IEEE Transactions on Biomedical Engineering, Nov 1, 2022
OBJECTIVE Electroencephalogram (EEG) is one of the most widely used signals in motor imagery (MI)... more OBJECTIVE Electroencephalogram (EEG) is one of the most widely used signals in motor imagery (MI) based brain-computer interfaces (BCIs). Domain adaptation has been frequently used to improve the accuracy of EEG-based BCIs for a new user (target domain), by making use of labeled data from a previous user (source domain). However, this raises privacy concerns, as EEG contains sensitive health and mental information. It is very important to perform privacy-preserving domain adaptation, which simultaneously improves the classification accuracy for a new user and protects the privacy of a previous user. METHODS We propose augmentation-based source-free adaptation (ASFA), which consists of two parts: 1) source model training, where a novel data augmentation approach is proposed for MI EEG signals to improve the cross-subject generalization performance of the source model; and, 2) target model training, which simultaneously considers uncertainty reduction for domain adaptation and consistency regularization for robustness. ASFA only needs access to the source model parameters, instead of the raw EEG data, thus protecting the privacy of the source domain. We further extend ASFA to a stricter privacy-preserving scenario, where the source model's parameters are also inaccessible. RESULTS Experimental results on four MI datasets demonstrated that ASFA outperformed 15 classical and state-of-the-art MI classification approaches. SIGNIFICANCE This is the first work on completely source-free domain adaptation for EEG-based BCIs. Our proposed ASFA achieves high classification accuracy and strong privacy protection simultaneously, important for the commercial applications of EEG-based BCIs.
Some concepts of the graphical unitary group approach (GUGA) have been applied in a method of eva... more Some concepts of the graphical unitary group approach (GUGA) have been applied in a method of evaluation of matrix elements based on the permutational symmetry of the wavefunction. The new method seems to be simpler than both the original method of evaluation of matrix elements and the unitary group approach formulated in the Gelfand-Tsetlin basis. Introduction of a graphical representation of the orbital parts of the configuration state functions allows us to give explicit formulae for matrix elements of two-electron operators. The resulting formalism may be applied in the conventional and in the direct configuration interaction methods. The reference state may be taken as an arbitrary multiconfiguration function.
An amendment to this paper has been published and can be accessed via a link at the top of the pa... more An amendment to this paper has been published and can be accessed via a link at the top of the paper.
ObjectivesDisorders of consciousness are very big medical and social problem. Their variability, ... more ObjectivesDisorders of consciousness are very big medical and social problem. Their variability, problems in precise definition and proper diagnosis make difficult assessing their causes and effectiveness of the therapy. In the paper we present our point of view to a problem of consciousness and its most common disorders.MethodsFor this moment scientists do not know exactly, if these disorders can be a result of simple but general mechanism, or a complex set of mechanisms, both on neural, molecular or system level. Presented in the paper simulations using neural network models, including biologically relevant consciousness’ modelling, help assess influence of specified causes.ResultsNonmotoric brain activity can play important role within diagnostic process as a supplementary method for motor capabilities. Simple brain sensory (e.g. visual) processing of both healthy subject and people with consciousness disorders help checking hypotheses in the area of consciousness’ disorders’ mec...
The topic of brain stem computational simulation still seems understudied in contemporary scienti... more The topic of brain stem computational simulation still seems understudied in contemporary scientific literature. Current advances in neuroscience leave the brain stem as one of the least known parts of the human central nervous system. Brain stem lesions are particularly damaging to the most important physiological functions. Advances in brain stem modeling may influence important issues within the core of neurology, neurophysiology, neurosurgery, and neurorehabilitation. Direct results may include both development of knowledge and optimization and objectivization of clinical practice in the aforementioned medical areas. Despite these needs, progress in the area of computational brain stem models seems to be too slow. The aims of this paper are both to recognize the strongest limitations in the area of computational brain stem simulations and to assess the extent to which current opportunities may be exploited. Despite limitations, the emerging view of the brain stem provided by its...
