基于MRI的脑肿瘤分割技术研究进展

The Research Progress of Brain Tumor Segmentation Technology Based on MRI

目的:对脑肿瘤的准确分割在临床上具有重要应用价值,但由于脑肿瘤结构复杂、边界模糊且与正常脑组织混叠在一起,因此,要实现对脑肿瘤的正确分割非常困难。为给相关研究者提供有益参考,本文对基于MRI的脑肿瘤分割技术研究进展进行了探讨。方法:查阅国内外相关资料的基础上,对现有的基于MRI的脑肿瘤分割方法进行分类,然后对近年来基于MRI的脑肿瘤分割技术及其研究进展进行了比较详细的综述和讨论,并介绍了脑肿瘤分割算法的评价方法,最后对脑肿瘤分割方法的发展趋势进行展望。结果:基于MRI的脑肿瘤分割方法主要包括:区域生长法、聚类分割方法、基于形变模型的分割方法、基于形态学分水岭的分割方法、图谱匹配、多谱MR图像分割和基于异常检测的分割方法等。结论:基于MRI的脑肿瘤分割方法将向全自动、实时、准确的分割方向发展,并有效地结合多种分割方法,综合利用多种图像信息和先验知识,有望在新的理论技术上做出突破。

Objective： In Clinical medicine, accurate segmentation of brain tumors has important application value, but brain tumor has complex structure, fuzzy boundaries and mixes with normal brain tissue, so it is very difficult to achieve the correct segmentation of brain tumors. To provide a useful reference for researchers, this paper explores the research progress of brain tumor Segmentation based on MRI. Methods： Consulting the related literatures home and abroad, we classified the methods for brain tumor MR image segmentation, reviews and discusses brain tumor segmentation technology based on magnetic resonance imaging and its research progress in recent years, and introduce the evaluation measures for segmentation methods. Finally, we concluded with a discussion on the prospect of future research in it. Results： The methods for brain tumor segmentation mainly include region growing method, clustering segmentation method, deformable model, watershed, segmentation based on atlas, Multispectral MR image segmentation, outlier detection and so on. Conclusions： With the further study, brain tumor segmentation method based on MRI to the segmentation tends to full automatic, real-time, accurate, effectively combining multiple segmentation methods, comprehensive use of a variety of image information and prior knowledge and will make a breakthrough in the new theory and technology.