Fast Thin-layer Structural Magnetic Resonance Imaging-based Hippocampal Characterization And Automatic Identification Of Patients With Mesial Temporal Lobe Epilepsy

Objective: 1)To study the differences of quantitative parameters such as structural volume,cortical thickness,and cortical surface area in the left and right white and gray matter structures of healthy human brains in structural magnetic resonance images,and to evaluate the influence of age and gender on quantitative parameters of gray and white matter structures,to explore the quantification of volume,cortical thickness,and cortical surface area in Magnetic Resonance Imaging negative Mesial Temporal Lobe Epilepsy(MRIn-MTLE)patients’ structurl Magnetic Resonance Images(s MRI).The difference between the parameters and the healthy people,the difference parameters were selected for classification research;(2)Utilize the hippocampus and amygdala subfiled automated segmentation and measurement technology for the segmentation and calculation of the hippocampus,amygdala and subregions of healthy controls and MRIn-MTLE patients,compare the differences in quantitative parameters of hippocampus,amygdala and subregions between MRIn-MTLE patients and healthy controls,select the different parameters for classification research;(3)Use artificial intelligence technology to analyze the gray matter and white matter of MRIn-MTLE patients with differences in s MRI.,hippocampus,amygdala,and other different parameters for model training and feature screening to evaluate the performance of parameters used for automated classification training and the performance of the technique for MRIn-MTLE patient classification.Methods:(1)Collect s MRI images of MRIn-MTLE patients retrospectively,and recruit MRIn-MTLE patients who meet the research requirements as well as age-and sex-matched healthy volunteers,using magnetic resonance equipment with a field strength of ≥3.0T(GE Signa HDxt,GE Signa Architect,Simense MEGNTOM Skyra)performed T1WI s MRI structural magnetic resonance sequence scans on patients and volunteers,and used Freesurfer7.1.1 to process the obtained images,such as motion correction,individual spatial registration,and atlas segmentation,and calculated cortex and cortex.Quantitative structural parameters such as volume of subregions,cortical thickness,cortical and white matter surface area,and quantitative structural parameters of the telencephalon region of interest(ROI)in healthy volunteers and MRIn-MTLE patients were analyzed,and t-test was used to compare quantitative structural parameters of telencephalic ROI in patients with MRIn-MTLE and healthy controls,including age and gender,and analyzed their effects on the quantitative parameters of T1 WI s MRI images of MRIn-MTLE patients and healthy controls.Statistically significantly different structures were used for classification analysis;(2)The collected T1 WI s MRI images of MRIn-MTLE patients and healthy volunteers were further segmented and quantitative parameters were calculated using refined atlases,and the hippocampus of MRIn-MTLE patients and healthy volunteers was obtained.The volume of the hippocampus,amygdala and subregions were analyzed,and the differences in the volume of the hippocampus,amygdala and subregions between patients and healthy controls were analyzed,and the structures with statistically significant differences in volume were selected for the next classification parameters;(3)The characteristic parameters of different brain regions of healthy and MRIn-MTLE patients obtained by T1 WI s MRI image analysis were used for image classification training of patients and healthy people,respectively,after upsampling,data normalization,principal component analysis,Pearson correlation coefficient screening and extract feature parameters,KW feature selection,and used support vector machines,random forests and other algorithms to build feature-based classification models and train them to evaluate the classification performance and diagnostic value of T1 WI s MRI image features in MRIn-MTLE.Results:(1)Whole-brain segmentation and measurement results showed that there were differences between genders and sides in parameters such as volume,cortical thickness,and cortical surface area of both cerebral hemispheres.Compared with the HC group,it was found that MRIn-MTLE patients had multiple brain regions were different in volume,cortical thickness,and surface area,and MRIn-MTLE affected male and female brains in different ways,the brain regions that differed included subregional volume,cortical thickness,and cortical surface area in the frontal and some temporal lobes;(2)The segmentation and volume measurement results of the amygdala,hippocampus and subregions in the HC group and the MRIn-MTLE group indicated that there were differences in gender and between left and right cerebral hemispheres in the amygdala,hippocampus and its subregions.Compared with the HC group,it was found that,there are differences in the volume of some hippocampal subfields and amygdala subregions in MRIn-MTLE;(3)the volume of the whole brain region of interest obtained by Freesurfer segmentation,cortical thickness,cortical surface area,amygdala and its subregions volume,hippocampus and its subregions The subregion volume and cortical feature parameters can be used for the classification task of MRIn-MTLE patient images,among them,the AUC of the female telencephalon ROI volume and the model have the highest accuracy,reaching 0.761 and 0.736 on the test data set,the AUC and model accuracy of the classification model constructed with volumetric parameters of the hippocampal subregions on the test dataset were 0.659 and 0.642,respectively.Conclusion:(1)Using Freesurfer’s brain atlas to segment and calculate the T1 WI s MRI images of MRIn-MTLE can find differences in brain regions that cannot be identified by naked eyes;(2)The volume of the amygdala,hippocampus,and its subregions in MRIn-MTLE patients is significantly different from that of HC.There are differences between groups,and it is affected by factors such as age and gender;(3)The measurement results based on T1 WI s MRI can assist in the determination of epileptogenic foci.Automated volume measurements of the whole brain and of the amygdala,hippocampus,and their subregions based on rapid thin-slice T1 WI s MRI may have the potential to predict cell loss patterns in hippocampal sclerosis before surgery and improve the diagnostic accuracy of MRIn-MTLE.