Introduction: Morphological changes are often observed in condyles after orthognatic surgery. CBCT enables interpretation of volumetric changes and 3D visualization of condylar changes; however, accurate 3D reconstruction and volume rendering of the condyles is still hampered by low contrast resolution and isovalue distortion. The objective of this study is to present a novel semi-automatic segmentation method which overcomes these obstacles in 3D condyle analysis, with considerable time saving over manual segmentation.
Methods : A semi-automatic segmentation method, combining local thresholding and region growing, was used to segment both condyles in CBCT data of ten patients, previously used in another study utilizing a more manual segmentation approach. Segmentation was performed by two independent observers, and inter-observer Dice-coefficients were calculated. Similarity was compared by calculating Dice-coefficients and surface distances.
Results : Dice-coefficients between both observers were excellent (mean 0.98; range [0.95–0.99]), indicating high reproducibility. Comparison with the manually segmented condyles also yielded high Dice-coefficients (mean 0.96, range [0.94–0.98]). No systematic errors were observed between both methods.
Conclusion and discussion : The proposed segmentation method yielded excellent results, and could be valuable in assessing condylar morphological changes post-operatively. Fully automated segmentation methods are currently researched to further reduce time consumption.
Key words : condylar resorption; CBCT; segmentation