基于内容的MRI脑肿瘤图像特征提取及检索方法

Content-based feature extraction and retrieval method of MRI brain tumorimage

医学图像检索中特征提取方法对检索的效果、性能具有重要影响,针对这个问题,设计了一个基于内容的医学图像检索系统。为了给医学图像检索系统的临床应用提供参考价值,该系统以哈佛大学医学院开发的脑肿瘤MRI医学图像数据库为背景,比较了颜色相关图、颜色矩、灰度共生矩阵、金字塔小波变换和树型小波变换这5个特征提取技术对MRI脑肿瘤医学图像的检索性能。实验结果表明树型小波变换和金字塔小波变换的检索效果较好。

Feature extraction method has an important influence on medical image retrieval result and performance. To compare the performance of these methods, a medical image retrieval system was designed and implemented. Brain tumor medical image database developed by Harvard Medical School was used to offer clinical applications valuable advice for reference. The retrieval performances of five feature extraction techniques： color eorrelogram, color moment, gray level cooccurrence matrix, Pyramid wavelet transform and tree wavelet transform were compared. The experimental results show that the retrieval effectiveness of Pyramid wavelet transform and tree wavelet transform methods are better.