Lung nodules detection using semantic segmentation and classification with optimal features
Lung nodules detection using semantic segmentation and classification with optimal features
复制标题
DOI：
10.1007/s00521-020-04870-2
复制
发表时间：
2020-05-11
影响因子：
6
通讯作者：
Shoaib, Umar
中科院分区：
文献类型：
--
作者：
Meraj, Talha;Raul, Hafiz Tayyab;Shoaib, Umar
肺癌是一种致命的疾病，如果没有在早期诊断。然而，由于其结节的形状和大小，肺癌的早期检测是一项具有挑战性的任务。放射科医生使用自动化工具来获得更精确的意见。由于健康和不健康组织之间的形状相似性，受影响的肺结节的自动检测是复杂的。多年来，已经开发了一些专家系统，帮助放射科医生有效地诊断肺癌。在这篇文章中，我们提出了一个框架，以精确地检测肺癌分类的良性和恶性结节。使用公开可用数据集的子集来测试所提出的框架，即，肺部图像数据库联盟图像采集（LIDC-IDRI）。我们在预处理阶段应用滤波和噪声去除。此外，自适应阈值技术（大津）和语义分割用于准确地检测不健康的肺结节。总体而言，13个结节特征提取使用主成分分析算法。此外，四个最佳的功能选择的基础上的分类性能。在分类阶段，采用了9种不同的分类器进行实验。实证分析表明，该系统优于其他技术，并提供99.23%的准确率使用logit提升分类。
Lung cancer is a deadly disease if not diagnosed in its early stages. However, early detection of lung cancer is a challenging task due to the shape and size of its nodules. Radiologists use automated tools for more precise opinion. Automated detection of the affected lung nodules is complicated because of the shape similarity among healthy and unhealthy tissues. Over the years, several expert systems have been developed that help radiologists to diagnose lung cancer effectively. In this article, we have proposed a framework to precisely detect lungs cancer to classify the benign and malignant nodules. The proposed framework is tested using the subset of the publicly available dataset, i.e., the Lung Image Database Consortium image collection (LIDC-IDRI). We applied filtering and noise removal in the pre-processing phase. Furthermore, the adaptive thresholding technique (OTSU) and the semantic segmentation are used to accurately detect the unhealthy lung nodules. Overall, 13 nodules features have extracted using principal components analysis algorithm. In addition, four optimal features are selected based on the classification performance. In the classification phase, 9 different classifiers are employed for the experimentation. Empirical analysis shows that the proposed system outperformed other techniques and provides 99.23% accuracy using a logit boost classifier.