题名HIFU治疗中的超声图像分割算法研究
作者闫晟
学位类别博士
答辩日期2007-06-08
授予单位中国科学院声学研究所
授予地点声学研究所
关键词高强度聚焦超声 超声图像分割 图像增强 边缘增强 水平集 马尔科夫随机场 小波变换
其他题名Ultrasound Image Segmentation in HIFU Therapy
学位专业信号与信息处理
中文摘要高强度聚焦超声(HIFU)技术作为一种无侵害的、安全、有效的肿瘤治疗手段,近年来取得了快速的发展,当前HIFU设备对多种恶性肿瘤的治疗以达到临床应用阶段。在HIFU设备中,通常利用B超设备对治疗区进行实时监控和引导,并对治疗效果进行实时评估。B超图像处理算法的水平,将直接影响到整套HIFU设备的性能及其治疗效果。本文在此背景下,对HIFU治疗中的超声处理算法,特别是超声图像分割算法,进行了一些研究工作,并将部分算法应用于当前的HIFU应用研究中。 超声图像由于自身成像机制问题,图像质量相对较差,直接进行图像分割的准确性要受到很大影响。因此在进行分割处理前,应先对超声图像进行预处理。本文在对几种当前常用超声图像增强算法进行了介绍和实现。随后,提出在小波收缩算法中,通过预分割的过程从图像中得到背景并进而自动获得小波收缩方法的降噪阈值。此方法和利用传统阈值的降噪效果相比,在去除散斑噪声方面更为有效。随后本文利用FCM和模糊增强的方法提取低对比度超声图像的边缘信息,此方法在后面章节的图像分割算法中起到了重要的作用。 水平集是近年来新兴的图像分割算法,在对超声图像的分割中有着很好的效果。本文对基于区域、基于边缘以及基于内外能量的三种水平集分割方法进行了研究,提出了小波收缩与Chan-Vese水平集分割方法相结合的超声图像分割方法,使得分割的结果更为准确,并明显的提高了算法的分割速度。之后,对于包含弱边缘的超声图像,本文改进了水平集演化的速度函数,解决了零水平集曲线跳过目标边缘的问题。 由于成像方式的原因,超声图像中存在失真场,影响分割算法的结果。利用基于马尔可夫随机场的最大后验概率分析,可以对图像中的失真场进行最佳估计,并实现对图像的分割。传统方法中,估计随机场和分割图像需要迭代的进行,运算量大。本章提出利用超声图像的先验特征对图像失真场进行一次估计,随后依据估计的随机场对超声图像进行分割。此方法即能自动的完成分割,也保证得到的分割结果准确有效。 最后,本文利用图像分割的方法在HIFU离体实验中,实现了对治疗过程的引导和监控。并对治疗后离体组织超声图像的纹理特征值进行了提取,所得特征值经加权处理可以对超声剂量进行正确反映。
英文摘要As a new non-invasive, safe and effective extracorporeal technique, High-intensity focused ultrasound (HIFU) has achieved rapid development these years. Now, HIFU systems have been successfully used in clinical application in the treatment of various malignant tumors. Conventional B-mode ultrasound is currently a leading imaging modality for monitoring, guiding and measuring the lesion in HIFU therapies for its advantages in cost, portability, and real-time implementation. The ultrasound image processing algorithms affect the performance of HIFU systems directly. In this background, the paper is based on some research of ultrasound image processing, especially ultrasound image segmentation. And the algorithm implementation in HIFU therapy is included. For the artifacts of B-mode ultrasound imaging such as speckle noise, shadows which make the segmentation task complicated, image enhancement is a pre-process of the segmentation. This paper introduces some common ultrasound image enhancement algorithms and their performance. Then, we present a new wavelet shrinkage method that getting the de-noising threshold from the image background which is obtained by a pre-segmentation algorithm automatically. This approach is more efficient than other ways in speckle noise compressing. Extracting grads information of low-contrast ultrasound image by FCM cluster and fuzzy enhance method plays an important role in image segmentation algorithm. Level set is an emerging algorithm for image segmentation and can achieve better result when processing ultrasound image. The paper studied three level set methods which are based on area, margin and both area and margin, and then introduced a new method of ultrasound image segmentation which combines wavelet shrinkage and Chan-Vese level set method. Besides, the paper improved the speed function of level set evolution when processing weak boundary ultrasound images. And solve the problem that zero level set across the weak boundary. To segment the ultrasound with distortion field, this paper introduces a new way to estimate the distortion field using the prior information of the ROI. This method makes the segmentation processing using MAP-MRF algorithm to segment ultrasound images without iteration steps. The method can also improve the accuracy of initialization and make the segmentation process automated. Finally, monitoring and guiding the treatment are achieved by using the image segmentation in HIFU vitro test. Weighting the eigen value extracted from ultrasound images of vitro organize, the result can reflect the HIFU energy properly.
语种中文
公开日期2011-05-07
页码117
内容类型学位论文
源URL[http://159.226.59.140/handle/311008/214]  
专题声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式
GB/T 7714
闫晟. HIFU治疗中的超声图像分割算法研究[D]. 声学研究所. 中国科学院声学研究所. 2007.
个性服务
查看访问统计
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。


©版权所有 ©2017 CSpace - Powered by CSpace