光学技术, 2020, 46 (2): 230, 网络出版: 2020-07-16  

传统解剖学特征与深度学习相结合的肺叶分割算法

Segmentation of Lung Lobes Based on traditional anatomical features and deep learning
作者单位
上海理工大学 医疗器械与食品学院, 上海 200093
摘要
CT图像中肺叶位置的确定对于肺部疾病的准确定位以及定性定量分析具有重要意义。为了提高肺叶自动分割准确率,提出了一种结合气管,血管等传统解剖学特征以及深度学习的肺叶分割算法。对原始图像进行预处理,获取肺实质、气管、血管以及基于深度学习网络的肺裂分割结果; 整合来自多个解剖结构的信息生成分水岭分割所需成本图像; 通过基于深度学习网络的肺叶粗分割结果,获取肺叶标记区域; 执行基于标记的分水岭分割,实现肺叶的自动分割。选取了来自上海市肺科医院的20例含有肺部疾病患者的CT图像对该方法进行验证,最终的Jaccard相似性系数为92.4%。实验结果表明方法具有较高的肺叶分割精度,并且具有较强的鲁棒性。
Abstract
The location of lung lobes in CT images is of great significance for accurate localization of lung diseases and qualitative and quantitative analysis. In order to improve the accuracy of lung lobes segmentation, an algorithm combining deep learning with traditional anatomical features such as trachea and vessels is proposed. Firstly, the original image is preprocessed to obtain the segmentation results of lung parenchyma, trachea, blood vessel and lung fissure based on deep learning. Then, the cost image of watershed segmentation is generated by integrating the information from multiple anatomical structures. Last, the lung lobe markers are obtained through the segmentation results based on deep learning. The automatic segmentation of lung lobes is realized by marker-based watershed segmentation. CT images of 20 patients with lung diseases from Shanghai Pulmonary Hospital were selected to validate the method. The Jaccard similarity coefficient was 92.4%. The experimental results show that the method has high segmentation accuracy and strong robustness

高磊, 段辉宏, 聂生东. 传统解剖学特征与深度学习相结合的肺叶分割算法[J]. 光学技术, 2020, 46(2): 230. GAO Lei, DUAN Huihong, NIE Shengdong. Segmentation of Lung Lobes Based on traditional anatomical features and deep learning[J]. Optical Technique, 2020, 46(2): 230.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!