低剂量胸腔CT肺部影像的肺结节计算机辅助诊断方法研究

发布时间：2018-01-16 04:03

本文关键词：低剂量胸腔CT肺部影像的肺结节计算机辅助诊断方法研究　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：肺癌目前被称为了全世界的头号癌症,并且发病率一直在上升,在我国表现的尤为显著。各项研究表明,早期的诊断和治疗,能够提升肺癌患者的治愈率,对愈后的恢复也有重要作用。近几年来,随着各种医学成像设备的普遍使用,医学设备成像的各种先进技术也越来越受到医学工作者的重视。肺结节的定位和定性有赖于这些技术的应用。为此,胸腔CT影像的肺结节计算机辅助检测识别系统的研究引起了越来越多科研工作者和医生的重视。由于医学影像本身就存在灰度值不均匀、个体差异、伪影、噪声、边缘模糊等因素,所以有关医学影像处理的算法,要达到较高的灵敏度和精确度具有很大的难度。本文研究对象基于低剂量胸腔CT图像,在CT图像预处理、肺实质分割、疑似肺结节提取、肺结节检测这几个方面,进行了实验和研究。提出了基于低剂量胸腔CT肺部影像的结节检测方法:首先,对原始数据进行转格式换,并针对CT图像成像过程中所携带的噪声问题,引进了快速自适应加权中值滤波器(Faster Weighted Median Filter,FWMF)进行预处理;其次,为缩小目标范围及原始数据存在异常图片而导致难以基于单张切片处理得到较好分割结果的问题,提出使用自动区域生长法对多张连续预处理后的CT帧切片序列进行肺实质分割;接着,对预处理后的CT帧序列利用形态学方法进行疑似结节区域的分割;最后,根据医生提供的真阳性结节的空间位置和规律,提出了基于肺边缘距离的最小距离算法,对肺实质中的疑似病变区域进行筛选和检测,最终实现对肺结节进行分类,以及疑似结节的初步定位,从而提高肺癌早期诊断的准确率,降低假阳性率,提高阅片诊断效率和减轻放射科医生的工作量。
[Abstract]:Lung cancer is currently known as the world's number one cancer, and the incidence has been rising, especially in China. Studies show that early diagnosis and treatment can improve the cure rate of lung cancer patients. In recent years, with the widespread use of various medical imaging equipment. Various advanced techniques of medical equipment imaging have been paid more and more attention by medical workers. The location and characterization of pulmonary nodules depend on the application of these techniques. More and more researchers and doctors pay more and more attention to the computer aided detection and recognition system of pulmonary nodules in chest CT image. Because of the uneven gray value individual difference and artifact in medical image itself. Because of the noise, edge blur and so on, it is very difficult to achieve high sensitivity and accuracy in the medical image processing algorithm. The object of this study is based on low dose chest CT images. In the aspects of CT image preprocessing, lung parenchyma segmentation, suspected pulmonary nodule extraction and pulmonary nodule detection, we have carried out experiments and studies. The original data is changed to format, and the noise problem in CT image imaging is analyzed. The fast adaptive weighted median filter (Faster Weighted Median filter) is introduced for preprocessing. Secondly, in order to narrow down the target range and the existence of abnormal images in the original data, it is difficult to get better segmentation results based on single slice processing. An automatic region growth method is proposed to segment the lung parenchyma of CT frame slices after continuous preprocessing. Then, the preprocessed CT frame sequence is segmented by morphological method. Finally, according to the spatial position and rule of true positive nodules provided by doctors, a minimum distance algorithm based on lung edge distance is proposed to screen and detect suspected lesion areas in lung parenchyma. Finally, the classification of pulmonary nodules and the initial localization of suspected nodules are realized, so as to improve the accuracy of early diagnosis of lung cancer, reduce the false positive rate, improve the efficiency of X-ray diagnosis and lighten the workload of radiologists.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;TP391.41

【参考文献】