corporate Technology and Siemens Healthcare, Computed Tomography have launched a new solution to save radiologists time and increase diagnostic confidence for thoracic bone assessment. The syngo.CT Bone Reading technology automatically extracts rib centerlines and virtually unfolds the rib cage and spine into a single image. Additionally, all ribs and vertebras are automatically labeled.
Machine learning and computer vision theories underpin the action of syngo.CT. Siemens’ technology acquires voxels with a new approach that uses a probabilistic boosting tree of CT image data sets that represent all known bone pathologies. This allows the instrument to automatically generate curved planar reformats (CPRs) of the ribs. The algorithm unfolds the rib cage, flattening the bones onto a plane that is much easier to examine than prior methods, which looked at rib pairs in slices on the head-to-toe axis. External clinical evaluation has shown that syngo.CT Bone Reading can shorten radiologists’ bone reading time by 50% and increase the sensitivity of rib fracture detection by 10%.
The syngo.CT Bone Reading Development Team
Grzegorz Soza, Principal Developer, Siemens Healthcare, Computed Tomography
Shaohua Kevin Zhou, Principal Developer, Siemens Corporate Technology
Atilla P. Kiraly
David Liu, Siemens Corporate Technology
Andreas Wimmer, Siemens Healthcare, Computed Tomography