Electroencephalography(EEG)analysis extracts critical information from brain signals,enabling brain disease diagnosis and providing fundamental support for brain–computer interfaces.However,performing an artificial i...Electroencephalography(EEG)analysis extracts critical information from brain signals,enabling brain disease diagnosis and providing fundamental support for brain–computer interfaces.However,performing an artificial intelligence analysis of EEG signals with high energy efficiency poses significant challenges for electronic processors on edge computing devices,especially with large neural network models.Herein,we propose an EEG opto-processor based on diffractive photonic computing units(DPUs)to process extracranial and intracranial EEG signals effectively and to detect epileptic seizures.The signals of the EEG channels within a second-time window are optically encoded as inputs to the constructed diffractive neural networks for classification,which monitors the brain state to identify symptoms of an epileptic seizure.We developed both free-space and integrated DPUs as edge computing systems and demonstrated their applications for real-time epileptic seizure detection using benchmark datasets,that is,the Children’s Hospital Boston(CHB)–Massachusetts Institute of Technology(MIT)extracranial and Epilepsy-iEEG-Multicenter intracranial EEG datasets,with excellent computing performance results.Along with the channel selection mechanism,both numerical evaluations and experimental results validated the sufficiently high classification accuracies of the proposed opto-processors for supervising clinical diagnosis.Our study opens a new research direction for utilizing photonic computing techniques to process large-scale EEG signals and promote broader applications.展开更多
The analysis of microstates in EEG signals is a crucial technique for understanding the spatiotemporal dynamics of brain electrical activity.Traditional methods such as Atomic Agglomerative Hierarchical Clustering(AAH...The analysis of microstates in EEG signals is a crucial technique for understanding the spatiotemporal dynamics of brain electrical activity.Traditional methods such as Atomic Agglomerative Hierarchical Clustering(AAHC),K-means clustering,Principal Component Analysis(PCA),and Independent Component Analysis(ICA)are limited by a fixed number of microstate maps and insufficient capability in cross-task feature extraction.Tackling these limitations,this study introduces a Global Map Dissimilarity(GMD)-driven density canopy K-means clustering algorithm.This innovative approach autonomously determines the optimal number of EEG microstate topographies and employs Gaussian kernel density estimation alongside the GMD index for dynamic modeling of EEG data.Utilizing this advanced algorithm,the study analyzes the Motor Imagery(MI)dataset from the GigaScience database,GigaDB.The findings reveal six distinct microstates during actual right-hand movement and five microstates across other task conditions,with microstate C showing superior performance in all task states.During imagined movement,microstate A was significantly enhanced.Comparison with existing algorithms indicates a significant improvement in clustering performance by the refined method,with an average Calinski-Harabasz Index(CHI)of 35517.29 and a Davis-Bouldin Index(DBI)average of 2.57.Furthermore,an information-theoretical analysis of the microstate sequences suggests that imagined movement exhibits higher complexity and disorder than actual movement.By utilizing the extracted microstate sequence parameters as features,the improved algorithm achieved a classification accuracy of 98.41%in EEG signal categorization for motor imagery.A performance of 78.183%accuracy was achieved in a four-class motor imagery task on the BCI-IV-2a dataset.These results demonstrate the potential of the advanced algorithm in microstate analysis,offering a more effective tool for a deeper understanding of the spatiotemporal features of EEG signals.展开更多
Background Frontal lobe injury(FLI)is related to cognitive control impairments,but the influences of FLI on the internal subprocesses of cognitive control remain unclear.Aims We sought to identify specific biomarkers ...Background Frontal lobe injury(FLI)is related to cognitive control impairments,but the influences of FLI on the internal subprocesses of cognitive control remain unclear.Aims We sought to identify specific biomarkers for long-term dysfunction or compensatory modulation in different cognitive control subprocesses.Methods A retrospective case-control study was conducted.Event-related potentials(ERP),oscillations and functional connectivity were used to analyse electroencephalography(EEG)data from 12 patients with unilateral frontal lobe injury(UFLI),12 patients with bilateral frontal lobe injury(BFLI)and 26 healthy controls(HCs)during a Go/NoGo task,which included several subprocesses:perceptual processing,anticipatory preparation,conflict monitoring and response decision.Results Compared with the HC group,N2(the second negative peak in the averaged ERP waveform)latency,and frontal and parietal oscillations were decreased only in the BFLI group,whereas P3(the third positive peak in the averaged ERP waveform)amplitudes and sensorimotor oscillations were decreased in both patient groups.The functional connectivity of the four subprocesses was as follows:alpha connections of posterior networks in the BFLI group were lower than in the HC and UFLI groups,and these alpha connections were negatively correlated with neuropsychological tests.Theta connections of the dorsal frontoparietal network in the bilateral hemispheres of the BFLI group were lower than in the HC and UFLI groups,and these connections in the uninjured hemisphere of the UFLI group were higher than in the HC group,which were negatively correlated with behavioural performances.Delta and theta connections of the midfrontal-related networks in the BFLI group were lower than in the HC group.Theta across-network connections in the HC group were higher than in the BFLI group but lower than in the UFLI group.Conclusions The enhancement of low-frequency connections reflects compensatory mechanisms.In contrast,alpha connections are the opposite,therefore revealing more abnormal neural activity and less compensatory connectivity as the severity of injury increases.The nodes of the above networks may serve as stimulating targets for early treatment to restore corresponding functions.EEG biomarkers can measure neuromodulation effects in heterogeneous patients.展开更多
基金supported by the National Major Science and Technology Projects of China(2021ZD0109902 and 2020AA0105500)the National Natural Science Fundation of China(62275139 and 62088102)the Tsinghua University Initiative Scientific Research Program.
