三维病例CT图像中的快速挖掘方法研究与仿真

Research and Simulationon the Fast Mining Method for Three-dimensional Case CT image

在三维CT病例图像的检测中,为提高纹理图像的检测精度,单纯的病例纹理二维特征在三维病例里带有较强的相似性和盲目关联性,传统的方法仅仅是从纹理特征唯一性特征的角度进行挖掘,没有考虑三维纹理变化的关联性,挖掘误差较大。提出纹理联合关联规则与自适应遗传算法相结合的病例CT三维图像快速挖掘方法,首先在纹理关联规则定义基础上通过图像降噪预处理和数据挖掘预处理,采用模板统计挖掘方法挖掘低维和高维图像纹理联合关联规则;然后引入兴趣度阈值,提出了一种根据适应度值自动调整交叉概率和变异概率的新的自适应遗传算法以达到对病例CT三维图像进行快速挖掘的目的。仿真结果表明,提出方法的识别速度和精度对比传统方法有明显提高,证实了研究方法的可行性。

In the detection of three-dimensional case CT image,in order to improve the detection accuracy of texture image,simple two-dimensional feature of case texture features in three-dimensional case has strong similarity and blind relevance.The traditional method is only from the point of view of the texture feature to make mining,but the correlation is not considered for three-dimensional texture changes,thus the error of mining is large.A fast mining method for three-dimensional case CT image based on texture combined association rule and adaptive genetic algorithm is proposed.Firstly,on the basis of the definition of texture association rules,through the image noise reduction preprocessing and data mining preprocessing,the image texture combined association rules of low dimension and high dimension is mined by using template statistics mining method：Then interestingness threshold is introduced,a new adaptive genetic algorithm of which the crossover probability and mutation probability are automatic adjusted based on the adaptive value is proposed,in order to achieve the purpose of the fast mining of the three-dimensional case CT image.The simulation results show that the recognition speed and accuracy of this method improve obviously compared to traditional methods,verifying the feasibility of the research method.