Experimental analysis of the quantitative ultrasonographic findings in maxillo-facial soft tissue structures

颌面部软组织结构超声定量结果的实验​​分析

• 批准号: 08672159
• 负责人: MIWA Kunihiro
• 金额: $ 1.41万
• 依托单位: Kyushu University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 1996
• 资助国家: 日本
• 起止时间: 1996 至 1997
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-08672159/
• 关键词: pattern recoqnition; smoothing algorithms; texture analysis; カラードプラ像; テクスチェア解析; 定量的画像診断; 均質度; 粗造度

The new measure for texture analysis is based on the intuition that the important texture information for the human visual system is contained in the relative frequency of local extremas in intensity. The principal measurement in this process is the determination of the number of local gray level maximum and minimum along a one-dimensional scan direction. The gray level values are first sent through a smoothing process which eliminates reversals of small amplitude, thereby retaining only the principal extrema. The smoothing algorithm is the digital equivalent of the familiar analog mechanical process known as gear backlash and it was originally described as a preprocessing method for character recognition.The ultrasonographic images of the tissue-mimicking phantom, normal parotid and submandibular glands, sialadenitis and salivary gland tumors were used. The images were scanned, digitized. In the digitized images, the selected ROI was analysed with the Max-Mm Measure in one dimension. The two-dimensional extension measured the max-mm features in the ROI.As the results, the threshold was increased, fewer extrema were detected and when the threshold exceeded the pixel density range of the image, no extrema were detected. Each image produced a characteristic curve which can be used to identify the texture. Each different density image of the same ROI in the base-material of the phantom produced a slightly different curve. Obviously, a larger size region of the ROT would give less variance in the features and more accurate classification results.

纹理分析的新方法是基于这样一种直觉，即人类视觉系统的重要纹理信息包含在局部极值强度的相对频率中。这一过程中的主要测量是确定沿一维扫描方向的局部灰度级最大值和最小值的数目。灰度值首先通过平滑过程发送，该过程消除了小幅度的反转，从而仅保留主极值。该平滑算法是一种被称为齿隙的常见模拟机械过程的数字等价物，最初被描述为一种用于字符识别的预处理方法。使用的是仿组织体模、正常腮腺和颌下腺、涎腺炎和唾液腺肿瘤的超声图像。图像被扫描，数字化。在数字化图像中，选择的ROI用一维的Max-mm测量进行分析。二维扩展测量了感兴趣区域的最大-mm特征，结果是阈值增加，极值检测减少，当阈值超过图像的像素密度范围时，检测不到极值。每幅图像产生一条特征曲线，可以用来识别纹理。在体模的基础材料中，相同ROI的每个不同密度的图像产生的曲线略有不同。显然，ROT的区域越大，特征的变化就越小，分类结果就越准确。