Prediction Of Biological Behavior Of Brain Invasion Of Atypical Meningioma Based On Preoperative MRI

Objective In 2016,the World Health Organization（WHO）took"brain invasion（BI）"as an independent diagnostic criterion for II grade atypical meningioma（AM）.Therefore,the original grade I benign meningioma with the characteristics of"brain invasion"will be upgraded to grade II according to the new standard.At the same time,the patient’s operation plan and adjuvant treatment plan also need to be changed accordingly.Pathology is still the gold standard of"brain invasion",but it can only analyze the tissue mass of the specimen submitted for examination,but not the whole tumor,which makes the brain invasion at the tumor-brain interface have the risk of missed diagnosis to a large extent.Magnetic resonance imaging（MRI）can analyze the tumor noninvasively and comprehensively before operation and reflect the pathological characteristics of the tumor to a certain extent,providing a useful basis for judging its invasiveness.Therefore,the purpose of this study is to retrospectively analyze the MRI features of atypical meningiomas and try to find the imaging indicators to predict brain invasive behavior before operation,in order to give clinical and pathological guidance before operation and reduce the misdiagnosis rate of meningioma grading.Materials and Methods A total of 114 patients with primary atypical meningioma confirmed by pathology in Tianjin Huanhu Hospital from March 2016 to September2019 were retrospectively enrolled,including 56 males and 58 females,with an average age of 55.54±12.16 years（range 18-82 years）,58 cases of brain invasion（BI group）and 56 cases of non-brain invasion（NBI）（NBI group）.All patients were evaluated according to MRI without knowing any clinical and pathological data.The evaluation indexes included tumor location,tumor morphology,tumor lobulation growth,tumor enhancement uniformity,tumor-brain interface smoothness,tumor minimum apparent diffusion coefficient(ADC_min),relative minimum ADC value(r ADC_min),shortest tumor diameter,longest tumor diameter,longest tumor diameter and longest edema diameter.Intraclass correlation coefficient（ICC）was used to evaluate the consistency of imaging indexes of two doctors.The imaging data of the patients were uploaded to the radiology cloud platform（https://mics.radcloud.cn）,and the region of interest（ROI）of the tumor was delineated on the T1 weighted imaging-contrast enhancement（T₁WI-CE）.The ROI of peritumoral brain edema（PTBE）on the fluid attenuated in version recovery-contrast enhancement（FLAIR-CE）sequence was delineated by the platform to calculate the tumor volume and edema volume.Then 14 indexes,such as sex,age,tumor location,tumor shape,tumor lobulated growth,tumor enhancement uniformity,tumor-brain interface smoothness,ADC_min,r ADC_min,shortest diameter,longest diameter,longest diameter,edema volume,edema volume and so on,were analyzed by univariate binary Logistic regression analysis.Spearman correlation analysis was used to analyze the statistically significant indexes in the single factor.The statistically significant indexes in univariate analysis with no strong correlation were included in multivariate binary Logistic regression analysis to screen the indicators closely related to the occurrence of brain invasion of meningioma.Draw the receiver operating characteristic（ROC）curve to evaluate the predictive value of the indexes selected in multivariate analysis for the state of brain invasion,and calculate the area under the ROC curve（AUC）,optimal threshold,sensitivity and specificity.Results 1、The indexes evaluated by two doctors were consistent（ICC=0.796～0.949）.2、The univariate binary Logistic regression analysis of the above indexes showed that the following seven indexes were significantly correlated with the occurrence of brain invasion:when the tumor enhancement uniformity was uneven[OR=0.450,95%CI（0.210～0.966）],when the tumor-brain interface was not smooth[OR=0.311%,95%CI（0.144～0.672）],the r ADC_min value decreased[OR=38.940,95%CI（1.817～834.563）),maximum tumor diameter reduction hours[OR=1.029,95%CI（1.004～1.054）],increased maximum edema diameter[OR=0.944,95%CI（0.920～0.968）],tumor volume reduction[OR=1.011,95%CI（1.001～1.021）],increased edema volume[OR=0.980,95%CI（0.970～0.990）].There were significant differences in other 7 indexes:sex,age,tumor location,tumor morphology,tumor lobulation growth,tumor ADC_min value and tumor shortest diameter（P>0.05）.3、The longest diameter of tumor was highly correlated with tumor volume（r=0.948,P<0.01）,and the longest diameter of edema was highly correlated with edema volume（r=0.913,P<0.01）.There was a weak correlation or no correlation between the other two variables.4、When the longest diameter of tumor and the longest diameter of edema were selected into the binary Logistic regression equation,the following four indexes were significantly correlated with the occurrence of brain invasion:when the smoothness of tumor-brain interface was not smooth[OR=0.189%,95%CI（0.052-0.686）],r ADC_minvalue decreased[OR=2693.910,95%CI（18.975～382460.060）],and the longest diameter of tumor decreased[OR=1.126,95%CI（1.070～1.186）,when the longest diameter of edema increased[OR=0.914,95%CI（0.877～0.952）].When the tumor volume and edema volume were selected into the binary Logistic regression equation,the results showed that the following four indexes were significantly correlated with the occurrence of brain invasion:when the tumor-brain interface was not smooth[OR=0.227,95%CI（0.076-0.676）],r ADC_min decreased[OR=247.436%,95%CI（3.908-15666.086）],and tumor volume decreased[OR=1.042,95%CI（1.021-1.065）],when the edema volume increased[OR=0.969,95%CI（0.954-0.985）].5、In a single index,the longest diameter of edema has the best predictive effect on the occurrence of BI（AUC=0.778,sensitivity 67.2%,specificity 78.6%）,edema volume has a good predictive effect on the occurrence of BI（AUC=0.738,sensitivity 58.6%,specificity 83.9%）.The other single index AUC is between 0.6 and 0.7,which is slightly weaker.The highest AUC of the comprehensive index-diameter is 0.899,which is larger than the AUC,of any single index,and the sensitivity and specificity are also the highest,which are 81%and 87.5%,respectively.The prediction performance of the comprehensive index-volume is slightly lower than that of the comprehensive index-diameter,but its prediction efficiency is also higher than that of any single index,with an AUC of0.876,a sensitivity of 82.8%and a specificity of 82.1%.Conclusions Preoperative MRI images can better predict the biological behavior of meningioma brain invasion,which is helpful for patients to formulate personalized treatment plans and improve the detection rate of brain invasion neuropathology.The longest diameter of edema can independently predict the effectiveness of BI,but the combined application of the longest diameter of edema,the longest diameter of tumor,r ADC_min and the smoothness of tumor-brain interface is more helpful to improve the predictive efficiency of BI.