Clinical value of MRI texture analysis for preoperative grading of meningiomas based on conventional MRI images / 中华放射学杂志

ABSTRACT

Objective To explore the efficacy and clinical value of texture analysis of conventional MRI image for the preoperative classification of meningiomas. Methods The MR images of 101 meningioma cases were retrospectively analyzed, all of them underwent routine MRI before operation, and confirmed by surgery and pathology,of whom 81 cases in WHO Ⅰ,19 cases in Ⅱand 1 case in Ⅲ.The maximum diameter of the tumor in MRI images was outlined as ROI by the Image J software.The histograms and grey-level co-occurrence matrix(GLCM)were used to measure the texture parameters,such as maximum value,minimum value,standard deviation(SD),skewness,kurtosis,angular second moment,contrast,inverse different moment, entropy, correlation and so on, and compared with postoperative pathological grading results.The independent sample t test or the Mann-Whitney U test was used to compare the difference of the histogram and the gray level co-occurrence matrix parameter of the MRI images between low-grade and high-grade meningiomas. ROC curves for statistically significant parameters were used to confirm their efficacy in predicting the preoperative grade of meningiomas.Results (1)In histogram parameters,the SDs of T2WI,FLAIR,ADC(SDT2WI,SDFLAIR,SDADC)of low grade meningioma were 14.764(10.928,17.932),9.753 (7.385,13.618) and 12.171 (9.138,19.691), the high grade meningiomas were 19.252 (13.580,25.715), 13.568 (8.936,17.108) and 16.636 (13.166,21.498), respectively. The difference between the two groups was statistically significant(Z values were-2.863,-2.250,-2.247,P respectively 0.004,0.024,0.025).The skewness values (SkeADC, SkeFLAIR) of ADC sequence and FLAIR sequence of low-grade meningiomas were 1.377 ± 1.172 and-1.327 ± 0.930, respectively, and the high-grade meningiomas were 2.503 ± 1.613 and-0.827±0.834,respectively.The differences between the two groups were statistically significant (t values were-2.196,-3.129, P values were 0.002,0.030, respectively). (2) In the GLCM parameters, entropy (EntT1WI, EntT2WI, EntFLAIR, EntADC) of the enhanced T1WI, T2WI, FLAIR and ADC sequences of low grade meningiomas were 6.881(6.174,7.305),6.534±0.598,6.019±0.588,6.040±0.588,high-grade meningiomas were 7.079 (6.742,7.739), 7.014 ± 0.514,6.370 ± 0.703,6.576 ± 0.450, respectively. And the difference between the two groups was statistically significant (Z=-2.007, t values were-3.294,-2.327,-3.245, P values were 0.045,0.001,0.022,0.002,respectively).(3)The ROC curves of EntADCand SkeADCwere plotted, and the area under the curve(AUC)was 0.768 and 0.710,respectively.Combined EntADC,SkeADCand EntT2WI were the best for the classification of meningiomas,and AUC was 0.799.Conclusion Texture analysis can provide more quantification information, which can be more accurately distinguishing high grade meningiomas and low grade meningiomas before surgery.