Research On Panoramic Image Processing Methods In Oral Implant Planning

With the rapid development of the digital oral process and the increasing demand for oral implants,cone beam computed tomography(CBCT)has gradually become a medical imaging technology widely used in assisting oral diagnosis,modeling,simulation and preoperative planning due to its own advantages.Especially in the preoperative diagnosis and planning stage of oral implant,medical staff usually use the panoramic image reconstructed from CBCT data to observe and evaluate the oral condition of patients,and then use the panoramic image to simulate the implantation of metal implants.Therefore,the high quality and automatic reconstruction of panoramic image of oral CBCT and the automatic segmentation of key anatomical structures in panoramic image have become the key issues in the research of assisted implant surgery technology.Focusing on this key issue,the main content of this paper is the research of automatic panoramic image reconstruction based on CBCT,the construction of panoramic image data set for the mandibular region,the research on oral panoramic image segmentation method based on Transformer,and the design of semi-automatic dental implant selection method.The main research contents are as follows:(1)Research on automatic reconstruction of high-quality panoramic images based on oral CBCT data.This paper presents a method that can automatically reconstruct a high-quality panoramic image of the mouth.The proposed method has two characteristics: firstly,it can detect the patient’s dental arch curve by projecting a small number of axial maximum intensity images from different slice ranges;Secondly,it can redistribute the intensity of key areas such as teeth,upper and lower jaw tissues and oral plants in the panoramic image,so as to reduce the interference of oral plants on the quality and contrast of the panoramic image.At the same time,the objective evaluation criteria for the display characteristics of panoramic images and the subjective evaluation criteria in line with the actual use of doctors are designed.The proposed method has been tested on 50 CBCT datasets for objective and subjective evaluation.The results of both evaluation methods show that the panoramic images generated by the method have good image quality.(2)Construct panoramic image data set for mandibular region.The original images of the mandibular region dataset(MRDataset)are from multiple open access panoramic X-ray image sets of the oral cavity.After removing blurry,overly overlapping,and repetitive images that are difficult to distinguish,the equipment parameters and other information on the filtered panoramic X-ray image are cropped and renumbered.This dataset consists of a total of 711 images,divided into 8 categories based on the number of teeth and the presence or absence of treatment areas,and includes 4 annotation labels: mandible,normal teeth,treatment teeth,and implant teeth.An effective dataset has been constructed for computer-aided dental image research.(3)Research on Segmentation Method of Oral Panorama Image Based on Transformer.Based on the characteristics of medical image segmentation datasets,and by analyzing the advantages and disadvantages of models based on CNN and Transformer,it is proposed to use the Transformer module and CNN module to jointly construct a hybrid segmentation model CBTrans that can extract different features of input images.This model parallelizes the CNN based on bottleneck module with the Swin Transformer module to form a hybrid module,and adds the hybrid module to the U-shaped encoder decoder architecture based on the U-shaped architecture.The model was tested using the mandibular region dataset and the retinal vascular dataset,and the experimental results showed that CBTrans exhibited excellent segmentation performance on both the mandibular region dataset and the retinal vascular dataset.