At ImFusion, we were working on bone ultrasound data for an industry customer for many years. Some of the research work is published, other work is confidential.
Our original bone segmentation method was recently compared with other works, and fared comparable on a variety of different datasets, see details in the CAOS paper below. The full dataset and thorough comparison is currently under review.
Within the scope of the publicly funded FOMIPU project (BMBF grant number 13GW0293B), we were investigating 3D ultrasound scans for pediatric cases. A common challenge in developing algorithms in ambulatory clinical settings is the lack of tracking information of the bone: While in orthopedic surgery, all bones are typically equipped with reference trackers, this is unfeasible in screenings. We developed a powerful method to compensate for relative bone movements in pediatric cases during the acquisition, by iteratively registering sections together to achieve full bone surface coverage to measure clinically relevant angles.
Reconstruction workflow: A Starting from multiple 3D US recordings (exemplary recording in blue) on the thigh (1), a bone detection extracts a point cloud from each sweep (2). Location-dependent filtering for wrong bone detections from the top of the images (3). Poisson surface reconstruction of individual sweeps (4). Iterative alignment and successive mesh combination globally aligns the individual recordings (illustrated on the right) and reduces to one reconstruction (red). B Illustration of the iterative overlap matrix calculation, registration, and mesh combination strategy. C Exemplary cross sections at the shaft before (left) and after registration (right).
2023
-
Femur Reconstruction in 3D Ultrasound for Orthopedic Surgery Planning
Christian Gebhardt, Lara Göttling, Lukas Buchberger, Christian Ziegler, Felix Endres, Quirin Wuermeling, Boris M. Holzapfel, Wolfgang Wein, Ferdinand Wagner, and Oliver Zettinig
International Journal of Computer Assisted Radiology and Surgery, Jun 2023
2022
-
Standardized Evaluation of Current Ultrasound Bone Segmentation Algorithms on Multiple Datasets
Prashant Pandey, Benjamin Hohlmann, Peter Broessner, Ilker Hacihaliloglu, Keiran Barr, Tamas Ungi, Oliver Zettinig, Raphael Prevost, Guillaume Dardenne, Zian Fanti, Wolfgang Wein, Eric Stindel, Fernando Arambula Cosio, Pierre Guy, Gabor Fichtinger, Klaus Radermacher, and Antony Hodgson
In CAOS 2022. The 20th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery, Jun 2022
