基于一致度、覆盖度和包含度的感兴趣区域不一致性决策算法

发布时间：2018-01-08 14:18

本文关键词：基于一致度、覆盖度和包含度的感兴趣区域不一致性决策算法　出处：《计算机应用》2015年10期 　论文类型：期刊论文

【摘要】：医学影像感兴趣区域(ROI)的噪声和疾病误判是一个典型的不一致性决策问题,同时也是困扰临床诊断的一个难题。针对这个问题,基于宏观与微观结合、全局与局部相结合的思想,提出了基于一致度、覆盖度和包含度的磁共振成像(MRI)前列腺肿瘤ROI不一致决策算法(Ito C-CIC)。首先提取MRI前列腺肿瘤ROI的高维特征,得到完备不一致决策信息表;然后通过计算不一致度找到不一致样本所在的等价类;再计算不一致等价类的覆盖度和包含度得到Score值,利用Score值筛选不一致样本,实现不一致性决策向一致性决策的转换;最后通过典型算例、UCI数据集和实验提取的前列腺肿瘤ROI特征构成的不一致决策信息表等进行验证。实验结果表明,所提算法能有效地找到并筛选掉不一致性样本。
[Abstract]:Medical image region of interest (ROI) of the noise and the disease is a misjudgment of inconsistency typical problems, and it is also a difficulty problem in clinical diagnosis. In order to solve this problem, based on the combination of macro and micro, local and global combining ideas is proposed based on the degree of consistency, cover and contain magnetic resonance imaging of prostate cancer (MRI) ROI inconsistent decision (Ito C-CIC) algorithm. Firstly, extract the high dimensional features of MRI prostate cancer ROI, to obtain completely inconsistent decision information table; then through calculating the degree of inconsistency found inconsistent with the kind of equivalence class is inconsistent; then calculate the equivalence class coverage and include get the Score value, the Score value were inconsistent samples, transform the inconsistency to consistency of decision-making; finally, through typical examples, ROI features of prostate tumor UCI data set and the experimental extraction The experimental results show that the proposed algorithm can effectively find and screen out inconsistency samples.

【作者单位】：宁夏医科大学理学院;陕西师范大学计算机学院;宁夏医科大学总医院核医学科;
【基金】：国家自然科学基金资助项目(81160183,61561040) 宁夏自然科学基金资助项目(NZ12179,NZ14085) 宁夏高校科研项目(NGY2013062) 陕西省语音与图像信息处理重点实验室开放课题资助项目(SJ2013003) 宁夏医科大学特殊人才项目(XT2011004)
【分类号】：TP391.41;R737.25
【正文快照】： 0引言前列腺癌是欧美国家男性常见的恶性肿瘤,发病率居所有恶性肿瘤的第二位[1],我国前列腺癌的发病率远低于西方国家,但由于人口老龄化和前列腺特异抗原(Prostate SpecificAntigen,PSA)筛查的普及,近10年来前列腺癌的发病率呈直线上升的态势[2]。磁共振成像(Magnetic Resonan

【参考文献】