基于视觉注意的肺结节显著性区域分割方法  

A Segmentation Method for Lung Nodules Based on Visual Attention

作　　者：韩贵来[1] 詹何庆[1] 邓丹琼[2] HAN Gui-lai1, ZHAN He-qing1, DENG Dan-qiong2 （1.Faculty of Medical Informatics, Hainan Medical College, Haikou 571199,China;2.The First Affaliated Hospital of Hainan Med- ical College, Haikou 571199,China）

机构地区：[1]海南医学院医学信息学院,海南海口571199 [2]海南医学院第一附属医院,海南海口571199

出　　处：《电脑知识与技术》2017年第6期165-167,共3页Computer Knowledge and Technology

基　　金：国家青年科学基金项目（81501557）;海南省自然科学基金项目（20156224）

摘　　要：针对肺实质图像中肺结节可能存在多发的实际情况,本文改进了ITTI模型中“赢者全拿”的策略,对多个显著性区域进行分析;针对基于视觉注意所检测到的肺结节显著性区域与肺实质图像中肺结节的实际位置并不完全一致的问题,该文先在肺实质图像中找出那些显著区域所对应的区域,将这些区域分割成各个连通区域,在每个连通区域中找到一个灰度值最大的点作为种子点,并在肺实质图像中进行区域生长,最终得到可疑肺结节图像。相比直接将显著性区域作为可疑肺结节的方法,本文方法分割得到的可疑肺结节更加准确,更利于后续的肺结节特征提取和识别,对肺癌的早发现,早诊治具有重要意义。For there may be multiple pulmonary nodules in the lung parenchyma images, this paper improved the ＂winner take all＂ strategy in the ITTI model to analysize multiple saliency regions. Aiming at the problem that the significant area of lung nodules detected by visual attention is not completely consistent with the actual location of pulmonary nodules in the lung parenchyma images, this paper found out the regions corresponding with the salient regions in the lung parenchyma images firstly, These regions are divided into each connected region. Then, the maximum value of gray value in each connected region was found out. These maximum values of gray value were used as seed points for regional growth in lung parenchyma images. Finally the suspected pulmonary nodule image was obtained. Compared with the method of directly using the significant region as a suspected pulmonary nodule, this method is more accurate and more suitable for the feature extraction and recognition of pulmonary nodules, which is of great significance for early detection and early diagnosis and treatment of lung cancer.

关 键 词：视觉注意机制 肺结节 显著图 图像分割 区域生长 

分 类 号：TP391[自动化与计算机技术—计算机应用技术;自动化与计算机技术—计算机科学与技术]