Objective:To explore the value of energy spectrum CT monochromatic images combined with histogram texture analysis for identifying parapelvic cysts and hydronephrosis.Methods:Retrospectively collected data were obtained from 25 patients with parapelvic cysts and 18 patients with hydronephrosis.The spectral imaging analysis platform was applied to assess monoenergy CT values from 40 to 140 keV at 10 keV increments,and the monoenergy CT values of the two groups were compared.If there were significant differences in monoenergy CT values,the corresponding monoenergy images were reconstructed on AW 4.5 workstation.GE Omni-Kinetics software was used to conduct histogram texture analysis,the largest layer of lesion was selected to outline the ROI,and texture parameters were generated based on grayscale histograms,including the minimum value,maximum value,mean value,standard deviation,variability,skewness,kurtosis,uniformity,energy,entropy,and the 5th,10th,25th,50th,75th,90th,95th percentiles.Single factor analysis was used to compare the differences in each parameter between the two groups.ROC curve was used to analyze the diagnostic value of each parameters.A multivariate logistic regression model was used to screen features and construct the final prediction model,and ROC curve of the final model was drawn,and its value was analyzed.Results:At 40～50 keV,the monoenergy CT values of parapelvic cysts were greater than those of hydronephrosis,the differences were statistically significant(both P＜0.05).There were no significant differences in the monoenergy CT values between the two groups at 60～140 keV.At 40 keV,the mean value,kurtosis,energy,entropy and the 5th,10th,25th,50th,75th percentiles were significantly different between the two groups(all P＜0.05),while the other parameters were not significantly different(all P＞0.05).At 50 keV,the mean value and the 10th,25th,50th percentiles were significantly different between the two groups(all P＜0.05),while the other parameters had no significant differences(all P＞0.05).ROC curve analysis revealed that AUC for distinguishing parapelvic cysts and hydronephrosis with 40 keV texture parameters was 0.696～0.756,greater than that with 50 keV texture parameters.Multivariate logistic regression revealed that kurtosis,energy and the 25th percentile at 40 keV were independent predictors of parapelvic cysts and hydronephrosis,and AUC for the three combined diagnosis was 0.915,the sensitivity was 93.3%and the specificity was 83.3%.Conclusion:The multivariate logistic regression model constructed based on the histogram texture parameters of 40 keV energy spectrum CT images can be used to effectively distinguish between parapelvic cysts and hydronephrosis.
国家科技期刊平台
NSTL
万方数据