文摘Electroencephalography(EEG)analysis extracts critical information from brain signals,enabling brain disease diagnosis and providing fundamental support for brain–computer interfaces.However,performing an artificial intelligence analysis of EEG signals with high energy efficiency poses significant challenges for electronic processors on edge computing devices,especially with large neural network models.Herein,we propose an EEG opto-processor based on diffractive photonic computing units(DPUs)to process extracranial and intracranial EEG signals effectively and to detect epileptic seizures.The signals of the EEG channels within a second-time window are optically encoded as inputs to the constructed diffractive neural networks for classification,which monitors the brain state to identify symptoms of an epileptic seizure.We developed both free-space and integrated DPUs as edge computing systems and demonstrated their applications for real-time epileptic seizure detection using benchmark datasets,that is,the Children’s Hospital Boston(CHB)–Massachusetts Institute of Technology(MIT)extracranial and Epilepsy-iEEG-Multicenter intracranial EEG datasets,with excellent computing performance results.Along with the channel selection mechanism,both numerical evaluations and experimental results validated the sufficiently high classification accuracies of the proposed opto-processors for supervising clinical diagnosis.Our study opens a new research direction for utilizing photonic computing techniques to process large-scale EEG signals and promote broader applications.
基金funded by National Nature Science Foundation of China,Yunnan Funda-Mental Research Projects,Special Project of Guangdong Province in Key Fields of Ordinary Colleges and Universities and Chaozhou Science and Technology Plan Project of Funder Grant Numbers 82060329,202201AT070108,2023ZDZX2038 and 202201GY01.
文摘The analysis of microstates in EEG signals is a crucial technique for understanding the spatiotemporal dynamics of brain electrical activity.Traditional methods such as Atomic Agglomerative Hierarchical Clustering(AAHC),K-means clustering,Principal Component Analysis(PCA),and Independent Component Analysis(ICA)are limited by a fixed number of microstate maps and insufficient capability in cross-task feature extraction.Tackling these limitations,this study introduces a Global Map Dissimilarity(GMD)-driven density canopy K-means clustering algorithm.This innovative approach autonomously determines the optimal number of EEG microstate topographies and employs Gaussian kernel density estimation alongside the GMD index for dynamic modeling of EEG data.Utilizing this advanced algorithm,the study analyzes the Motor Imagery(MI)dataset from the GigaScience database,GigaDB.The findings reveal six distinct microstates during actual right-hand movement and five microstates across other task conditions,with microstate C showing superior performance in all task states.During imagined movement,microstate A was significantly enhanced.Comparison with existing algorithms indicates a significant improvement in clustering performance by the refined method,with an average Calinski-Harabasz Index(CHI)of 35517.29 and a Davis-Bouldin Index(DBI)average of 2.57.Furthermore,an information-theoretical analysis of the microstate sequences suggests that imagined movement exhibits higher complexity and disorder than actual movement.By utilizing the extracted microstate sequence parameters as features,the improved algorithm achieved a classification accuracy of 98.41%in EEG signal categorization for motor imagery.A performance of 78.183%accuracy was achieved in a four-class motor imagery task on the BCI-IV-2a dataset.These results demonstrate the potential of the advanced algorithm in microstate analysis,offering a more effective tool for a deeper understanding of the spatiotemporal features of EEG signals.
基金This work was supported by the National Natural Science Foundation of China(82271933,81971800 and 81871536)Priority Academic Program Development of Jiangsu Higher Education Institutes(PAPD)(SYSD2012063)Shandong Provincial Natural Science Foundation(ZR2023QH434),China.
文摘Background Frontal lobe injury(FLI)is related to cognitive control impairments,but the influences of FLI on the internal subprocesses of cognitive control remain unclear.Aims We sought to identify specific biomarkers for long-term dysfunction or compensatory modulation in different cognitive control subprocesses.Methods A retrospective case-control study was conducted.Event-related potentials(ERP),oscillations and functional connectivity were used to analyse electroencephalography(EEG)data from 12 patients with unilateral frontal lobe injury(UFLI),12 patients with bilateral frontal lobe injury(BFLI)and 26 healthy controls(HCs)during a Go/NoGo task,which included several subprocesses:perceptual processing,anticipatory preparation,conflict monitoring and response decision.Results Compared with the HC group,N2(the second negative peak in the averaged ERP waveform)latency,and frontal and parietal oscillations were decreased only in the BFLI group,whereas P3(the third positive peak in the averaged ERP waveform)amplitudes and sensorimotor oscillations were decreased in both patient groups.The functional connectivity of the four subprocesses was as follows:alpha connections of posterior networks in the BFLI group were lower than in the HC and UFLI groups,and these alpha connections were negatively correlated with neuropsychological tests.Theta connections of the dorsal frontoparietal network in the bilateral hemispheres of the BFLI group were lower than in the HC and UFLI groups,and these connections in the uninjured hemisphere of the UFLI group were higher than in the HC group,which were negatively correlated with behavioural performances.Delta and theta connections of the midfrontal-related networks in the BFLI group were lower than in the HC group.Theta across-network connections in the HC group were higher than in the BFLI group but lower than in the UFLI group.Conclusions The enhancement of low-frequency connections reflects compensatory mechanisms.In contrast,alpha connections are the opposite,therefore revealing more abnormal neural activity and less compensatory connectivity as the severity of injury increases.The nodes of the above networks may serve as stimulating targets for early treatment to restore corresponding functions.EEG biomarkers can measure neuromodulation effects in heterogeneous patients.