Jamzad Amoon, Setarehdan Seyed Kamaledin
Control and Intelligent Processing Center of Excellence, School of Electrical and Computer Engineering, College of Engineering, University of Tehran, North Kargar Street, Tehran 14395-515, Iran.
J Ultrasound Med. 2014 Apr;33(4):597-610. doi: 10.7863/ultra.33.4.597.
The twinkling artifact is an undesired phenomenon within color Doppler sonograms that usually appears at the site of internal calcifications. Since the appearance of the twinkling artifact is correlated with the roughness of the calculi, noninvasive roughness estimation of the internal stones may be considered as a potential twinkling artifact application. This article proposes a novel quantitative approach for measurement and analysis of twinkling artifact data for roughness estimation.
A phantom was developed with 7 quantified levels of roughness. The Doppler system was initially calibrated by the proposed procedure to facilitate the analysis. A total of 1050 twinkling artifact images were acquired from the phantom, and 32 novel numerical measures were introduced and computed for each image. The measures were then ranked on the basis of roughness quantification ability using different methods. The performance of the proposed twinkling artifact-based surface roughness quantification method was finally investigated for different combinations of features and classifiers.
Eleven features were shown to be the most efficient numerical twinkling artifact measures in roughness characterization. The linear classifier outperformed other methods for twinkling artifact classification. The pixel count measures produced better results among the other categories. The sequential selection method showed higher accuracy than other individual rankings. The best roughness recognition average accuracy of 98.33% was obtained by the first 5 principle components and the linear classifier.
The proposed twinkling artifact analysis method could recognize the phantom surface roughness with average accuracy of 98.33%. This method may also be applicable for noninvasive calculi characterization in treatment management.
闪烁伪像是彩色多普勒超声图像中一种不良现象，通常出现在内部钙化部位。由于闪烁伪像的出现与结石的粗糙度相关，因此内部结石的无创粗糙度估计可被视为闪烁伪像的一种潜在应用。本文提出了一种用于测量和分析闪烁伪像数据以进行粗糙度估计的新定量方法。
制作了一个具有7个粗糙度量化水平的模型。首先按照所提出的程序对多普勒系统进行校准，以利于分析。从该模型获取了总共1050张闪烁伪像图像，并为每张图像引入并计算了32种新的数值测量指标。然后使用不同方法根据粗糙度量化能力对这些指标进行排序。最后针对不同的特征和分类器组合，研究了所提出的基于闪烁伪像的表面粗糙度量化方法的性能。
11个特征被证明是粗糙度表征中最有效的闪烁伪像数值测量指标。线性分类器在闪烁伪像分类方面优于其他方法。像素计数测量指标在其他类别中产生了更好的结果。顺序选择方法显示出比其他个体排名更高的准确性。前5个主成分和线性分类器获得了98.33%的最佳粗糙度识别平均准确率。
所提出的闪烁伪像分析方法能够以98.33%的平均准确率识别模型表面粗糙度。该方法也可能适用于治疗管理中的无创结石表征。
