一种基于k-means聚类和半监督学习的医学图像分割算法

A medical image segmentation algorithm based on k-means clustering and semi-supervised learning

医学图像分割是计算机视觉和图像处理领域近年来研究的热点问题之一。一种基于k-means聚类和半监督学习的医学图像分割新算法被提出。在k-means聚类模型中,相似度函数是关系到聚类效果好坏的关键因素。所使用的相似度函数通过基于side-information的半监督学习方法来确定;确定后的相似度函数又被运用回k-means聚类模型中来实现对医学图像的分割。为了检验该算法效果,脑部肿瘤患者的磁共振图像被运用在实验中。分析结果表明:该算法在本文所采用的实例中能获得优于传统算法的分割效果。

Medical Image Segmentation is one of the most popular applications in contemporary computer vision and image processing fields. A novel algorithm based on k-means clustering and semi-supervised learning was presented in this study. The similarity function is one of the most important factors in clustering algorithms. It was determined via a semi-supervised learning process based on side-information in k-means method in our study. The learned similarity function was thereafter incorporated in the clustering model to differentiate tumor pixels from non-tumor pixels. In order to evaluate the presented algorithm,experiments incorporating MRI from patients with brain tumor were conducted as well. The superiority of the introduced algorithm over several existing ones was demonstrated therein.