Opinions

Letter to the Editor: Non-contrast CT texture analysis in predicting intracerebral hemorrhage growth

by Fan Xia, Zhiyuan Yu, Jun Zheng, Chao You, Hao Li (ns_lihao@126.com)

Quantitative parameters of CT texture analysis as potential markers for early prediction of spontaneous intracranial hemorrhage enlargement

Letter to the Editor:

With great interest, we read the article by Shen et al [1]. (Shen Q, Shan Y, Hu Z, et al: Quantitative parameters of CT texture analysis as potential markers for early prediction of spontaneous intracranial hemorrhage enlargement. Eur Radiol [2018 Oct; 28(10): 4389-4396. DOI: 10.1007/s00330-018-5364-8.]) regarding the role of non-contrast CT texture analysis (CTTA) in predicting intracerebral hemorrhage (ICH) growth. The results of this study suggest that quantitative parameters generated by CTTA could objectively quantify the heterogeneity of intracerebral hematoma and independently predict early hemorrhage enlargement.

Hematoma heterogeneity is formed by collapsed blood vessels bleeding and military microvascular aneurysms [2; 3], indicating measurable features in the digital image, which is the basis of radiological analysis. It was first qualitatively described by categorization of density heterogeneity [4]. Other semi-quantitatively scoring systems were later reported [5]. However, they still heavily rely on qualitative criteria and the experience of neuroradiologists. In Shen et al.’s study, quantitative parameters calculated via CTTA proved to be potential predictors in early hematoma enlargement, which suggests that CTTA could be a more effective and practical method in the radiological analysis. However, there were some issues that should be discussed.

Two commonly used parameters of texture analysis, skewness and kurtosis, are important in the characterization of the gray-scale image [4-7]. However, in Shen et al.’s study, neither skewness nor kurtosis was discussed. In previous research of heterogeneity of hematoma, Kim et al.’s study introduced these two parameters, which were not found to be the independent predictors of outcome [8]. Thus, we are interested in whether these two parameters have predicting values in early hemorrhage enlargement after Laplacian of Gaussian (LoG) operator filtering.

In addition, the capacity of quantitative parameters in characterizing texture features is worth discussion. It is noticeable that uniformity is more sensitive with the heterogeneity compared to other parameters in Shen et al.’s study. The uniformity quantifies the distribution of pixels. It directly represents homogeneity of the image and it is proven to be positively correlated with the heterogeneity of hemorrhage [1]. In the digital image field, uniformity is interchangeable with energy, defined as the quadratic sum of all elements in the image [9]. Hence, Shen et al.’s study led us to a question whether another indicator of complexity, entropy, which closely related to energy, could be studied in CTTA [10]. On account of low entropy represents well-aligned status in the image, high energy could potentially indicate low entropy in the image [11]. Therefore, as uniformity shows a great value in predicting in hematoma enlargement, the role of entropy in interpreting heterogeneity of hematoma could point us to new research directions.

In Shen et al.’s study, CTTA after image filtering offers a new direction for radiological analysis. However, methods implemented to choose filtering operator and to set Standard Deviation of Gaussian Distribution were not mentioned in this study. Since non-contrast CT texture analysis of intracranial hematoma is a new and promising field, further studies are still required to examine the predicting value of CTTA in ICH growth. Hence, strategies of convolution kernel choosing and setting are needed to be briefly mentioned at least, to assist further research and discussion.

 

References