IEEE Transactions on Biomedical Engineering, Nov 1, 2022
OBJECTIVE Electroencephalogram (EEG) is one of the most widely used signals in motor imagery (MI)... more OBJECTIVE Electroencephalogram (EEG) is one of the most widely used signals in motor imagery (MI) based brain-computer interfaces (BCIs). Domain adaptation has been frequently used to improve the accuracy of EEG-based BCIs for a new user (target domain), by making use of labeled data from a previous user (source domain). However, this raises privacy concerns, as EEG contains sensitive health and mental information. It is very important to perform privacy-preserving domain adaptation, which simultaneously improves the classification accuracy for a new user and protects the privacy of a previous user. METHODS We propose augmentation-based source-free adaptation (ASFA), which consists of two parts: 1) source model training, where a novel data augmentation approach is proposed for MI EEG signals to improve the cross-subject generalization performance of the source model; and, 2) target model training, which simultaneously considers uncertainty reduction for domain adaptation and consistency regularization for robustness. ASFA only needs access to the source model parameters, instead of the raw EEG data, thus protecting the privacy of the source domain. We further extend ASFA to a stricter privacy-preserving scenario, where the source model's parameters are also inaccessible. RESULTS Experimental results on four MI datasets demonstrated that ASFA outperformed 15 classical and state-of-the-art MI classification approaches. SIGNIFICANCE This is the first work on completely source-free domain adaptation for EEG-based BCIs. Our proposed ASFA achieves high classification accuracy and strong privacy protection simultaneously, important for the commercial applications of EEG-based BCIs.
Some concepts of the graphical unitary group approach (GUGA) have been applied in a method of eva... more Some concepts of the graphical unitary group approach (GUGA) have been applied in a method of evaluation of matrix elements based on the permutational symmetry of the wavefunction. The new method seems to be simpler than both the original method of evaluation of matrix elements and the unitary group approach formulated in the Gelfand-Tsetlin basis. Introduction of a graphical representation of the orbital parts of the configuration state functions allows us to give explicit formulae for matrix elements of two-electron operators. The resulting formalism may be applied in the conventional and in the direct configuration interaction methods. The reference state may be taken as an arbitrary multiconfiguration function.
An amendment to this paper has been published and can be accessed via a link at the top of the pa... more An amendment to this paper has been published and can be accessed via a link at the top of the paper.
ObjectivesDisorders of consciousness are very big medical and social problem. Their variability, ... more ObjectivesDisorders of consciousness are very big medical and social problem. Their variability, problems in precise definition and proper diagnosis make difficult assessing their causes and effectiveness of the therapy. In the paper we present our point of view to a problem of consciousness and its most common disorders.MethodsFor this moment scientists do not know exactly, if these disorders can be a result of simple but general mechanism, or a complex set of mechanisms, both on neural, molecular or system level. Presented in the paper simulations using neural network models, including biologically relevant consciousness’ modelling, help assess influence of specified causes.ResultsNonmotoric brain activity can play important role within diagnostic process as a supplementary method for motor capabilities. Simple brain sensory (e.g. visual) processing of both healthy subject and people with consciousness disorders help checking hypotheses in the area of consciousness’ disorders’ mec...
The topic of brain stem computational simulation still seems understudied in contemporary scienti... more The topic of brain stem computational simulation still seems understudied in contemporary scientific literature. Current advances in neuroscience leave the brain stem as one of the least known parts of the human central nervous system. Brain stem lesions are particularly damaging to the most important physiological functions. Advances in brain stem modeling may influence important issues within the core of neurology, neurophysiology, neurosurgery, and neurorehabilitation. Direct results may include both development of knowledge and optimization and objectivization of clinical practice in the aforementioned medical areas. Despite these needs, progress in the area of computational brain stem models seems to be too slow. The aims of this paper are both to recognize the strongest limitations in the area of computational brain stem simulations and to assess the extent to which current opportunities may be exploited. Despite limitations, the emerging view of the brain stem provided by its...
Uploads
Papers by Wlodzislaw Duch