关键词: 纹理分析, 脑出血, 体层摄影术, X线计算机

Abstract: Objective The aim of the study was to explore the feasibility of predicting early hematoma enlargement of hypertensive intracerebral hemorrhage based on texture analysis of non-contrast CT images. Methods A retrospective analysis of the CT image data of 74 patients with hypertensive intracerebral hemorrhage in our hospital, the patients were divided into two groups with （group A） and without（group B）hematoma enlargement. The Omni-Kinetics software was used to extract the texture features of the hematoma region in the first CT images of the two groups patients for comparison, incorporate statistically different parameters into the binary logistics regression analysis, and perform receiver operating characteristic curves （ROC） for all statistically significant parameters and the regression model analyze diagnostic performance. Results The differences in texture characteristics of the two groups in std-Deviation, Variance, Entropy, UPP, sumEntropy, Compactness, MaxSize, Sphericity were statistically significant （P<0.05）. The prediction model of the binary logistics regression equation was： Logit（P）=-43.929+sumEntropy×81.731+Compactness×-1.300, the area under the ROC curve of Logit（P）was 0.921, when the threshold was 0.694, the sensitivity and specificity were 90.5% and 83.0%. Conclusion Texture analysis based on non-contrast CT images is helpful to predict the early expansion of hematoma in hypertensive intracerebral hemorrhage, and the predictive model has the best diagnostic efficiency.

Key words: texture analysis, intracerebral hemorrhage, tomography, X-ray computer