Methods For CBCT Image Processing And Intelligent Analysis For Dental Implant Planning

The dental office workflow is rapidly evolving towards a more efficient and cost-effective chain using state-of-the-art technology.The rapid development of dental digitalization has placed new demands on CBCT(Cone Beam CT)image processing.The three-dimensional model of the patient,which is acquired by the optical imaging device,needs to be registered and fused with the anatomical structure in the CBCT image to play its due role.Therefore,the automatic segmentation of key anatomical structures in dental CBCT images is an important research issue in the development of dental digitalization.This thesis mainly completed the research on automatic panoramic image reconstruction from dental CBCT,automatic segmentation of mandibular canal based on panoramic image,automatic segmentation of tooth region and key technology design of dental implant planning.The main research works and contributions are as follows:(1)Automatic reconstruction method for hight-contrast panoramic image from dental CBCT data.In this study,we developed a fully automatic reconstruction method to improve the global and detail contrast of panoramic images.The thickness detection of the dental arch effectively overcomes the inaccuracy of the manual setting and reduces the non-interesting tissues in panoramic image reconstruction.Then,a new synthesis algorithm is proposed at image synthesis stage to reduce the effect of non-interest tissues on image contrast.Finally,an image enhancement algorithm based on image filtering is applied to the synthesized image to improve the detail contrast of the final panoramic image.A total of 129 real clinical dental CBCT data sets were used to test the proposed method.The panoramic images generated by three methods were subjectively scored by three experienced dentists who were blinded to the generated method.The overall image contrast score revealed that the proposed method scored the highest of 11.03±2.46,followed by the ray sum and x-ray methods with corresponding scores of 6.4±1.65 and 5.35±1.56.The results of expert subjective scoring indicated that the image contrast of the panoramic image generated by the proposed method is higher than those of existing methods.(2)Automatic segmentation of IAN(Inferior Alveolar Nerve,IAN)based on panoramic image.First,the ray casting algorithm is used to obtain the rendered image of the multi-planar reconstruction data.Then,the mental foramen is detected in the rendered image using the gLoG algorithm.The panoramic image containing the jaw bones is reconstructed.EfficientUNet is used to segment the IAN canal in the panoramic image to obtain the mask of the IAN canal.Finally,according to the mask of the mental foramen and IAN canal positon,the IAN canal in the data was segmented and then passed through the Dijkstra algorithm.Perform a global search to get the final segmentation result.Compared with the results of manual segmentation,the recall rate,accuracy and average DSC of the three-dimensional IAN canal were 84.26%,87.53%,and 86.71%,respectively.(3)Automatic tooth segmentation.The tooth area accounts for about 9%of the total dental CBCT data.In order to reduce the amount of training data and improve the accuracy of prediction,an automatic detection algorithm for fast tooth region of interest is improved.Then,the tooth are segmented by 3D U-Net network.A total of 49 patients tooth were labeled in the CBCT data,40 for network training and 9 for testing.The recall rate,accuracy and average DSC of the segmentation results were 87.21%,98.23%,and 92.10%,respectively.(4)Key technology design and development of dental implant planning software.To be able to apply the research content to clinical practice,the dental implant planting software system was designed and implemented.A combined transfer function design is proposed.The design is user-centric and uses a two-dimensional image as an interface,which simplifies the user’s operation steps.We proposed an automatic measurement method for the amount of bone removal of the alveolar at the planting position and designed as a widget.We have completed the MPPS(Modality Performed Procedure Step,MPPS)service and multi-threaded DICOM(Digital Imaging and Communication in Medicine,DICOM)storage service process design.Currently,the 3D rendering module and panoramic image processing module in the system have been clinically verified.