Research On The Algorithm Of Organ Recognition And Image Segmentation Based On Geometric Active Contour Model

Magnetic induction therapy(or says magnetic thermoseed heat therapy)is a new technique of tumor heat therapy developed in recent years.At present,the segmentation of clinical tumor or organ depend on the doctor's manual operation greatly,conformal heat treatment is a major technical difficulty due to the low speed and efficiency.The planning system can assist doctors to identify the lesions accurately,and calculate the thermal fields according to the pattern of the lesion to inactivate the pathological tissue,reduce or avoid normal tissue and important structure to be damaged.Thesis aim to provide a feasible segmentation algorithm for planning system.Focus on the core issue that how to determine the geometric properties and space of human tissues,organs and treatment areas during the process of induction targeting and conformal thermal therapy planning,implement the outline method and tissue segmentation method respectively,based on geometric active contour model and CT sequence.These two methods cover the two-dimensional and three-dimensional segmentation of medical imaging,and can meet the clinical application requirements of the planning system for the target area extraction.The combined image energy level set algorithm is used for contour sketch,this algorithm introduces the regular term and integrates the appropriate energy functional,reduced the deviation of the level set function and the signed distance function by minimizing the energy functional,so the level set not required to be initialized again.Use the finite difference method to solve the partial differential equation of the image energy function,which accelerates the evolution of the curve.By introducing the regional detection function,variable weight coefficient and new termination function,the algorithm not only self-adjust the evolution direction,can divide the multi-objective,the weak edge and the concave shape,the inner cavity pattern accurately,and have a nice performance in CT image segmentation experiment.The segmentation algorithm with local adaptive and robust feature statistics for 3-D medical image(LARFS)is used for the recognition of human tissues and organs,so,the problem of insufficiency or leakage of traditional algorithms is solved preliminarily.This algorithm uses the 3d reconstruction data field information,combine with traditional regional growth algorithms and statistical methods,and the driver of contour evolution is based on the geometric active contour model,so,the regional information(mean,variance and probability density function)of the image is fully utilized to represent the local segmentation characteristics.The experimental evidence suggests that the LARFS using three-dimensional body element information can achieve good segmentation results under suitable growth factors for different organs.