文章摘要

文章信息/Info

Title:: Ultrasound image segmentation of thyroid nodule based on CV-RSF algorithm

作者:: 邵蒙恩1; 严加勇2; 3△; 崔崤峣3; 于振坤4; 1.上海理工大学医疗器械与食品学院，上海 200093；2.上海健康医学院附属周浦医院，上海 201318；3.中科院苏州生物医学工程技术研究所，苏州 215163；4.南京同仁医院，南京 211102

Author(s):: SHAO Mengen1; YAN Jiayong2; 3; CUI Xiaoyao3; YU Zhenkun4; 1.University of Shanghai for Science and Technology, Shanghai 200093,China;2.Zhoupu Hospital affiliated to Shanghai University of Medical&Health Science, Shanghai 201318;3. Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, Suzhou 215163,China; 4. Tongren Hospital of Nanjing, Nanjing 211102,China

Keywords:: CV model; RSF model; CV-RSF model; Thyroid nodule; Ultrasound images; Segmentation

摘要:: 结合CV模型和RSF模型，建立CV-RSF模型，实现甲状腺结节超声图像半自动分割算法。基于CV模型的全局信息对图像粗分割，并以粗分割结果作为RSF模型的初始轮廓，然后再利用RSF模型的局部信息对病灶实现最终的分割。根据设定的两组不同初始轮廓，分别利用RSF模型和CV-RSF模型对病灶分割。结果表明，CV-RSF模型解决了RSF模型对初始轮廓敏感的问题，而且通过重叠率的对比，CV-RSF模型分割更准确。对比RSF模型，CV-RSF模型实现的甲状腺结节超声图像半自动分割算法，更加有效、准确。

Abstract:: To establish a CV-RSF model based on CV model and RSF model, and to realize semi-automatic segmentation algorithm of thyroid nodules. Based on the global information of the CV model, the image was roughly segmented, and the rough segmentation result was used as the initial contour of the RSF model. Then the local information of the RSF model was used to achieve the final segmentation of the lesion. According to the set of different initial contours, the lesions were segmented by RSF model and CV-RSF model respectively. The results showed that the CV-RSF model solved the problem that the RSF model was sensitive to the initial contour, by comparing the overlap ratio, the CV-RSF model segmentation was more accurate.Compared with the RSF model, the semi-automatic segmentation algorithm of thyroid nodules with CV-RSF model is more effective and accurate.