【作者】 王晶；

【导师】 范立南；

【作者基本信息】 沈阳大学 ， 控制理论与控制工程， 2010， 硕士

【摘要】 随着各种医学影像设备的应用,大大的提升了临床诊断和医学研究的速度和质量,医学影像的各种后处理技术也越来越受到研究者的重视。在众多的医学影像设备之中,CT计算机断层扫描以其图像分辨率高,解剖组织关系明确和获取方式对病人伤害小的特点逐渐成为了主要的医学影像设备之一。医学图像分割是医学图像各种后续处理的必要基础,在医学图像处理中占有重要的地位。在肺部疾病诊断中,肺实质分割为计算机辅助诊断工具设计、肺结节检测、肺功能评估和肺部图像三维重建提供了条件。因此,准确的分割出完整的肺实质对临床诊断和后续处理及研究都有着重要的意义。本文的研究目的在于将肺实质从背景中分离出来,对于由于病变造成的左右肺粘连的现象,成功的将左右肺实现分离并且尽量保持其边缘,在众多的病变肺部图像中,存在一些肺结节处于靠近肺部边缘的位置,这部分结节很容易在分割中被误去除,造成有价值信息的丢失,这是对病变诊断非常不利的,文本通过增加一个边缘修补的处理来降低这种遗漏的产生,为后续操作提供一个完整准确的肺实质分割结果。针对以上所述问题,本文主要做了以下研究,提出了一个完整的基于多种算法融合的肺实质分割策略。(1)基于肺部图像具有目标与背景具有较高的对比度的特点,选取简单有效的阈值法对图像进行初分割,由于阈值法容易受到噪声的干扰,所以本文在采用阈值法分割之前,对原始图像进行去噪以此改善阈值分割的质量。(2)对多种阈值选取方法做了比较研究。(3)在剔除肺实质图像中气管的步骤中,根据肺部CT图像的解剖结构特性实现了种子点自动选取并在此基础上进行区域生长。(4)将连接区域标记方法引入左右肺连接情况作出判断,为后续的分离左右肺处理提供了范围,提高了分割的效率。(5)运用迭代的形态学复合操作对左右肺实现分离,提高了分割的准确性。(6)针对肺部边缘上的缺口,设计了一种端点检测方法应用于边缘的修补。通过用本文方法对多组临床图像进行仿真实验,将处理结果与医生手工分割结果进行对比,采用平均距离来评价本文方法,通过实验可以证明应用本文方法可以实现将肺实质从背景中分离出来,得到分开的左右肺,并且能够修补肺部边缘缺口,通过本方法可以得到一个与手工分割结果相近的、令人满意的结果用于后续研究和处理。更多还原

【Abstract】 The application of many kinds of medical image technology improves the speed and quantity of clinical diagnosis and medical research it has been given more and more attention by researchers. The image of computerized tomographic scanning technology (CT) become one of the most important technology because the character of high resolution and clear relationship between organizations. It is also high light that there is few heart to the patient in the process of image acquisition.Clinical image segmentation is the basis of the following process it takes a important place in clinical image process technology. In the clinical diagnosis process the result of lung segmentation make the CAD design the detection of lung nodules the evaluation of lung function and the rebuild of 3D lung image possible. It is every important to get a exact lung image to provide the clinical diagnosis and research a reliable basis.The aim of this research is to segment the lung field from the background. To divide the connected right and left lung which are coursed by pathology. There are many nodules lie near the lung boundary. This kind of nodules will be set to background pixel in the threshold process and the useful information will loose. It is un-expectant. This paper adds a lung boundary repair process to solve the problem to get a more accurate lung boundary for the following steps.Aiming at the problems above the following work has been scheduled and a segmentation strategy based on compositive arithmetic has been approved.(1) The threshold method is base on the gray value information it is easier and more efficient to segment the lung field from the around issue because the high contrast between the two parts. But the method is easy to be impacted by noise so a filter has been used before threshold method to improve the quality of the initial image.(2) Apply the optimal threshold to get the segmentation after the research on several thresholds. (3) To segment the air-way which rely in the lung field through a region growing method in which the seed can be selected automatically.(4) To introduce the connective region labeling method into the estimate of the status of connection of left and right lung region. The results of estimate provide a range of the following process.(5) Design an iteratively morphological multi-operation to separate the right and left lung to get a more accurate result of segmentation.(6) Provide a extreme point detection method to repair the gap of lung boundary.The results have been tested by processing the clinical image sets by emulator. We compare the result of the method in this paper to the result of manual method by image analyst. The result indicates the method in this paper can be used to segment the lung region from background and get the right and left lung separately. The gap on the lung boundary can be repaired by this method to get a result which is similar with the result of manual method can be used in the following research and process.更多还原