结合改进型主动外观模型与马尔科夫随机场的椎间盘核磁共振图像分析  

Spine Disc MR Image Analysis Using Improved Independent Component Analysis Based Active Appearance Model and Markov Random Field

作　　者：郝世杰[1] 詹曙[1,3] 蒋建国[1] 李鸿[2] Ian Ross 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009 [2]安徽医科大学第一附属医院骨科,合肥230022 [3]London Health Science Center, London Ontario Canada N6A 5A5

出　　处：《生物医学工程学杂志》2010年第1期6-9,15,共5页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(60474035);教育部博士点基金项目资助(20060359004);教育部留学归国人员科研启动基金资助项目(413117)

摘　　要：针对目前腰部医学图像处理较少涉及分割并量化分析软组织的现状,提出一种腰部核磁共振图像(magnet-ic resonance imaging,MRI)中椎间盘分割及定量分析的算法:首先用改进的基于独立分量分析的主动外观模型(in-dependent component analysis based active appearance model,ICA-AAM)分割椎体,并基于最小描述长度(minimumdescription length,MDL)准则获得腰部弯曲曲线;利用上述分割结果,结合椎间盘成像特点和局部图像灰度投影,进一步实现基于马尔科夫随机场(Markov random field,MRF)的椎间盘快速无监督分割;通过分割结果判断并定量计算椎间盘突出程度。实验表明该算法快速有效,可以辅助医生诊断、治疗腰椎间盘突出等疾病。As there are not many research reports on segmentation and quantitative analysis of soft tissues in lumbar medical images,this paper presents an algorithm for segmenting and quantitatively analyzing discs in lumbar Magnetic Resonance Imaging（MRI）.Vertebrae are first segmented using improved Independent component analysis based active appearance model（ICA-AAM）,and lumbar curve is obtained with Minimum Description Length（MDL）;based on these results,fast and unsupervised Markov Random Field（MRF） disc segmentation combining disc imaging features and intensity profile is further achieved;finally,disc herniation is quantitatively evaluated.The experiment proves that the proposed algorithm is fast and effective,thus providing doctors with aid in diagnosing and curing lumbar disc herniation.

关 键 词：医学图像处理与分析 腰部核磁共振图像 基于独立分量分析的主动外观模型 马尔科夫随机场 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]