Identification of fractured bone tissue from CT images

Abstract

The identification of fractured bone from computed tomographic (CT) images is a helpful task in medical visualisation and simulation. In many cases, specialists need to manually revise 2D and 3D CT images and detect bone fragments in order to check a fracture. The automation of this process would allow them to save time. In visualisation, it allows the reduction of image noise and the removal of undesirable parts. In simulation, the utilisation of models reconstructed from CT images of patients allows customisation of the simulation, because the result of the segmentation can be used to perform a reconstruction that provides a 3D model of the patient anatomy. In this paper, the main issues to be considered in order to identify both healthy and fractured bones are described. The identification of fractured bone requires not only to segment bone tissue but also to label bone fragments. Moreover, some fragments can appear together after the segmentation process, hence additional processing can be required. Currently proposed approaches for identifying healthy and fractured bones are analysed and classified. Finally, the aspects to be improved in the described methods are outlined and the work to do is identified.