Study Of Segmentation Methods Of Thyroid Nodules Based On Active Contour Model

Thyroid nodule is a common disease of endocrine system,and malignant nodule indicates the coming of the thyroid cancer.However,the mortality rates of thyroid cancer accounts for about 0.2% of all cancer deaths.It shows that most of the thyroid cancer can be cured.Ultrasonic inspection is considered as the most commonly used method for general survey and diagnosis of the thyroid diseases.It plays important roles in early detection and treatment for improving the cure rates of thyroid cancer.While,inspecting of the images by doctors is not only inefficient but also subject to cause misdiagnosis and missed diagnosis.In order to solve these problems,computer aided diagnosis technology based on ultrasonic inspection is widely used in clinic.This paper devotes itself in segmenting the thyroid ultrasonic image.In order to improve the segmentation accuracy and efficiency,we studies the segmentation method of geometric active contour model.In this paper,we study the geometric active contour model.Because it is based on the curve evolution theory and level set method,and its implicit expression overcomes the parameterized expression,moreover,curve evolution process is based on the geometrical characteristics of the curve.Among them,the DRLSE model use variational level set method. The addition of distance regularization item avoids the problem of the re-initialization during the curve evolution and improves the speed of the curve evolution.However,in the CV model,the image is divided into two regions,and use the average gray level to distinguish between target and background.When the average gray-scale difference reaches its maximum, the segmentation task is completed.Therefore,CV model is a kind of global optimal segmentation method.First,we use DRLSE model as the research object.While,this model can't accurate segment thyroid ultrasound images which the boundary aren't clear.So,this paper proposes DRLSE segmentation model improving edge stopping function.The edge stopping function plays a key role in the evolution process of the curve.We put global information into the edge stopping function,obtain a new model that gradient combined with global information.The simulation results not only reduce the sensitivity of the initial position,but also enhance the ability of detection the fuzzy edge of thyroid ultrasound images.Secondly,we use CV model as the research object.As CV model is difficult to segment the image with intensity inhomogeneity.So,the improved CV segmentation model of combining local information is proposed. We constructed a new speed function by using local fitting information.Due to the nuclear window has controllability,a pixel can strictly dependent on its neighbor pixels,and overcomes intensity inhomogeneity problem.Then, the speed function is incorporated into the CV model,avoid the problem of weight allocation.The improved CV model can accurately segment ultrasound images with intensity inhomogeneity.Finally,the two improved models were compared.The simulation results show that the improved CV model obtained the better segmentation effect.It not only has global segmentation ability, and can extract calcified plaque inside the nodules. At the same time, the accuracy of segmentation and segmentation efficiency is better than the improved DRLSE model.