| Abstract [eng] |
Problem. In order to produce accurate anatomical model, which could help to plan and implement orthognathic surgery, it is important to correctly extract the jaw segments from the series of CBCT images. The variety of 3D modelling softwares‘ algorithms are: global threshold determining method – based on global Hounsfield scale value selected manually, despite of the fact, that the density of jaw bone is spatialy variable; active contour – when the region competition or edge as attractor is used – which handling is demanding and complicated, because of the planty of user predefined parameters. Method. The algorithm determining local threshold values via Otsu method was proposed for the segmentation of bone regions in CBCT images. The filter is involved into the algorithm, which checks if analysed CBCT image hystogram is bimodal. The filter was used for removal of irrelevant fields for the segmentation. For algorithm testing the synthetic 3D model was created, which voxels intensities, size, distribution and resolution are close to the real CBCT images. Results. The optimal parameters controlling the algorithm were determined by using synthetic data. It was found that: the optimal volumetric area (799 mm3) is needed to obtain local threshold; the maximal analysed field volumetric shift is 4.5 mm in the case of the rapidity increasing (til 10 times). The pilot study, with CBCT images from real clinical cases, was performed and it was found that proposed algorithm can be used for segmentation of bone fields in CBCT images. Conclusion. The algorithm is more efficient, more reliable and simpler to use comparing to the global threshold determining method. It provides a possibility to use the local information for bone segmentation. |
