基于CT图像的肺癌前期辅助检测研究

发布时间：2018-05-04 00:21

本文选题：肺结节 + LIDC-IDRI　；参考：《电子科技大学》2017年硕士论文

【摘要】：早发现,早治疗,早诊断是提高肺癌患者生存率的关键。但是不容乐观的是只有15%的肺癌在早期被发现,这是由于早期肺癌在临床上的表现并不明显,患者通常察觉不到任何症状。因此如何做到早发现早治疗是关键所在,肺癌早期计算机辅助检测有着重大的研究意义。首先介绍了预处理相关的内容,包括去噪,增强和肺部掩膜的分割;然后介绍了本文的重点内容肺结节的分割,特征的提取,特征的选优和分类。具体来说,本文的主要研究内容如下所示:1)给出预处理中去噪,增强和肺部分割的具体算法和流程,为后续核心部分做好基础工作。其中去噪使用了中值滤波去噪,增强使用的是基于高斯核函数的增强,肺部分割部分给出了一个基于阈值的逻辑框架。2)给出了肺结节分割的逻辑框架和具体算法,特征提取的计算公式或者思想,特征选优采用SFS方法,阳性阴性的分类采用SVM分类器。最后对实验结果做出了具体的分析和总结。为了验证本文框架的有效性,采用了公开数据集LIDC-IDRI数据库来进行效果的对比操作。对比效果表明,该论文提出的算法具有较优的效果。
[Abstract]:Early detection, early treatment, early diagnosis is the key to improving the survival rate of lung cancer patients. But it is not optimistic that only 15% of the lung cancer is found early in the early stage, which is due to the early detection of lung cancer in clinic, and the patient is usually not aware of any symptoms. Therefore, how to find early treatment is the key to early lung cancer. Auxiliary detection is of great significance. First, it introduces preprocessing related contents, including denoising, enhancement and segmentation of lung mask, and then introduces the segmentation of lung nodules, feature extraction, feature selection and classification. The main contents of this paper are as follows: 1) denoising in preprocessing. The concrete algorithms and processes of enhancement and lung segmentation are done for the following core parts. The denoising is used to denoise with median filtering, the enhancement is based on the enhancement of the Gauss kernel function, the lung segmentation part gives a logic framework based on the threshold based on the logic framework and the specific algorithm of the lung nodule segmentation. The feature extraction formula or thought, features selected to adopt the SFS method, the positive negative classification adopts the SVM classifier. Finally, the experimental results are analyzed and summarized. In order to verify the validity of the framework, the open data set LIDC-IDRI database is used to compare the effect. The proposed algorithm has a better effect.

【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R734.2;R730.44

【相似文献】