Research On The Diagnosis Of Hippocampal Sclerosis Temporal Lobe Epilepsy Based On T1WI Brain Microstructure Radiomics

Objective: To explore the diagnostic value of radiomics feature values based on T1 WI brain microstructure in the diagnosis of hippocampal sclerosis temporal lobe epilepsy.Methods: From September 2018 to December 2020,20 patients with epilepsy were enrolled in the Department of Neurology,Handan Central Hospital,Hebei Province.A total of 30 cases were included in the temporal lobe epilepsy group,and another 33 cases of healthy control group were included in this study.In this study,the Freesurfer6.0.0 software is used for image segmentation,and the Python program developed by Pyradiomics is used to extract the feature values of radiomics.SPSS 25.0 software was used for statistical analysis.The t-test of independent samples was used for the two groups of cases to calculate the p and t values of each radiomics feature values,when P<0.05,the difference was considered statistically significant,and the ROC curve was drawn by the eigenvalues with significant statistical differences,which used to evaluate the diagnostic value of each eigenvalue inhippocampal sclerotic temporallobe epilepsy.And calculat eAUC,accuracy,sensitivity,specificity,positive predictive value and negative predictive value of each radiomics feature values.Results:(1)The independent-sample t-test between the TLE group and HC group showed that: We compared the extracted 14 shape eigen values,18 first-order eigenvalues,and 75 texture features made pairwise comparisons on the left and right sides.It was found that shape＿Major Axis Length(left),shape＿Major Axis Length(right),shape＿Maximu m2Ddiameter(left),shape＿Maximum2Ddiameter(right)and shape＿Surface Area(left),glcm＿Idn(right),gldm＿low Gray Level Emphasis(right),the five shape feature values and two texture features have significant statistical differences(P<0.05),and the other feature values have no statistical differences(P> 0.05).(2)Based on the ROC curve obtained from the seven eigenvalues,the results are calculated respectively:(1)The AUC of shape＿Major Axislength(left)and shape＿Major Axislength(right)are 0.791(95%CI0.594-0.988)and 0.782(95%CI 0.582-0.982)respectively,its accuracy rates were 71.4%,71.4%,sensitivity were 90.9%,90.9%,and specificity were 50.0%,50.0%,positive predictive value were 66.7%,66.7%,and negative predictive value were 83.3%,88.3% respectively;(2)The AUC of shape＿Maximum2DDiameter(left)and shape＿Maximu m2Ddiameter(right)are 0.795(95%CI 0.605-0.986),0.791(95%CI 0.598-0.984),and their accuracy rates are 71.4% and 71.4%,the sensitivity were 54.5%,54.5%,the specificity were 90.0%,90.0%,the positive predictive value were 85.7%,85.7%,and the negative predictive value were 64.3%,64.3%,respectively;(3)The AUC of shape＿Surface Area(left)was 0.736(95%CI 0.501-0.972),the accuracy is 76.2%,the sensitivity is 81.8%,the specificity is 70.0%,and the positive and negative predictive values ??are 75.0% and 77.8% respectively;(4)The AUC of glcm＿Idn(right)and gldm＿low Gray Level Emphasis(r ight)are 0.755(95%CI 0.537-0.972)and 1.000(95%CI 1.000-1.000)respectively,the accuracy is 60.0%,100.0%,the sensitivity is 55.0%,100.0%,the specificity is 100.0%,100.0%;the positive predictive v is 100.0%,100.0%;and the negative predictive value is 66.7%100.0%,respectively;(5)The comparison between the ROC curves shows that the gldm＿Low Gray Level Emphasis value is the highest in all the indicators of AUC,accuracy,sensitivity,specificity positive predictive value and negative predictive value.The specificity and positive predictive value of glcm＿Idn(right)was also as high as 100.0%.Conclusions: The radiomic shape feature values and texture features of the hippocampus based on T1 anatomical images have potential value for exploring the biological markers of hippocampal sclerosis temporal lobe epilepsy.