基于LGDF模型的粥样硬化斑块HRMR图像分割  

HRMR Image Segmentation Based on LGDF Model for Atherosclerotic Plaque

作　　者：严静 刘启榆[2] 周莹[2] 张顺源[2] 刘知贵[1] YAN Jing;LIU Qiyu;ZHOU Ying;ZHANG Shunyuan;LIU Zhigui(College of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010,China;Department of Radiology,Mianyang Central Hospital,Mianyang 621000,China;College of Information Engineering,Southwest University of Science and Technology,Mianyang 621000,China)

机构地区：[1]西南科技大学计算机科学与技术学院,四川绵阳621010 [2]绵阳市中心医院放射科,四川绵阳621000

出　　处：《自动化仪表》2018年第11期21-26,共6页Process Automation Instrumentation

基　　金：四川省科技计划基金资助项目(18ZDYF3046);西南科技大学研究生创新基金资助项目(18YCX005)

摘　　要：颅内动脉粥样硬化斑块可视化对颅内动脉粥样硬化疾病(ICAD)的辅助诊断与治疗具有重要意义。颅内动脉血管及斑块的分割结果将直接影响三维可视化效果。由于高分辨率核磁共振(HRMR)图像灰度分布不均匀、动脉血管结构复杂、目标边界模糊等原因,导致颅内病变动脉血管及斑块的定位与分割困难。局部高斯分布拟合(LGDF)模型可以准确分割灰度分布不均匀图像,但是对初始轮廓较敏感。针对上述问题,提出基于数学形态学修正阈值分割结果的方法,以实现初始轮廓的自动提取。利用阈值法,实现斑块的定位与预分割;通过数学形态学,进行细分割;将斑块的分割结果作为LGDF模型的初始轮廓,实现动脉血管的自动分割。对23组试验数据进行处理,斑块与动脉血管的平均分割精度分别达到93. 95%和96. 62%。试验结果表明:该方法能够准确定位和分割斑块,较好地避免了毗邻动脉血管的干扰,成功地分割出病变的颅内动脉血管。该方法在分割精度和效率上都优于单一的LGDF模型,具有较好的实用性。后期的三维可视化不仅可以用于脑血管狭窄患者的临床诊断与评估,还能进一步引导颅内斑块介入手术。The visualization of intracranial atherosclerotic plaque is of great significance in the diagnosis and treatment of intracranial atherosclerotic disease(ICAD),and the segmentation results of intracranial arteries and plaques will directly affect the three-dimensional visualization.Due to the inhomogeneity of high-resolution magnetic resonance(HRMR)images,complex structure of arteries,and blurred target boundaries,it is difficult to locating andsegmenting the intracranial lesions arteries and plaques.The LGDF model can segment the inhomogeneous images accurately,but it is sensitive to the initial contour.To solve problems above,a method based on mathematical morphology to correct the threshold segmental results is proposed to realize automatic extraction of initial contours.Threshold method is used to realize the locating and pre-segmentation of plaque;fine segmentation is performed by using mathematical morphology;finally,the plaque segmentation result is used as the initial contour of the LGDF model to achieve automatic segmentation of artery.This experiment is carried out based on 23 groups of image data,the average segmentation accuracy of plaques and arteries reaches 93.95%and 96.62%respectively.The experimental results show that the method is very efficient and accurate to locate and segment the plaque;it can avoid the intrusion of the adjacent arteries for the successful segmentation of the intracranial lesions’artery.This method is better than single LGDF model in segmental accuracy and efficiency,and it shows a good practicability.The later three-dimensional visualization can be used not only for the clinical diagnosis and evaluation of patients with cerebral vascular stenosis,but also for guiding intracranial plaque intervention.

关 键 词：颅内动脉粥样硬化斑块 图像分割 阈值法 数学形态学 LGDF模型 脑卒中 

分 类 号：TH776[机械工程—仪器科学与技术] TP391[机械工程—精密仪器及机械]