An Imaging Study Of MRI-negative Epilepsy Based On Multimodality MRI

Part ?:Impairments of gray matter in MRI-negative epileptic patients with different seizure typesObjective:To investigate the damage of gray matter structure in MRI-negative epilepsy patients with different symptoms by voxel-based morphometry(VBM).Methods:From June,2009 to October,2016,ninety MRI-negative epilepsy patients and thirty-five healthy volunteers were underwent the 3T magnetic resonance imaging scan in Nanjing General Hospital.The patients were divided into three groups(including I-GTCS,S-GTCS,and PS)according to different symptoms.The three-dimensional high-resolution T1 structural MRI data was scanned for the voxel-based morphometry.Data of four groups were compared using one-way analysis of variance(ANOVA)were performed to map the overall GMV.An independent-sample t test was performed in order to compare gray matter volume differences between every patient groups and controls.According the results of ANOVA,impaired brain regions were selected as seed regions in order to carried out correlation analysis between gray matter volume and duration.Results:ANOVA showed significant differences in bilateral thalamus and frontal lobe between four groups(alphasim correction,P<0.01).Independent-sample t test showed that the bilateral thalamus and frontal lobe volume decreased in all three patients groups(alphasim correction,P<0.01).The volume of bilateral thalamus showed significantly negatively correlated with duration in I-GTCS patients(P<0.01).Conclusions:Generalized seizures and partial seizures all can cause damage to the gray matter structure,especially in thalamus and frontal lobe.Along with the duration getting longer,the impairments of thalamus and frontal lobe getting different.Suggesting that there might be different influence of the thalamo-cortical network between different epilepsy seizures.Part II:Using machine learning to classify MRI-negative temporal lobe epilepsy based on multi-modality MRIObjective:Through machine learning methods combined with multi-modality MRI features,the diagnostic effect of MRI-negative temporal lobe epilepsy was evaluated and the imaging criteria for efficient classification were searched.At the same time,the potential pathophysiological mechanism of MRI-negative temporal lobe epilepsy was explored by further analysis of the classification model.Methods:In our study,102 patients with MRI negative temporal lobe epilepsy(MRI-negative TLE)were diagnosed in Nanjing Jinling hospital from June 2009 to December 2017.The diagnosis was performed by 2 senior neurologists based on their clinical manifestations,EEG findings and treatment evidence according to the latest classification criteria of ILAE.102 age-and sex-matched healthy controls(HC)were enrolled in this study.The BOLD-fMRI?three-dimensional high-resolution T1WI and DTI data was scanned for the machine learning models.The functional image features(ALFF,fALFF,ReHo,DC,DCglobal)and the structural image features(VBM,FA,MD)of the subjects were calculated to construct the classification model by the support vector machines(SVM)method and multi kernel learning(MKL)method.Then we calculate the accuracy and weight of different features.The classification results of functional features and structural featuress are compared.Results:This study found that the MKL method combined with functional image classification was the best classification model with the highest accuracy rate.The highest rate of accuracy was up to 77.45%(ReHo and DCglobal,sensitivity 72.55%,specificity 82.35%,P<0.05).The classification weight of DCglobal features in the model was about 56.12%,while ReHo was about 43.88%.By further calculating the classification weights of each brain region of,bilateral cerebellum,bilateral temporal lobe and left hippocampus were found to play an important role in classification prediction.Conclusions:MKL method combined with some functional image features can achieve superior classification accuracy and provides a new medcial imaging tool for the clinical diagnosis of MRI negative temporal lobe epilepsy.In addition,our study found that the bilateral cerebellar,temporal and left hippocampal regions of MRI-negative TLE group were different from those of HC group,suggesting that these brain regions may play an important role in the development of disease and provide imaging evidence for further study of its mechanism.