Technical Brief

Influence of Scan Resolution, Thresholding, and Reconstruction Algorithm on Computed Tomography-Based Kinematic Measurements

[+] Author and Article Information
Christopher John Tan

Surgical and Orthopaedic Research Laboratories,
Prince of Wales Clinical School,
University of New South Wales,
Sydney, NSW 2031, Australia;
Sydney Veterinary Emergency & Specialists,
675 Botany Road,
Rosebery, NSW 2018, Australia
e-mail: chris.tan@sydneyvetspecialists.com.au

William C. H. Parr, William R. Walsh

Surgical and Orthopaedic Research Laboratories,
Prince of Wales Clinical School,
University of New South Wales,
Sydney, NSW 2031, Australia

Mariano Makara, Kenneth A. Johnson

Sydney School of Veterinary Science,
The University of Sydney,
Sydney, NSW 2006, Australia

1Corresponding author.

Manuscript received November 2, 2016; final manuscript received August 4, 2017; published online August 23, 2017. Assoc. Editor: Joel D Stitzel.

J Biomech Eng 139(10), 104503 (Aug 23, 2017) (5 pages) Paper No: BIO-16-1436; doi: 10.1115/1.4037558 History: Received November 02, 2016; Revised August 04, 2017

Radiographic data, including computed tomography (CT) and planar X-ray, is increasingly used for human and animal kinematic studies. There is a tendency toward using as high-resolution imaging as possible. Higher resolution imaging is one factor (in conjunction with the reconstruction algorithm), which may increase the precision of reconstructed three-dimensional (3D) surface models in representing true bone shape. However, to date no study has tested the effects of scan resolution, threshold, and 3D model reconstruction algorithm on the accuracy of bone kinematic results. The present study uses a novel method to do this where canine tarsal bones were positioned on a radiolucent Lego board and scanned before and after undergoing known translations and/or rotations. The digital imaging and communications in medicine (DICOM) images were acquired using two different CT scanning resolutions and processed using three different segmentation threshold levels and three different reconstruction algorithms. Using one bone as the reference bone, an iterative closest point (ICP) algorithm was used to register bones to a global co-ordinate system and allow measurement of other bone kinematics in terms of translations and rotations in and around the x-, y-, and z-axes. The measured kinematics were compared to the “known” kinematics, which were obtained from the Lego board's manufacturing standards and tolerances, to give accuracy error metrics for all bones. The results showed error in accuracy of measured kinematics was at subvoxel levels (less than 0.5 mm). Despite altering the volume and surface area of the 3D bone models, variation in resolution, segmentation threshold and reconstruction algorithm had no significant influence upon the accuracy of the calculated tarsal bone kinematics.

Copyright © 2017 by ASME
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Fig. 1

Position of the eleven bones during the before scans (a) and after scans (b). All motion is described relative to the reference bone, the calcaneous (*). The direction of motion is described relative to the global co-ordinate axes shown.

Grahic Jump Location
Fig. 2

Workflow for the scan alignment and calculation of bone kinematics. Initial isosurface reconstructions of the lego boards with bones in the two positions are not aligned with the global coordinate system or one another (a) and (b). The first step is to align the before scan with the global coordinate system (c). The after scan is also roughly aligned with the global coordinate system (d). The second stage of alignment uses an ICP algorithm to minimize translational and rotational differences between the scans using the calcaneus (bone with * shown) of the before scan as the fixed entity and superimposing the calcaneus in the after scan along to this. The remainder of the after lego board and bone models are moved with the after calcaneus, but do not influence the alignment (e). The before model for each bone is then aligned (again using ICP alignment) with the after position and saved separately from the before model in the before position (f). Thereby, the same 3D model is stored in both before and after positions.

Grahic Jump Location
Fig. 3

Effect of scan resolution and smoothing on visual appearance of 3D surface model of canine calcaneous segmented with a threshold of 900 HU. Top: 3D surface models generated from high-resolution scans and high-quality model generation (left), optimal quality model generation (center), and optimal quality model generation and additional smoothing (right). Bottom: 3D surface models created from low resolution scans and high-quality model generation (left), optimal quality model generation (center), and optimal quality model generation and additional smoothing (right).

Grahic Jump Location
Fig. 4

Effect of segmentation threshold level of surface generation of 3D bone model. The inner blue line represents the surface generated with a threshold setting of 1300 HU, the middle red line represents thresholding at 900 HU and the outer green line represents the surface generated at 500 HU for a small portion of one of the bones used in the study. Changes in model volume and surface area will occur but there is equal effect across the entire bone surface.




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In