Three-dimensional Reconstructions Of The Upper Air Way And Paransal Sinuses

Cone beam computed tomography images are often used to help doctors accurately diagnose potential diseases of a patient such as lesions in the upper airway.Therefore,a new and simple algorithm is proposed to segment the upper airway and maxillary paranasal sinuses.This algorithm mainly uses region growing and thresholding techniques.It is necessary to preprocess the image before segmentation.Due to the limitations of imaging equipment and acquisition conditions,the obtained original medical images are often affected by degradation and noises,so it is necessary to smooth and enhance the image.There are three kinds of image segmentation methods: manual segmentation,manually interactive segmentation,and automatic segmentation.Much of the previous research has focused on the segmentation of the lower airway tract.This paper proposes an automatic upper airway segmentation algorithm,which is mainly based on threshold segmentation and three-dimensional region growing.In the algorithm,three-dimensional region growing is the main part,region growing method needs to determine the seed points.In many previous algorithms,the selection of seed points is generally chosen by hand,which usually depends on the experience of operators,so it is unstable.In this paper,an automatic segmentation algorithm for upper airway is proposed.The seed points are selected automatical ly according to the location distribution of the main upper airway tract in CBCT images.The airway is then grown based on some specific growth criteria.Finally,three-dimensional reconstruction of the upper airway is performed.In addition,the proposed algorithm is compared to the manually segmented images in terms of recall rate,structural similarity and accuracy,to verify the effectiveness of the proposed algorithm.For the maxillary sinus,manual segmentation is often adopted.It will take a lot of time and affect the working efficiency of doctors.Sinus segmentation can not only determine the volume of the sinus,but also determine whether the nasofrontal canal is open.Accurate segmentation of sinus tissue is an important basis for helping doctors to make surgical plans and surgical guidance.Therefore,an automatic segmentation algorithm is proposed for maxillary paranasal sinuses,and the volumes of left and right maxillary sinus are calculated.Due to the physiological structure of the maxillary sinuses,there are small gaps between the sphenoid sinus and the maxillary sinus on some sections of CBCT images.In the three-dimensional region growing,the six-neighborhood of the seed point pixel,are the left,top,right and lower voxel of the seed point in the same slice,and the two nearest points in the adjacent upper and lower slices,which are named before and after.The maxillary sinus can be separated from other tissues by limiting the intensity of its before and after pixels in the six points.