基于CT图像的肺部肿瘤识别-电子与通信工程专业论文.docxVIP

上海交通大学硕士学位论文 上海交通大学硕士学位论文 万方数据 万方数据 - 77 - 基于 CT 图像的肺部肿瘤识别 摘 要 计算机辅助诊断是通过影像学、图像处理等技术，结合计算机的 分析、建模和模式识别等方法，帮助医师发现病灶，提高诊断的效率 和准确率。本文搭建了一个载入连续 CT 图像，利用形态特征分辨可自 动识别肺部肿瘤的计算机辅助诊断（CAD）系统，并提出形状特征矩 和三维细化两种新的肿瘤识别方法。此系统通过三维几何形状信息来 识别肿瘤，包括一系列的处理，如插值、分割、疑似区域搜索和识别。 首先，将连续的 CT 图像插值成在三个维度上等距的立体图形，以便于 恢复肿瘤原有的三维形态。其次，对肺实质区域进行分割，使整个灰 度图的分布更有利于进行下一步的识别操作。分割的步骤分为最大类 间方差 OTSU 算法选取自适应阈值，连通域处理及数学形态学法如膨 胀、腐蚀、开运算和闭运算的组合。通过灰度分布衰减得到阈值和区 域增长选取出感兴趣区域（ROI）作为潜在的肿瘤目标。最后，利用矩 不变理论分辨 ROI 得出真正的肿瘤。并运用三维细化的方法对三维的 CT 图像组进行处理，在较规则的肿瘤识别上取得了很好的效果。临床 CT 扫描数据的实验表明以上提到的方法能够在肿瘤检测上有良好的表 现。在所试验的案例中此系统表现出 77.55%肿瘤检出率，在可接受范 围内。此系统可以区分出肿瘤和其他组织如支气管，器官等，并提供 友好的用户界面，支持切片间的自由浏览，是辅助诊断的有效方法。 关键词：计算机辅助诊断、肿瘤识别、肺实质分割、CT 图像处理 I RECOGNITION OF LUNG NODULES ON CT SCANS ABSTRACT Computer-Aided Diagnosis is proposed for help radiologists increase efficiency and accuracy when finding focuses by combining HYPERLINK dict://key.0895DFE8DB67F9409DB285590D870EDD/iconography iconography, image processing, computer analysis, modeling and pattern recognition. This paper describes a Computer-Aided Diagnosis (CAD) system for automatic pulmonary nodules detection on serial CT scans based on shape features. It also proposes two novel recognition methods which are moment invariant features and 3D thinning. The system recognizes nodules by 3D geometric information through the process of interpolation, segmentation, suspicious area searching and recognition. Firstly, the serial CT images are interpolated to equal scales in each dimension, in order to recover the original 3D shape of nodules. Secondly, pretreatment is implemented to segment the lung parenchyma region, which is beneficial to the next step recognition with a simple distributed grey level image. The segmentation process includes maximum between-class variance method, OTSU algorithm, to select thresholds adaptively, connected-area judgment and mathematical morphology such as the combine of open algorithm, close algorithm, dilation and erosion. Thirdly, detect objects called regions