甲状腺结节钙化特征的自动提取方法研究.PDF

28 Thesis 论 著 甲状腺结节钙化特征的自动提取方法研究 李 涛，李怡勇，米永巍，丁明跃，张 冀 [摘要] 目的 提出一种甲状腺结节钙化特征的自动提取方法，从而提高应用钙化特征对甲状腺结节进行鉴别诊断 的可重复性和准确性。方法 首先采用基于GVF Snake 模型的分割算法和钙化灶的图像识别算法完成甲状腺结节 钙化特征的自动提取。然后，在对 30 例甲状腺结节的回顾性研究中，通过受试者工作特征曲线（receiver operating characteristic curve ，ROC ）分析评估所提取的钙化指数（calcification index ，CI ）对甲状腺结节的鉴别诊断价值。结果 提取的钙化指数在良、恶性结节分类中具有一定的准确性（ROC 曲线下面积为0.720）。以约登指数最大的分界点作 为最佳的诊断界限点，准确性可达80%。结论 该方法能有效地提取甲状腺结节的钙化特征，所提取的钙化指数对 甲状腺恶性结节的超声临床诊断具有一定的辅助作用。 [关键词] 甲状腺结节；钙化特征；自动提取；超声图像 [中国图书资料分类号] R318 ；R445.1 [文献标志码] A [文章编号] 1003-8868 渊2015 冤12-0028-03 DOI 10.7687/J.ISSN1003-8868.2015.12.028 Automatic feature extraction algorithm for thyroid nodules calcification 1,2 2 2 1 2 LI Tao , LI Yi-yong , MI Yong-wei , DING Ming-yue , ZHANG Ji 渊1. School of Life Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China; 2. Department of Medical Engineering, Wuhan General Hospital of Guangzhou Military Area Command, Wuhan 430070, China冤 Abstract To propose a kind of thyriod nodule calcification feature automatic extraction method so as to im鄄 prove repeatability and accuracy when using calcification feature to differentiate thyroid nodules. The automatic extraction algorithm of the calcification feature of thyroid nodules was fulfilled based on GVF snake model segmentation and image recognition method of calcifications. Then, in a retrospective study of 30 cases of thyroid nodules, the value of extracted calcification index (CI) in the differential diagnosis of t