Study For Subsolid Pulmonary Nodule Segmentation Methods Based On Active Contour Model

Lung cancer is the most common malignancy in the world.Early lung cancer presents as pulmonary nodules on CT images.Clinical study showed that the partial solid pulmonary nodules(PSPN),juxta-pleural sub-solid pulmonary nodules(JP-SSPN)and juxta-vascular partial solid pulmonary nodules(JV-PSPN)as malignant nodules in high probability.In addition,the size and variation of solid components in PSPN are of great value in the qualitative diagnosis of nodules.In the computer-aided diagnosis system of pulmonary nodules,the accurate segmentation of pulmonary nodules is the basis of improving the diagnosis accuracy.Therefore,in order to accurately segment JP-SSPN and the solid components in PSPN,JV-PSPN,several improved active contour models were proposed in this paper.The specific contents are as follows:Firstly,an active contour model combining wavelet energy and Bayesian probability was proposed for the solid components segmentation problem in PSPN.On the one hand,in view of the low contrast and uneven brightness between the solid components and the surrounding ground glass,the model take the blurred wavelet energy to construct the regional term,introduces the local regional information,and drive the contour curve to the boundary of the solid components.Due to the multi-scale characteristics of Wavelet energy,which can enhance the differentiation between the solid components and the surrounding ground glass.On the other hand,for the fuzzy boundary of the solid components,the speed function based on Bayesian probability is constructed.The speed function approaches to zero at the boundary of the solid components,and complete the segmentation of the solid components in PSPN.Then,an active contour model combining wavelet energy and Hessian matrix index is proposed for the solid component segmentation problem in JV-PSPN.On the one hand,the blurred wavelet energy is used to construct the regional term of the model,combine with the local regional information,and drive the evolution of the contour curve to the boundary of the solid components.On the other hand,aiming at the problem that the solid components and blood vessel have similar gray value,combining with blurred wavelet energy and Hessian matrix index,a fuzzy clustering algorithm is constructed to calculate the fuzzy membership degree of the image.The fuzzy membership degree can enhance the differentiation between the solid components and the background tissue.Then,according to the fuzzy membership degree,the speed function of the model is calculated.The speed function approaches to zero at the boundary of the solid components,and complete the solid components segmentation in JV-PSPN.Finally,an active contour model based on robust speed function was proposed for the segmentation problem of JP-SSPN with fuzzy boundary and similar gray value to pleura.Firstly,a fuzzy k-nearest neighbor algorithm combining wavelet energy and local binary pattern features was constructed to calculate the membership degree of the image.The membership degree can enhance the distinction between pulmonary nodules and pleura and other background tissues.Then,three robust speed functions are calculated according to the membership degree and introduced into the model.At the boundary of JP-SSPN,the three robust velocity functions approach to zero,and complete the precise segmentation of JP-SSPN.