Journal of Microscopy, 2014 doi: 10.1111/jmi.12159 Received 24 September 2013; accepted 23 June 2014 Influence of segmentation on micro-CT images of trabecular bone S. TASSANI ∗ , V. KORFIATIS † & G. K. MATSOPOULOS † ∗ Institute of Communication and Computer System, National Technical University of Athens, Zografou, Athens, Greece †National Technical University of Athens, Zografou, Athens, Greece Key words. Chan–Vese, micro-CT, Otsu, segmentation, trabecular bone. Summary Segmentation of biomedical images is of great impor- tance in various studies aiming to both the identifica- tion of regions of interests within the image and the performance of quantified measurements. Nevertheless, the segmentation of the biomedical images represents a wide range of medical cases and there is not a unique technique applicable to all kinds of medical images. In this study, three popular techniques for segmenting micro- CT images of bone microstructures are evaluated. Fixed threshold, Otsu’s algorithm and a modified version of the Chan–Vese segmentation technique have been applied on micro-CT images and have been compared to higher resolution golden standard, that is histological images. The modification of the Chan–Vese technique is based on the novel implemen- tation of a new initialization process called the Branch Point Initialization. Stereological measurements were performed on all the segmented images and statistically compared to the golden standard. Fixed threshold and the modified Chan–Vese technique have shown comparable results, with a maximum significant error of about 10%. However, Chan–Vese showed an easier, faster and more reliable segmentation procedure for optimal settings identification. The Otsu’s method showed a maximum error larger than 20%. Given the limits and advan- tages of the known segmentation techniques, the proposed modified Chan–Vese active contour technique shows high po- tential for use in the segmentation of micro-CT images as well as in other high-resolution X-ray images. This potential is augmented by the recent introduction of high-resolution clin- ical technologies for which standard techniques have already shown to be insufficient. Introduction Image segmentation is a very wide subject of image pro- cessing applied to biomedical imaging. It is a prerequisite Correspondence to: Simone Tassani, Institute of Communication and Computer System, National Technical University of Athens, 9 Iroon Polytechniou Street, 157 80 Zografou, Athens, Greece. Tel: +30-210-7722288; fax: +30-210-7723557; e-mail: tassani.simone@gmail.com for identification and further analysis and quantification of biomedical features. Since a single segmentation technique capable of segmenting all kinds of biomedical images does not exist, the choice of a technique for segmentation image oriented. In high-resolution micro-CT images of trabecular bone, a fixed threshold was found to be a reasonable solution for segmentation of trabecular structure (Kuhn et al., 1990; Muller et al., 1994, 1998; Perilli et al., 2007). Fixed threshold requires minimum computational cost and can be easily used on micro-CT datasets, despite the size of 1 GB or more. To select the appropriate threshold level, however, a comparison with the golden standard is required, usually by means of synchrotron facilities, histological studies or Archimedes’ principle (Ding et al., 1999; Perilli et al., 2007; Kazakia et al., 2008). In the first case, the researcher needs to use a structure with high costs and not easily accessible whereas, in the case of comparison with the histology, the development of the study requires a long period of time. Finally, the application of the Archimedes’ principle on trabecular bone is still questionable (Schileo et al., 2008, 2009). However, by comparing micro-CT images to higher reso- lution images, it is possible to identify an appropriate thresh- old level, but it depends on the specific acquisition set-up. Moreover, limits of the fixed threshold for the segmentation of micro-CT images were already pointed out for specimens with bone volume fraction (BV/TV) below 15% (Hara et al., 2002). Generally, when the dimension of the analysed objects become comparable to the image resolution, the partial vol- ume effect significantly affects the image segmentation, and the estimation error of specific biomedical features increases. It is clear how the fixed threshold can introduce significant errors, especially when high BV/TV specimens are compared to low BV/TV ones, i.e. comparison between osteoporotic and control specimens. To reduce these errors, many adapting techniques were introduced. In the Otsu’s method, a fixed threshold is auto- matically identified for every new image depending on its his- togram (Otsu, 1979). Furthermore, new sophisticated tech- niques based on active contours have been also developed over the years towards image segmentation (Osher & Sethian, 1988, Chan & Vese, 2001, Thevenaz & Unser, 2008). The C  2014 The Authors Journal of Microscopy C  2014 Royal Microscopical Society