The value of MRI-T2WI texture analysis in the differentiation of atypical medulloblastoma and ependymoma of the fourth ventricle / 中国医师进修杂志

ABSTRACT

Objective@#To evaluate the value of MRI-T2WI texture analysis in the differentiation of atypical medulloblastoma and ependymoma of the fourth ventricle.@*Methods@#Preoperative MRI data of 36 cases of fourth ventricle tumor (19 cases of medulloblastoma and 17 cases of ependymoma) confirmed by the Central Theater General Hospital of the Chinese People′s Liberation Army were retrospectively analyzed. Manually sketch areas of interest (ROI) were made using texture analysis software to get histogram parameters, including mean, median, standard deviation, skewness, kurtosis, maximum, minimum, heterogeneity, entropy, the 5th percentile (T2WI5th), the 10th percentile (T2WI10th), the 25th percentile (T2WI25th), the 50th percentile (T2WI50th), the 75th percentile(T2WI75th), the 95th percentile(T2WI95th). Independent sample t test or mann-whitney U test was used for statistical analysis of each parameter value between the two groups. Logistic regression analysis was used to perform multiparameter joint analysis on meaningful parameters. The area under the curve (AUC) was obtained by using the subject operating characteristic (ROC) curve to determine the threshold and diagnostic efficacy for distinguishing the two group.@*Results@#The mean, median, T2WI5th, T2WI10th, T2WI25th, T2WI50th, T2WI75th values of the medulloblastoma group were smaller than those of the ependymoma group, and difference was statistically significant (P<0.05). The T2WI5th had the best diagnostic performance. When the critical value was 1 343, the diagnostic specificity was 75.0% and the sensitivity was 92.9%. Logistic regression was used to predict the joint analysis of probability parameters. T2WI5th+T2WI10th had the best performance, and AUC was 0.928. When the critical value was 0.75, the diagnostic specificity was 75.0% and the sensitivity was 100.0%.@*Conclusions@#MRI-T2WI texture analysis can provide quantitative information for preoperative diagnosis of atypical medulloblastoma and ependymoma and improve the accuracy of preoperative diagnosis, which has obvious clinical value.