基于改进K-Means的腹内脂肪自动定量检测算法

发布时间：2018-08-11 20:12

【摘要】：检测肥胖病人腹部脂肪的分布及含量情况,判定腹型肥胖的种类,对评估和治疗糖尿病和心血管代谢等相关疾病有重要的临床价值.通过分析人体腹部磁共振(MR)图像中脂肪成像的特点,提出一种无监督的腹部脂肪自动检测算法.该算法运用SLIC算法对腹部磁共振图像进行预处理,生成超像素;然后用泛洪填充算法进行背景剔除,再将改进的K-means算法用于脂肪区域与非脂肪区域以及皮下脂肪与内脏脂肪的自动分割;最后基于分割结果实现对腹部脂肪的定量分析.实验结果表明,文中算法能精确地检测出腹部脂肪的含量,并能够区分脂肪的类别,相比以往的半自动或全自动算法,其准确率得到了有效的提高.
[Abstract]:Detecting the distribution and content of abdominal fat in obese patients and determining the types of abdominal obesity have important clinical value in the evaluation and treatment of diabetes mellitus and cardiovascular metabolism and other related diseases. An unsupervised automatic abdominal fat detection algorithm is proposed by analyzing the characteristics of fat imaging in human abdominal magnetic resonance (MR) images. The algorithm uses SLIC algorithm to preprocess the abdominal magnetic resonance image to generate super-pixel, and then uses flooding fill algorithm to remove the background. Then the improved K-means algorithm is applied to the automatic segmentation of fat region and non-fat region, subcutaneous fat and visceral fat. Finally, the quantitative analysis of abdominal fat is realized based on the segmentation results. The experimental results show that the proposed algorithm can accurately detect the abdominal fat content and can distinguish the type of fat. Compared with the previous semi-automatic or automatic algorithm, the accuracy of the algorithm has been improved effectively.
【作者单位】：上海交通大学计算机科学与工程系;中国科学院软件研究所计算机科学国家重点实验室;澳门大学科技学院电脑及资讯科学系;
【基金】：国家自然科学基金(61572316,61133009) 国家高技术研究发展计划(2015AA015904) 浙江大学CAD&CG国家重点实验室开放课题(A1401) 中国科学院自动化研究所模式识别国家重点实验室开放基金澳门大学科研基金(MYRG150(Y3-L2)/FST11/WW)
【分类号】：R589.2;TP391.41

【相似文献】