| Background: Epilepsy is a common chronic neurological disorder and a significant contributor to the global burden of non-communicable diseases.Over the past three decades,there has been a rapid increase in the global population,significant population aging,and substantial changes in the incidence and mortality rates of diseases in epidemiological indicators.However,it is still unclear how these factors have influenced the global burden of epilepsy over the past three decades.To address this issue,it is necessary to conduct in-depth research and analysis to better understand the trends and influencing factors of the global burden of epilepsy.This will help to implement targeted measures to effectively reduce the burden caused by epilepsy.Temporal lobe epilepsy(TLE)is the most common focal epilepsy syndrome in adults worldwide,characterized by recurrent,unpredictable seizures.Traditionally,TLE has been considered as a focal brain disorder,but recent research suggests that it is a widespread brain network disease,and the alterations in brain networks in TLE patients may be related to the pathogenesis of temporal lobe epilepsy.Objective: In this study,we aimed at constructing a long-term follow-up cohort of TLE patients,establishing EEG microstate patterns in TLE patients,comparing the differences in EEG microstates between TLE patients and healthy controls(HC),as well as between drugresistant TLE(DRE)and drug-sensitive TLE(DSE)patients,quantifying the brain network changes represented by the alterations in EEG microstates,and using the obtained brain network features to develop machine learning models for differentiating TLE patients from healthy controls and between drug-resistant and drug-sensitive TLE patients.Method: In the first part of the research,we described the current burden of epilepsy using data from the Global Burden of Disease 2019 database.We examined the relationship between indicators such as economic development and healthcare service quality and the burden of epilepsy and conducted frontier analysis to further explore the relationship between economic development and the burden of epilepsy.Subsequently,we used the Norpred ageperiod-cohort model to predict the future trends of the burden of epilepsy over the next 25 years.In the second to fourth parts of the research,we conducted a study on temporal lobe epilepsy(TLE)patients who sought treatment at the Neurology Department Outpatient Clinic of Xijing Hospital from September 2017 to June 2023.The inclusion criteria were as follows: 1.Age ≥ 18 years;2.Diagnosed with right temporal lobe epilepsy according to the 2017 International League Against Epilepsy(ILAE)diagnostic criteria,with a disease duration of more than 1 year;3.Completed 24-hour EEG monitoring at the EEG Monitoring Center of the Neurology Department of Xijing Hospital;4.Completed cranial magnetic resonance imaging(MRI)examination;5.Signed an informed consent form and agreed to participate in follow-up.The exclusion criteria were as follows: 1.Presence of significant structural abnormalities detected in cranial MRI;2.Presence of significant intellectual disabilities;3.Refusal to undergo follow-up.EEG data collected for this study were performed at the EEG Monitoring Center of the Neurology Department of Xijing Hospital.All EEG data were selected from the initial EEG recordings taken before patients received any medication,following the international 10-20 system for electrode placement,with an average reference electrode.We extracted 1-hour background EEG data between 9 AM and 11 AM,without any apparent epileptiform discharges,as the raw data for analysis.All EEG preprocessing and microstate analysis were conducted using MATLAB 2021 b,EEGLAB,and Cartool software.The data obtained in this study were statistically analyzed using the following procedures: Continuous variables were expressed as mean ± standard deviation,and discrete variables were expressed as median and interquartile range.Student’s t-test was used for normally distributed variables,and the Wilcoxon rank-sum test was used for nonnormally distributed variables.Multiple comparisons were corrected using false discovery rate(FDR)correction.Terrain analysis of group comparisons was performed using topographic analysis of variance(TANOVA),with significance set at P < 0.05.Machine learning models were constructed using Python software(version 3.11.0),with packages including Numpy(version 1.26.2),Pandas(version 1.5.3),and Scikit-learn(version 1.2.2).All statistical analyses and visualizations were conducted using MATLAB 2021 b,R software(version 4.3.1),and Python software(version 3.11.0).Results: 1.The analysis on global burden of epilepsy and its trendsIn the first part of the research,we utilized the GBD 2019 database to explore the trends in epilepsy burden from 1990 to 2019 globally.Globally,the prevalence,incidence,mortality,and