A Portable Software Application to Measure Fracture Callus in Radiographs

2014 
Introduction: The clinical evaluation of fracture fixation technology requires accurate methods to objectively measure clinically relevant metrics. The formation of fracture callus is visible on standard radiographs and has relevance to healing and mechanical integrity at the fracture site.[1-2] However, established techniques for measuring fracture callus in radiographs are timeintensive and subjective, with inter-physician variability of 20-25%.[3] A computational algorithm was recently introduced to improve the objectivity of callus measurement.[4] However, this algorithm was developed using non-portable, proprietary software and it did not have a graphic user interface. Furthermore, this algorithm had an unsatisfactory processing time and could only analyze images where the cortex was vertically aligned. These limitations have precluded a widespread adoption of this imaging tool by clinicians and scientists. The objective of this study was to overcome these limitations by developing a fast, portable and robust software application to accurately identify and quantify fracture callus in digital radiographs. Methods: A portable Java application called OrthoReadTM (Fig. 1) was developed to measure periosteal fracture callus in digital radiographs. This software tool automatically calculates callus size in a user-defined region of interest (ROI) after the external cortex is segmented. Cortex segmentation requires the user to define two points, or limits, on the external cortex of each fracture fragment: one limit near the ROI border and one limit near the fracture site (Fig. 2A, 2B). These limit positions are extrapolated to the ROI border, and are then input into an optimal boundary tracking method, known as intelligent scissors, that uses a series of local cost functions to detect the external cortex boundary.[5] Once the external cortex is defined, cortical fragments are bridged, and a Sobel edge detection algorithm is used in conjunction with a dynamic thresholding algorithm to outline the callus region. Filters and morphological operations are used to reduce noise while maintaining edge contrast. Conversion to metric area is achieved using PPI or a hardware feature of known dimension.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    0
    References
    0
    Citations
    NaN
    KQI
    []