Value Of CT Texture Analysis On Early Judgement Of Spontaneous Intracranial Hemorrhage Enlargement

Objectives:Hematoma expansion occurs in up to 73%of patients with primary intracerebral hemorrhage(ICH)and is independently associated with early neurological deterioration,disability and death.Prevention of expansion is the only intervention and potential therapeutic target after ICH symptom onset.Previous studies have identified heterogeneity within a hematoma on NCCT,to be predictive of expansion.CT texture analysis(CTTA),a new quantitative technique can be used to assess heterogeneous tumors.We quantified heterogeneity of hematoma and to provide new parameters for predicting hematoma expansion using CTTA.Materials and Methods:We retrospectively studied 82 ICH patients with baseline NCCT and 24-hour follow-up CT available from January 2013 to December 2015.Clinical data was obtained by a clinical neurologist who performed the physical and laboratory examinations,recording the patient’s basic information,blood pressure,blood glucose,onset time,National Institutes of Health Stroke Scale(NIHSS)score,international normalization rate activates part of the prothrombin time(APTT),and anticoagulant therapy history.Image data were assessed by a chief radiologist and a resident radiologist.Patients were dichotomized according to 24-hour hematoma expansion>33%,Regions of interest were drawn manually on the NCCT images according to the border of the hematoma.Through a Laplacian of Gaussian(LoG)band-pass filter,Mean grey-level intensity(M),variation(V)and uniformity(U)were derived with different spatial scales from fine texture(σ=1.0)to coarse texture(σ=2.5).All parameters were examined.Significantly different texture parameters were then subjected to ROC analysis,and AUC,sensitivity,and specificity were used to evaluate the diagnostic efficacy of parameters in the prediction of hematoma enlargement.Consistency analysis between observers was tested using the intraclass correlation coefficient(ICC),with an ICC less than 0.4 indicating poor consistency,0.4-0.75 indicating general consistency,and greater than 0.75 indicating good consistency.The above statistical analysis was performed in R version 3.1.1(http://www.r-project.org).Results:Clinical data:There were no significant differences in age,gender,past history of hypertension,anticoagulation,blood glucose and blood pressure,NIHSS score,APTT time,onset time,or onset time to CT imaging interval between the two groups(p>0.05).Imaging data:There was no significant difference in the baseline volume of the two groups(p>0.05).There was no significant difference in the texture parameters of the two groups before application of filtering(p>0.05),and the mean grayscale showed no significant difference in texture parameters after application of filtering(p>0.05).There was a significant difference in variance under parameter V1.0(p<0.01),but no significant difference under parameters V1.8 and V2.5(p>0.05).There were significant differences in uniformity(p<0.01)under parameters U1.0,U1.8,and U2.5.With variancein the texture parameter V1.0,the area under the curve(AUC)was 0.92,with a sensitivity of 0.85 and a specificity of 0.90.The areas under the curves of the uniformity parameters U1.0,U1.8,and U2.5 were 0.92,0.84,and 0.76,respectively,with sensitivities of 0.85,0.90,and 0.78,respectively,and specificities of 0.90,0.63,and 0.61,respectively.The intra-group correlation coefficients(ICC)were between 0.67 and 0.99 respectively.Conclusion:This study showed that the texture analysis of filtered cranial CT images,variance(V1.0),and uniformity(U1.0,U1.8,and U2.5)can effectively predict hematoma enlargement.This method of quantifying the hematoma by quantifying the CT texture through filtering is faster,more objective,more comprehensive,and more independently operable for heterogeneity analysis of hematoma than the swirl sign and CTA spot sign methods from previous qualitative and semi-quantitative scoring systems.This method may be used to maximize the effectiveness of predictive models in clinical practice.Combined with other CT predictive indicators for ICH patients,this method may aid in improving the clinical guidelines for personalized treatment of ICH patients.