Current Issue Cover
医学图像分割算法的评价方法

张石1, 董建威1, 佘黎煌1(东北大学信息科学与工程学院,沈阳 110004)

摘 要
对医学图像分割算法的客观评价是推进算法在临床上得到应用的关键。针对目前对医学图像分割方法的研究较多,而对分割算法的评价方法的研究却很少的问题,提出了一种判断和比较医学图像分割算法优劣的评价方法。首先对现有的几种评价方法进行了综述,并总结出了一套评价系统。可靠性、精确性、区域统计特性和效率是评价一个分割方法的4个重要参数,结合医学图像分割分别对它们的定义进行了说明。这些参数互相影响,评价分割算法时必须权衡这些指标,根据不同的应用背景赋予它们不同的权重。此外,还介绍了如何建立医学图像分割金标准数据库的方法。最后,通过Insight Toolkit(ITK)软件包中的两个算法例子,结合脑白质分割的医学背景,演示了如何利用本文评价系统来对这两种分割算法做出比较。实验结果表明,该评价方法可行,比较结果具有合理性。该研究为医学图像分割算法的评价提供了科学合理的方法,同时也指出了推动医学图像分割算法在临床上应用所应解决的问题。
关键词
The Methodology of Evaluating Segmentation Algorithms on Medical Image

()

Abstract
Objective evaluation of medical image segmentation algorithms is one of the important steps toward establishing validity and clinical applicability of an algorithm. Since there are a large number of articles presenting segmentation methods on medical image, with few studying the evaluation methods on their performance, this paper presents an evaluation method for different segmentation algorithms. The author first gives a survey of several available evaluation methods and presents a systematic summary. Reliability, precision, region statistical characteristics and efficiency are the four most important metrics. The definitions of them are then described based on the image segmentation process. For comparison, weights should be added to these metrics according to the application. Moreover, the author also presents a method on how to construct gold standard of medical images. At last, with the task of brain white matter segmentation, the author demonstrates how to make use of the proposed evaluation method to compare two segmentation algorithms in insight toolkit(ITK). The experiment results show that this method is practical and reasonable. This study gives a scientific method for the evaluation of segmentation algorithms on medical image. Meanwhile, it points out the problems to be solved before the segmentation algorithms could be put into use in clinic.
Keywords

订阅号|日报