Send to

Choose Destination
J Orthop Res. 2016 Jul;34(7):1224-33. doi: 10.1002/jor.23146. Epub 2016 Jan 18.

Automated measurement of fracture callus in radiographs using portable software.

Author information

Department of Computer Science, Boise State University, 1910 University Drive, Boise, Idaho, 83725-2055.
Department of Mechanical Engineering and Mechanics, Lehigh University, 19 Memorial Dr W, Bethlehem, Pennsylvania, 18015.
Department of Mechanical and Biomedical Engineering, Boise State University, 1910 University Drive, Boise, Idaho, 83725-2085.
Competence Center for Applied Biotechnology and Molecular Medicine, Equine Hospital, Vetsuisse Faculty, University of Zurich, Winterthurerstrasses 260, CH-8057, Zurich, Switzerland.
Cork University Hospital, University College Cork, Wilton, Cork, Ireland.


The development of software applications that assist the radiographic evaluation of fracture healing could advance clinical diagnosis and expedite the identification of effective treatment strategies. A radiographic feature regularly used as an outcome measure for basic and clinical fracture healing research is new bone growth, or fracture callus. In this study, we developed OrthoRead, a portable software application that uses image-processing algorithms to detect and measure fracture callus in plain radiographs. OrthoRead utilizes an optimal boundary tracking algorithm to semi-automatically segment the cortical surface, and a novel iterative thresholding selection algorithm to then automatically segment the fracture callus. The software was validated in three steps. First, algorithm accuracy and sensitivity were analyzed using surrogate models with known callus size. Second, the callus area of distal femur fractures measured using OrthoRead was compared to callus area manually outlined by orthopaedic surgeons. Third, the callus area of ovine tibial fractures was measured using OrthoRead and compared to callus volume measured from micro-CT. The software had less than a 5% error in measuring surrogate callus, and was insensitive to changes in image resolution, image rotation, and the size of the analyzed region of interest. Strong positive correlations existed between OrthoRead and clinicians (R(2)  = 0.98), and between 2D callus area and 3D callus volume (R(2)  = 0.70). The average run time for OrthoRead was 3 s when using a 2.7 GHz processor. By being accurate, fast, and robust, OrthoRead can support prospective and retrospective clinical studies investigating implant efficacy, and can assist research on fracture healing mechanobiology. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 34:1224-1233, 2016.


callus; computer aided diagnosis; fracture; radiography; segmentation

[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for Wiley Icon for PubMed Central
Loading ...
Support Center