Click here to


Are you sure ?

Yes, do it No, cancel

Multi-Body 3D-2D Registration for Image-Guided Reduction of Pelvic Dislocation in Orthopaedic Trauma Surgery

R Han1*, A Uneri1 , T De Silva1 , M Ketcha1 , N Sheth1 , G Kleinszig2 , S Vogt2 , G Osgood3 , J Siewerdsen1 , (1) Johns Hopkins Univ, Baltimore, MD, (2) Siemens Healthineers, Erlangen, Germany (3) Department of Orthopaedic Surgery, Johns Hopkins Medical Institute, Baltimore, MD


(Tuesday, 7/16/2019) 7:30 AM - 9:30 AM

Room: 225BCD

Purpose: Pelvic dislocation (at sacroiliac joint and/or pubic symphysis) is a devastating pelvic ring injury with poor long-term prognosis if reduction does not accurately restore natural morphology. Reduction is commonly assessed via intraoperative fluoroscopy, challenging even experienced surgeons. We present a registration method that relates a pelvis shape atlas to preoperative CT and intraoperative fluoroscopy for quantitative reduction guidance. The method solves a challenging multi-body 3D-2D registration for joint dislocation and yields a solution that is compatible with demanding workflows in fluoroscopy-guided orthopaedic trauma surgery.

Methods: A pelvic atlas, formed from manual segmentation of left/right hip and sacrum in 40 CT images, was used to construct a statistical shape model (SSM) and statistical pose model (SPM). Multiple dislocated bone segments definition is achieved by mapping the SSM to preoperative CT via active shape model (ASM). The poses of dislocated bone segments are estimated from the SPM, and the target reduction to restore morphology is computed. During reduction, multiple bone segments are registered to fluoroscopy via 3D-2D registration, obtaining 3D pose estimates from 2D images. Performance was tested in a simulation study of 40 CT images by simulating a range of dislocation patterns.

Results: The method successfully solved multi-body 3D2D registration without segmentation for the special case of joint dislocations (cf., comminuted fractures). The accuracy of patient-specific bone segmentations via ASM was (1.8±0.2) mm. Target reduction demonstrated translational error of (0.3±1.9) mm and rotational error of (0.1°±1.7°). 3D-2D registration achieved projection distance error of (1.7±0.3) mm with runtime of (42.5±10.2) s in initial implementation.

Conclusion: The system provides a novel means of image-based quantitative guidance and assessment of pelvic dislocation reduction using a patient’s preoperative CT and intraoperative fluoroscopy that are standard to common workflow. The method has the potential to reduce operating time, radiation dose, and improve surgical outcomes.


Image-guided Surgery


Not Applicable / None Entered.

Contact Email