Study On The Key Techniques Of3D Reconstruction System Of Stomatology Image

In three-dimensional reconstruction system of stomatology images, withthree-dimensional reconstruction technique, the data from CT and other tomographicimaging devices were employed to display salient anatomical features, closely related toimplantation and located in the maxillary and mandible, and other structures dentistsinterested in. And then the doctors could determine surgery scheme according to relativespatial relationships and geometry of the different anatomical structures. Structures suchas single-tooth shape, regional shape of the whole teeth, maxillary sinus bottom, inferioralveolar neural tube, mental foramen, nasal palatine, periodontal ligament, and so onwere the objects of attention for doctors. Currently, three-dimensional reconstructionsystem of stomatology images is primarily a foreign product, which can realizereconstruction of basic structures according to three-dimensional scan data, and exportcorresponding reconstructed surfaces for computer-aided design and production, whichhad a positive role on enhancing stomatology treatment of dental restored planting andcorrection. However, the three-dimensional reconstruction technique of existing systemhas lower efficiency in drawing the shape of a tooth, but also can't effectively solve theproblems of tissue occlusion in the volume reconstruction of the whole stomatologyregion, so users still give priority to two-dimensional reconstruction plane such ashorizontal plane (X-Y plane), coronal plane(X-Z plane), and sagittal (Y-Z plane).Moreover, system can't reconstruct directly structures that have lower contrast, forexample inferior alveolar neural tube, only to adopt hand drawing. In addition, theconstruction system is limited by the image registration of soft issue, and lack ofconstruction of the soft issue such as periodontal ligament. Relatively, the domestic3Dreconstruction system of the stomatology images remains in the initial stage.In order to realize the localization of3D reconstruction system of stomatologyimages, in this paper, we launched a study for the key technical issues encountered inthe3D visualization and reconstruction for related oral structures, including imagesegmentation of the inferior alveolar neural tube, image registration for paraffin slices of periodontal ligament sequences, the methods of feature point extraction forreconstruction of the curve of the tooth, and reconstruction for the spatial structures ofperiodontal ligament in the whole region of stomatology.In this paper, the main work is demonstrated in the following:1. Proposed an approach for segmentation of the image of inferior alveolar neuraltube, which according to shape driven level set segmentation model restrained by localinformation. This method introduced the local information of inferior alveolar neuraltube images based on the shape driven level set algorithm, and guided the evolution ofthe level set curve together with local, regional, boundary and shape information, andarrived the goal of effective segmentation of inferior alveolar neural tube with weakchanged boundary region. Two sets of patient data were compared, the average relativeultimate measurement accuracy of the improved segmentation method is1.716%and1.692%, which surpasses4.432%and4.115%of the traditional method based on shapedriven level set obviously.2. Proposed a registration method combining the regular step gradient descent of2D rigid registration and the symmetric logarithmic domain diffeomorphic Demonsnon-rigid registration in a cascading way. We realized the registration of micrographsimages of paraffin sections of the periodontal ligament. Compared to with the non-rigidregistration method based on finite element, the registration method has moreadvantages remaining the topology of images, and the average time for every image was3.1%of the finite element method, while the average registration error was89%of themethod.3. Proposed a method for3D dental feature point extraction based on theimproved discrete curve evolution model. This method can be adaptive to determine theamount of feature points for the curve of the edge between different layers of CT imagesof the teeth, and thus reduce the redundancy of data storage, and improve the noiseimmunity of the feature points, with higher efficiency. In contrast to existing discretecurve evolution algorithm, the time for the improved method, extracting feature pointson every layer of CT images, required is around50%of the discrete curve evolutionalgorithm, and the quantity of the extracted feature points is about80%of the discretecurve evolution algorithm. 4. To reconstruct periodontal ligament collagen fibers, proposed a3Dreconstruction method, which based on mathematical modeling of the transfer functionof the3D reconstruction, and reconstruction of collagen fibers with ray castingalgorithm. The method reduced excessive demands of a priori knowledge of the transferfunction design for the operator. The reconstructed spatial structures of collagen fiberswere consistent with anatomical structures. Furthmore, to reconstruct the wholestomatology region, we improved the selection method of threshold for the LMIPvolume rendering algorithm, and proposed an approach to set the threshold according tothe difference between the first peak and the global peak of the ray to determine thefinal peak. This method weakened the problem the reconstruction noise is too bigcaused by the difference. In addition, this paper presented a method fusing improvedLMIP and ICPVR to render target in the whole stomatology region, the method couldreconstruct simultaneously internal and external structure of each anatomical object, andimprove the visualization effect of target effectively.Improvements presented in this paper solved urgent problems in3D reconstructionsystem of stomatology images, and promoted the localization progress of3Dreconstruction system of stomatology images, but efficiency of the implementation ofthe algorithm need further improvement in the future.