基于支持向量机的高强度聚焦超声束损伤程度分类识别

• 作 者: 冯艳玲 陈真诚 何继善 钱盛友
• 机 构: 中南大学信息物理工程学院生物医学工程研究所,深圳信息职业技术学院计算机应用系,湖南师范大学物理与信息科学学院,
• 出 处: 生物医学工程学杂志
• 发表日期: 2010年05期
• 关键词: 高强度聚焦超声 程度分类 支持向量机 超声图 凝固性坏死 亚像素级 无损检测 矢量场 特征点 分类算法

基于超声的组织损伤无损检测在高强度聚焦超声(HIFU)的临床推广与应用中具有重要意义。提出一种亚像素级方法将HIFU束损伤所导致的图像变化程度量化为相关距离。采用相关距离作为样本进行基于支持向量机(SVM)的HIFU束损伤程度分类训练,并对方法进行分类识别检验。实验结果表明,超声图像亚像素级相关分析矢量场可反映组织发生凝固性坏死的位置,基于SVM的分类算法可有效识别HIFU束损伤程度。

Ultrasound-based tissue damage non-destructive testing is of great importance in the clinical application and promotion of high-intensity focused ultrasound (HIFU). A sub-pixel-level method is proposed to quantify the degree of image change caused by HIFU beam damage as the correlation distance. The correlation distance was used as a sample to classify the HIFU beam injury degree based on Support Vector Machine (SVM), and the method was classified and verified. The experimental results show that the vector field of sub-pixel-level correlation analysis of ultrasound images can reflect the location of tissue coagulation necrosis. SVM-based classification algorithm can effectively identify the degree of HIFU beam injury.