高强度聚焦超声图像纹理分析

Texture analysis of high intensity focused ultrasound image

通过纹理信息分析筛选出合适的纹理特征能为超声图像的目标定位和分割做准备。本文介绍了3种常用的纹理分析方法,并且利用共生矩阵重点对一种新的无创肿瘤治疗技术——高强度聚焦超声(High Intensity Focused Ultrasound,HIFU)治疗中超声图像的纹理信息进行了分析,包括角二阶矩(Angular Second Moment,ASM)、相关性、逆差矩(Inverse Difference Moment,IDM)、熵、对比度、不相似度等。最终确定了3种有效的子宫肌瘤图像纹理特征,即ASM、IDM和相关性。这些特征在肿瘤区域和正常组织之间有明显的差异,ASM和IDM的相对差异都在20%以上,而相关性的相对差异也在5%以上。这3种纹理特征有望运用于HIFU中超声图像的目标定位和分割,但还需要进一步研究噪声对纹理信息的干扰。

Texture information analysis method was applied to screen out proper texture features for the object location and segmentation in ultrasound images. Three commonly used texture analysis methods were introduced in this paper, and the texture information of ultrasound image in a new non-invasion tumor therapy, high intensity focused ultrasound （HIFU） therapy, was analyzed by co-occurrence matrix, including angular second moment （ASM）, correlation, inverse difference moment （IDM）, entropy, contrast and dissimilarity. ASM, IDM and correlation were finally confirmed as three effective texture features of uterine fibroid images. These three texture features appeared obvious differences between tumor region and normal region. The relative differences of ASM and IDM were higher than 20%, and that of correlation was higher than 5%. These texture features are of good potential to be applied in the object location and segmentation of ultrasound images in HIFU. However, the interference of noise on texture information is needed to be further studied.