disability-adjusted life years(DALYs)associated with epilepsy have all increased over the past three decades.There are significant differences in the burden of epilepsy among regions with different levels of development.There is a significant negative relationship between epilepsy burden and the quality of healthcare services(r =-0.64,p<0.001).Frontier analysis reveals that the burden of epilepsy in some countries does not align with their level of economic development.Over the next 25 years,there is a projected upward trend in the global burden of epilepsy.2.Study on Microstate Characteristics of Temporal Lobe Epilepsy Patients In the second part of the study,we constructed seven microstate topographic maps for the temporal lobe epilepsy(TLE)group,healthy control group,drug-resistant TLE(DRE) group,and drug-sensitive TLE(DSE)group.We calculated and compared the temporal parameters of different microstates and found that:(1)The microstate topographic maps constructed from normal background EEG of TLE patients showed significant differences compared to the healthy population(all P < 0.001);(2)Within the subgroups of TLE patients,significant differences were found between microstate E and microstate F topographic maps in the DRE group and DSE group(P = 0.019;P = 0.012);(3)After analyzing and comparing the temporal parameters of the seven microstates,we found that TLE patients exhibited significantly accelerated microstate dynamics,which was consistent across all microstates;(4)In the subgroups of TLE patients,although the differences in the temporal parameters of microstates were not pronounced between the DRE group and DSE group,significant differences were still observed in the temporal structure of microstates.3.Study on Spatial and Temporal Variability of Microstate Dynamic Functional Connectivity Networks in Temporal Lobe Epilepsy PatientsIn the third part of the study,using the microstates obtained in part two,we constructed dynamic functional connectivity networks corresponding to different microstates.We calculated the spatial and temporal variability of these dynamic functional connectivity networks at the regional and global levels and compared the differences in spatial and temporal variability of dynamic functional connectivity networks between TLE patients and healthy controls,as well as between DRE and DSE patients.This allowed us to quantify the microstate differences observed in the first part of the study.We found that:(1)The spatial variability of the microstate-based dynamic functional connectivity networks in TLE patients was significantly lower than that in healthy controls(FDR-corrected P < 0.0001);(2)The temporal variability of the microstate-based dynamic functional connectivity networks in TLE patients was significantly higher than that in healthy controls(FDRcorrected P < 0.01).Furthermore,we quantified the microstate differences between DRE and DSE patients using the same approach.We found that:(1)The spatial variability of the microstate E dynamic functional connectivity network in DRE patients was significantly higher than that in DSE patients(FDR-corrected P < 0.05),particularly in bilateral frontal regions,left temporal region,central region,and right parietal region;(2)The temporal variability of the microstate G dynamic functional connectivity network in DRE patients was significantly lower than that in healthy controls(FDR-corrected P < 0.05),particularly in bilateral frontal regions,left temporal region,and right occipital region.4.Construction of Temporal Lobe Epilepsy Diagnosis Model and Drug-Resistant Epilepsy Prediction Model based on spatial and temporal variability of microstate-based dynamic functional connectivity networksIn the fourth part of the study,utilizing the spatiotemporal variability of the microstate dynamic functional connectivity network obtained from the third part of the research,we first constructed a machine learning diagnostic model to distinguish between TLE patients and healthy individuals,with an average accuracy of 0.95±0.042,average F1 score of 0.96±0.035,average precision of 0.94±0.057,average recall of 0.98±0.034,and an area under the Receiver Operating Characteristic(ROC)curve of 0.997.Subsequently,we further utilized the spatiotemporal variability data of the microstate dynamic functional connectivity network to construct a machine learning prognostic model for distinguishing between DRE and DSE patients among TLE patients,with an average accuracy of 0.93±0.087,average F1 score of 0.94±0.072,average precision of 0.94±0.076,average recall of 0.94±0.096,and an area under the ROC curve of 0.972.Conclusion: 1.The analysis on global burden of epilepsy and its trendsEpilepsy burden remains a significant and unevenly distributed problem worldwide.The burden of epilepsy is closely associated with economic development and the quality of