Hierarchical Segmentation Of Facial Soft Tissue Based On High Resolution MRI Image

Hierarchical segmentation of facial soft tissue is mainly referring to the recognition and segmentation of the skin,subcutaneous fat,and muscle.This study can provide doctors with qualitative and quantitative measurements of the specific target tissue,which are a mathematical basis for medical diagnosis and clinical treatment.The three-dimensional model constructed on the basis of segmentation can also provide doctors with simulation of surgical program and simulation of postoperative effect.Therefore,this study can not only improve the accuracy and objectivity of the diagnosis,but also has an important significance to improve the quality of plastic surgery.At present,the domestic and foreign scholars have focused on the segmentation research of soft tissues in the brain,heart and lung tissues,for the facial soft tissue research work is relatively small.In addition,the facial soft tissue segmentation based on MRI images is susceptible to noise,artifact,intensity inhomogeneity,partial volume effect and adjacent similar gray-scale tissues.Therefore,it is of significance to study the segmentation of facial soft tissue based on MRI images.The main contents and achievements of this article are as follows:1.According to the spatial continuity of MRI images and the similarity of the anatomical structure between adjacent slices,the segmentation result of the above slice is regarded as a priori,and the SBGFRLS method is used to realize the continuous segmentation of skin tissue.At the same time,iterative stopping conditions based on the length of the evolution curve and the position changes of the points on the curve between two iterations are added to the SBGFRLS method in order to solve the false segmentation caused by the air cavity structure.2.An adipose tissue segmentation method is designed by combining MICO algorithm and region growing.This method firstly uses the MICO algorithm and the image operation to deal with the MRI images with intensity inhomogeneity to obtain the initial gray-corrected segmentation images of adipose tissue.Then using region growing to do the second segmentation on the primary images.And performing the morphological processing to the above results to remove the small holes.The segmentation results of this method are more accurate than the traditional methods like region growing,FCM,active contour model and graph cuts.3.A modified distance regularized level set evolution is presented to segment the masseter in MRI images.The presented method introduces the phase congruency into the edge indicator function to solve the segmentation problem caused by the edge indicator function which controlled by image gradient falls into a local minimum,and improves the segmentation accuracy.