Robust learning from corrupted EEG with dynamic spatial filtering

Banville, Hubert; Wood, Sean U. N.; Aimone, Chris; Engemann, Denis-Alexander; Gramfort, Alexandre

doi:10.1016/j.neuroimage.2022.118994

Cited by 30 publications

(22 citation statements)

References 70 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another avenue of research is to study the potential of EEG learned graphs as a backbone on which spatial filtering of EEG maps can be performed—e.g. for interpolation of missing channels (Banville et al, 2022; Svantesson et al, 2021)—via spectral graph diffusion filtering schemes (Abramian et al, 2021; Tarun et al, 2020). Finally, the proposed EEG-based graph learning and spectral representation via GSP can be readily extended to other data modalities, in particular, fMRI (Itani and Thanou, 2021; Preti and Van De Ville, 2019), near-infrared spectroscopy (Petrantonakis and Kompatsiaris, 2018), or Magnetoencephalography (Tewarie et al, 2019; Sareen et al, 2021).…”

Section: Discussionmentioning

confidence: 99%

Spectral Representation of EEG Data using Learned Graphs with Application to Motor Imagery Decoding

Miri

Abootalebi

Saeedi-Sourck

et al. 2022

Preprint

View full text Add to dashboard Cite

Objective: This paper presents a graph signal processing (GSP)-based approach for decoding two-class motor imagery EEG data via deriving task-specific discriminative features. Methods: First, a graph learning (GL) method is used to learn subject-specific graphs from EEG signals. Second, by diagonalizing the normalized Laplacian matrix of each subject graph, an orthonormal basis is obtained using which the graph Fourier transform (GFT) of the EEG signals is computed. Third, the GFT coefficients are mapped into a discriminative subspace for differentiating two class data using a projection matrix obtained by the Fukunaga-Koontz transform (FKT). Finally, an SVM classifier is trained and tested on the variance of the resulting features to differentiate motor imagery classes. Results: The proposed method is evaluated on Dataset IVa of the BCI Competition III and its performance is compared to i) using features extracted on a graph constructed by Pearson correlation coefficients and ii) three state-of-the-art alternative methods. Conclusion: Experimental results indicate the superiority of the proposed method over alternative methods, reflecting the added benefit of integrating elements from GL, GSP and FKT. Significance: The proposed method and results underpin the importance of integrating spatial and temporal characteristics of EEG signals in extracting features that can more powerfully differentiate motor imagery classes.

show abstract

Section: Discussionmentioning

confidence: 99%

Spectral Representation of EEG Data using Learned Graphs with Application to Motor Imagery Decoding

Miri

Abootalebi

Saeedi-Sourck

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…These are then converted in discrete selections by the Gumbel-Softmax module and applied to the input as a binary mask z, dropping a number of channels. This masked input X is then fed to the DSF module which re-weights the received channels with an attention mechanism, computing a weight matrix W and bias b that are multiplied with and added to the masked input: X = W X + b [8]. Finally, the original classifier is then applied to the resulting signal to obtain a prediction y.…”

Section: B Dynamic Feature Selectionmentioning

confidence: 99%

“…Thus, the question arises how we can make a single network be able to cope as efficiently as possible when multiple input sets are possible. We tackled this issue by extending our network with the Dynamic Spatial Filtering (DSF) proposed by Banville et al [8]. The idea of DSF is to re-weight the M input channels using an attention layer.…”

Section: Dealing With Different Channel Subsetsmentioning

confidence: 99%

“…The usage of this dynamic channel selection means that the inference model will be presented with different channel subsets for different inputs, as if channels were randomly missing. We show that applying dynamic spatial filtering (DSF) [8] to the masked input to re-weight the channels helps the inference model become more robust against the missing of channels and improves performance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bandwidth-Efficient Distributed Neural Network Architectures With Application to Neuro-Sensor Networks

Strypsteen

Member

2023

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

We propose a dynamic sensor selection approach for deep neural networks (DNNs), which is able to derive an optimal sensor subset selection for each specific input sample instead of a fixed selection for the entire dataset. This dynamic selection is jointly learned with the task model in an end-to-end way, using the Gumbel-Softmax trick to allow the discrete decisions to be learned through standard backpropagation. We then show how we can use this dynamic selection to increase the lifetime of a wireless sensor network (WSN) by imposing constraints on how often each node is allowed to transmit. We further improve performance by including a dynamic spatial filter that makes the task-DNN more robust against the fact that it now needs to be able to handle a multitude of possible node subsets. Finally, we explain how the selection of the optimal channels can be distributed across the different nodes in a WSN. We validate this method on a use case in the context of body-sensor networks, where we use real electroencephalography (EEG) sensor data to emulate an EEG sensor network. We analyze the resulting trade-offs between transmission load and task accuracy.

show abstract

“…For example, in polysomnography, changes in the subject's position during sleep might result in a loss of contact between several electrodes and the scalp. Beyond sleep applications, the spread of mobile wearable EEG devices raises new challenges, as they are more prone to noise and missing channels [43]. Finally, the EEG and machine learning communities consider with great interest the question of transferability across datasets, which raises major challenges regarding inconsistent numbers of channels or channels ordering.…”

Section: Rationale Of the Transformationsmentioning

confidence: 99%

Data augmentation for learning predictive models on EEG: a systematic comparison

Rommel,

Paillard,

Moreau

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

The use of deep learning for electroencephalography (EEG) classification tasks has been rapidly growing in the last years, yet its application has been limited by the relatively small size of EEG datasets. Data augmentation, which consists in artificially increasing the size of the dataset during training, has been a key ingredient to obtain state-of-the-art performances across applications such as computer vision or speech. While a few augmentation transformations for EEG data have been proposed in the literature, their positive impact on performance across tasks remains elusive. In this work, we propose a unified and exhaustive analysis of the main existing EEG augmentations, which are compared in a common experimental setting. Our results highlight the best data augmentations to consider for sleep stage classification and motor imagery brain computer interfaces, showing predictive power improvements greater than 10% in some cases.

show abstract

Robust learning from corrupted EEG with dynamic spatial filtering

Cited by 30 publications

References 70 publications

Spectral Representation of EEG Data using Learned Graphs with Application to Motor Imagery Decoding

Spectral Representation of EEG Data using Learned Graphs with Application to Motor Imagery Decoding

Bandwidth-Efficient Distributed Neural Network Architectures With Application to Neuro-Sensor Networks

Data augmentation for learning predictive models on EEG: a systematic comparison

Contact Info

Product

Resources

About