RSNA 2009 

Abstract Archives of the RSNA, 2009


Detection of Dynamic 3D Motion Patterns of Joints from 2D Projection Images

Scientific Papers

Presented on December 4, 2009
Presented as part of SST15: Physics (Image Reconstruction III)


Remmet Jonges, Abstract Co-Author: Nothing to Disclose
Bart Carelsen, Abstract Co-Author: Employee, Koninklijke Philips Electronics NV
Simon Strackee, Abstract Co-Author: Nothing to Disclose
Geert J. Streekstra PhD, Presenter: Nothing to Disclose


To understand the functioning of the wrist joint, the availability of dynamic 3D motion patterns of the carpal bones is essential. For this purpose we recently introduced the 4D-RX method. Although highly accurate, the drawback of this method is that it is limited to imposed cyclic motions. We propose a 2D-3D registration method based on bi-plane fluoroscopy images to allow for detection of arbitrary 3D dynamic motion patterns of carpal bones during voluntary motion of the hand.


For acquisition of bi-plane fluoroscopy images we used the Alura FD20/10 of Philips. During voluntary motion of the hand, image series of two flat panel detectors are simultaneously acquired. Additionally, a CT scan of the wrist is acquired from which radius, ulna and the carpal bones are segmented. Based on the segmented bones, two Digitally Reconstructed Radiographs (DRR’s) are constructed that are compared to the two fluoroscopy images obtained from the fluoroscopy unit. By manipulation of the poses of the segmented bones in the construction of the DRR’s, the two fluoroscopy images finally match the DRR’s. From the pose changes in the final match, the three translation and three rotation parameters of each of the 15 bones are obtained. The precision of the method was evaluated in a cadaver wrist. Moreover, the applicability of the method in vivo was tested during voluntary motion of the hand in a multitude of directions.  


The experiments on the cadaver specimen reveal that the precision of the estimated motion parameters depend slightly on the orientation of the wrist relative to the flat panel detectors (imprecision: maximally 0.05 mm for translations and 0.42 degrees for rotations). The in vivo tests reveal that the pose of the wrist relative to the flat panel coordinate system yield optimal results if the over projection of bones is minimized.


Precise estimation of translations and rotation parameters of carpal bones can be achieved using 2D-3D registration of dynamic bi-plane fluoroscopy images and static 3D CT images.


Estimation of 3D motion patterns of carpal bones during voluntary motion is feasible and potentially useful for patients with dynamic wrist problems.

Cite This Abstract

Jonges, R, Carelsen, B, Strackee, S, Streekstra, G, Detection of Dynamic 3D Motion Patterns of Joints from 2D Projection Images.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL.