Publication Details
Virtual 2D-3D Fracture Reduction with Bone Length Recovery Using Statistical Shape Models
Madeja Roman (FIT)
Španěl Michal, doc. Ing., Ph.D. (DCGM)
Čuta Martin, Mgr., Ph.D. (FIT)
Zemčík Pavel, prof. Dr. Ing., dr. h. c. (DCGM)
STOKLÁSEK, P.
MIZERA, A.
Preoperative planning, Fracture reduction, Fixation devices, 2D-3D registration, Statistical shape model
Computer-assisted 3D preoperative planning based on 2D stereo radiographs has been brought into focus recently in the field of orthopedic surgery. To enable planning, it is crucial to reconstruct a patient-specific 3D bone model from X-ray images. However, most of the existing studies deal only with uninjured bones, which limits their possible applications for planning. In this paper, we propose a method for the reconstruction of long bones with diaphyseal fractures from 2D radiographs of the individual fracture segments to 3D polygonal models of the intact bones. In comparison with previous studies, the main contribution is the ability to recover an accurate length of the target bone. The reconstruction is based on non-rigid 2D-3D registration of a single statistical shape model onto the radiographs of individual fragments, performed simultaneously with the virtual fracture reduction. The method was tested on a syntethic data set containing 96 virtual fractures and on real radiographs of dry cadaveric bones suffering peri-mortem injuries. The accuracy was evaluated using the Hausdorff distance between the reconstructed and ground-truth bone models. On the synthetic data set, the average surface error reached 1.48+1.16 mm. The method was built into preoperative planning software designated for the selection of the best-fitting fixation material.
@inproceedings{BUT155083,
author="KLÍMA, O. and MADEJA, R. and ŠPANĚL, M. and ČUTA, M. and ZEMČÍK, P. and STOKLÁSEK, P. and MIZERA, A.",
title="Virtual 2D-3D Fracture Reduction with Bone Length Recovery Using Statistical Shape Models",
booktitle="ShapeMI MICCAI 2018: Workshop on Shape in Medical Imaging Proceedings",
year="2018",
series="LNCS",
journal="Lecture Notes in Computer Science",
number="11167",
pages="207--219",
publisher="Springer International Publishing",
address="Granada",
doi="10.1007/978-3-030-04747-4\{_}20",
isbn="978-3-030-04746-7",
issn="0302-9743",
url="https://www.fit.vut.cz/research/publication/11806/"
}