healthcare services.Despite promising progress in reducing the burden of epilepsy in various countries,there is still considerable room for improvement.Over the next 25 years,there is a projected upward trend in the global burden of epilepsy.The burden of epilepsy should continue to be recognized as an urgent global public health priority.2.Study on Microstate Characteristics of Temporal Lobe Epilepsy PatientsEven in the absence of epileptic-like discharges,the resting state brain networks(RSN)of patients with temporal lobe epilepsy(TLE)are still significantly more active than those of healthy individuals,and the specific functional brain network structures also exhibit marked differences from those of healthy populations,leading to unstable states in various resting state brain networks of TLE patients.The differences in resting state brain networks between patients with drug-resistant epilepsy(DRE)and drug-sensitive epilepsy(DSE)are not significant,but DRE patients exhibit activation patterns in the salience network and anterior default mode network(DMN)that are distinct from those of DSE patients.There are significant differences in the specific microstate temporal structures between DRE and DSE,suggesting the presence of potentially more subtle distinctions.3.Study on Spatial and Temporal Variability of Microstate Dynamic Functional Connectivity Networks in Temporal Lobe Epilepsy PatientsCompared to healthy individuals,the resting state brain networks of TLE patients exhibit high instability and synchronization in activity,which may be one of the significant factors contributing to epileptic-like discharges in these patients.In contrast,DRE patients demonstrate highly independent activity in the salience network and highly stable and synchronized activity in the sensorimotor network.Compared to DSE patients,DRE patients exhibit highly independent,stable,and synchronized activity in the bilateral frontal and left temporal regions,implying the formation of stable epileptogenic networks in these regions,which may be one of the key reasons for poor responsiveness to antiepileptic drugs in DRE patients.4.Construction of Temporal Lobe Epilepsy Diagnosis Model and Drug-Resistant Epilepsy Prediction Model based on Spatiotemporal variability of Microstate-based dynamic functional connectivity networksIn the fourth part of the study,utilizing the spatiotemporal variability of microstate dynamic functional connectivity networks obtained in the second part,we first constructed a machine learning diagnosis model to differentiate between TLE patients and healthy controls.The model demonstrated good diagnostic performance and could accurately distinguish TLE patients from healthy individuals.Subsequently,using the spatiotemporal variability data of microstate dynamic functional connectivity networks,we constructed a machine learning prediction model to further differentiate between DRE and DSE patients in the TLE population.The model exhibited good predictive performance and could accurately distinguish DRE patients from DSE patients.The innovative aspects of this study are as follows:(1)The study employed the method of microstate analysis to analyze the EEG data of patients with temporal lobe epilepsy.We proposed the hypothesis that "even in the absence of epileptic discharges or seizures,the brains of patients with temporal lobe epilepsy are more active compared to healthy individuals.There are significant differences in the background EEG microstates between temporal lobe epilepsy patients and healthy controls,which may be related to changes in the brain networks of temporal lobe epilepsy patients." The study validated the reasonableness of this hypothesis;(2)Based on the results of microstate analysis,the study constructed dynamic functional connectivity networks specific to each microstate for the first time.We calculated the spatiotemporal specificity of each network,further quantifying the brain network changes represented by EEG microstate changes caused by temporal lobe epilepsy;(3)The study further analyzed the brain network differences between drugresistant temporal lobe epilepsy and drug-sensitive temporal lobe epilepsy.We discovered that there were highly independent,highly stable,and highly synchronized activities in the bilateral frontal regions and left temporal regions of drug-resistant temporal lobe epilepsy patients compared to drug-sensitive temporal lobe epilepsy patients,indicating the formation of stable epileptogenic networks in this area;(4)the study utilized the spatiotemporal variability of microstate dynamic functional connectivity networks to construct machine learning diagnostic models for temporal lobe epilepsy and machine learning predictive models for drug-resistant temporal lobe epilepsy.The models showed high efficacy and might have some clinical significance. |
PDF Full Text Request