Reply to Letter to the Editor: Quantitative parameters of CT texture analysis as potential markers for early prediction of spontaneous intracranial hemorrhage enlargement.
by Qijun Shen, Zhan Feng, Peipei Pang (email@example.com)Quantitative parameters of CT texture analysis as potential markers for early prediction of spontaneous intracranial hemorrhage enlargement
We thank Drs. Xia et al. for their interest in our study and appreciate the opportunity to address their comments.
First, there were so many feature parameters from gray-histogram. Different algorithms in different software can work out 30-40 parameters [1,2]. Adding all parameters to the study may confuse the purpose, more parameters were more likely to result in false positive statistics . The purpose of our study was to find the texture parameters that can more objectively, more comprehensively reflect the heterogeneity among the most commonly used parameters
Second, skewness represented the degree of asymmetric distribution in the image histogram. Kurtosis is a measure of whether the data are heavy-tailed or light-tailed relative to a normal distribution. Kurtosis and skewness can reflect heterogeneity, but may be unilaterally. Uniformity was literally easy to understand and reflected the homogeneity and heterogeneity as we mentioned in the article, it’s actually almost negatively correlated with entropy. The entropy measures the randomness of the distribution of the coefficients values over the intensity levels. If the value of entropy is high, then the distribution is among more intensity levels in the image. This measurement is the inverse of energy. A simple image has low entropy while a complex image has high entropy. But entropy was too abstract for clinicians to understand. Due to our study focused on extracting fewer and more understandable parameters to reflect heterogeneity more comprehensively, the uniformity we chose was more appropriate. In our study, we introduced the quantitative technique of CTTA as a heterogeneity measure for tumour characterization into the analysis of spontaneous intracerebral hemorrhage and the methods were consistent with previously published partial analyses [5-7].
Finally, We agree that setting of LoG convolution parameters were very important for CTTA. The convolution kernel we mentioned were consistent with previously published partial analyses of CT textures [4,7-9]. CCTA studies based on LOG also basically used these parameters collocation, because these convolution kernels were the most concentrated area of the LOG, containing 99% of the image information. There were not any widely accepted standards for setting of convolution kernels. The effect of LOG setting on CCTA needed to be analyzed in further studies. Our study was focus on extracting texture parameters after filtering, and the results shows our method can extract more significant parameters after filtering.