| Maxillary sinus(MS)is the largest paranasal sinus in the human bodyand the anatomical structure of MS is cone-shaped.Thickened membrane and cysts are common morphological changes in the maxillary sinus.Due to the complex anatomical structure of the MS and its relevance to oral pathology,we need to segment the maxillary sinus membrane and quantify the thickness of the mucosal thickening.Therefore,this article develops an automatic segmentation technology to segment the maxillary sinus membrane and cysts,which can be helpful to provide the relevant technical support for preoperative diagnosis.The methodology of this research can be summarized as follows:(1)The first step is to preprocess the CBCT data,including CBCT image denoising and ROI region cropping,and the segmentation of the MS air cavity can be obtained by the robust statistics segmenter,and then it can generate the 3D model of the MS air cavity.(2)The second step is to segment the MS bone cavity.First,the fuzzy C-means(FCM)algorithm is used to automatically segment the maxillary sinus bone cavity.On the basis of FCM segmentation result,the neighborhood membership retrieval algorithm is used to retrieve the thin bone voxel belonging to the condition of bone membership and divide them into bone clusters.Then we can use the posterior sinus wall thin bone enhancement method to segment the thin bone boundary on the posterior sinus wall.The proposed method can be used to locate the thin bone boundary and enhance them to a higher bone intensity,which is based on the voxel intensity distribution on the posterior sinus wall.Then the thin bone located posterior sinus wallwas segmented and then the 3D mesh model was generated,which can be combined with the FCM segmentation result to construct the complete MS bone cavity model.(3)After completing the segmentation of the MS cavity,it will perform a vertex screening algorithm on the MS cavity to select the vertices on the inner wall of the MS cavity.Before screening the vertices of the MS bone cavity,we need to mark the Apex point on the zygomatic bone and Medial1-4 on the medial wall of nasal cavityrespectively.The slicing path can be automatically generated by marking points and then the slicing path can be also automatically placed parallel to the inner wall of the sinus cavity.The vertex screening algorithm is performed on each slicing path.The judgment condition of the vertex screening is whether the vertex on the inner wall of the sinus bone cavity faces the origin and whetherthe projection angle of the normal vector on the slicing path less than the angle threshold.At the same time,the vertex cluster closest to the center of the circle on the inner wall of MS bone cavitywas selected by the distance clustering algorithm.Then,we proposed angular-based mean distance filtering algorithm to filter out irrelevant points outside the sinus cavity.The Poisson surface reconstruction algorithm can be used to reconstruct the 3D mesh model of MSbone cavity based on the selected vertices.Finally,the morphological changes of the maxillary sinus membrane and cystswere obtained by subtracting the air sinus cavity model from the reconstructed bone sinus cavity.The proposed method has been applied to 10 cases of maxillary sinus with thickened mucosa or cysts,and the result proved that this method can successfully segment the maxillary sinus thickened membrane(<2 mm).The experimental results also prove that the errors of the volume difference and Hausdorff distance are respectively within 3.29%and1.08%.The automatic segmentation technology proposed in this paper can effectively and accurately segment the maxillary sinus thickenedmembrane and cysts.At present,the manual segmentation method is not only time-consuming,but also inaccurate.It also exhibits inter-operator variability and repeatability problems.In the proposed method,manual operations are eliminated with bone segmentation and vertex screening processes.It makes the proposed method repeatable and greatly improves the segmentation result. |
PDF Full Text Request