Search

Objective. Supervised learning paradigms are often limited by the amount of labeled data that is available. This phenomenon is particularly problematic in clinically-relevant data, such as electroencephalography (EEG), where labeling can be costly in terms of specialized expertise and human processing time. Consequently, deep learning architectures designed to learn on EEG data have yielded relatively shallow models and performances at best similar to those of traditional feature-based approaches. However, in most situations, unlabeled data is available in abundance. By extracting information from this unlabeled data, it might be possible to reach competitive performance with deep neural networks despite limited access to labels. Approach. We investigated self-supervised learning (SSL), a promising technique for discovering structure in unlabeled data, to learn representations of EEG signals. Specifically, we explored two tasks based on temporal context prediction as well as contrastive predictive coding on two clinically-relevant problems: EEG-based sleep staging and pathology detection. We conducted experiments on two large public datasets with thousands of recordings and performed baseline comparisons with purely supervised and hand-engineered approaches. Main results. Linear classifiers trained on SSL-learned features consistently outperformed purely supervised deep neural networks in low-labeled data regimes while reaching competitive performance when all labels were available. Additionally, the embeddings learned with each method revealed clear latent structures related to physiological and clinical phenomena, such as age effects. Significance. We demonstrate the benefit of self-supervised learning approaches on EEG data. Our results suggest that SSL may pave the way to a wider use of deep learning models on EEG data., 32 pages, 9 figures

Supervised learning paradigms are often limited by the amount of labeled data that is available. This phenomenon is particularly problematic in clinically-relevant data, such as electroencephalography (EEG), where labeling can be costly in terms of specialized expertise and human processing time. Consequently, deep learning architectures designed to learn on EEG data have yielded relatively shallow models and performances at best similar to those of traditional feature-based approaches. However, in most situations, unlabeled data is available in abundance. By extracting information from this unlabeled data, it might be possible to reach competitive performance with deep neural networks despite limited access to labels. Here we report results using self-supervised learning (SSL), a promising technique for discovering structure in unlabeled data, to learn representations of EEG signals. Specifically, we consider a contrastive approach and provide results on two clinically-relevant problems: EEG-based sleep staging and pathology detection. Results report that linear classifiers trained on SSL-learned features consistently outperformed purely supervised deep neural networks in low-labeled data regimes while reaching competitive performance when all labels were available. Our results suggest that self-supervision may pave the way to a wider use of deep learning models on EEG data.

International audience; The supervised learning paradigm is limited by the cost - and sometimes the impracticality - of data collection and labeling in multiple domains. Self-supervised learning, a paradigm which exploits the structure of unlabeled data to create learning problems that can be solved with standard supervised approaches, has shown great promise as a pretraining or feature learning approach in fields like computer vision and time series processing. In this work, we present self-supervision strategies that can be used to learn informative representations from multivariate time series. One successful approach relies on predicting whether time windows are sampled from the same temporal context or not. As demonstrated on a clinically relevant task (sleep scoring) and with two electroencephalography datasets, our approach outperforms a purely supervised approach in low data regimes, while capturing important physiological information without any access to labels.; Le paradigme de l’apprentissage supervisé est limité par le coût - et parfois l’impraticabilité - de la collecte de données et de l’étiquetage dans de multiples domaines. L'apprentissage auto-supervisé, un paradigme qui exploite la structure de données non étiquetées pour créer des problèmes d'apprentissage qui peuvent être résolus avec des approches supervisées standard, s'est révélé très prometteur en tant qu'approche de pré-entraînement ou d'apprentissage de traits caractéristiques dans des domaines tels que la vision par ordinateur et le traitement de séries temporelles. Dans ce travail, nous présentons des stratégies d'auto-supervision pouvant être utilisées pour apprendre des représentations informatives à partir de séries temporelles multivariées. Une approche fructueuse consiste à prédire si des fenêtres temporelles sont échantillonnées dans le même contexte temporel ou non. Comme le démontre une tâche cliniquement pertinente (classification des stades du sommeil) et avec deux jeux de données d'électroencéphalographie, notre approche surpasse une approche purement supervisée dans des régimes de données faibles, tout en capturant des informations physiologiques importantes sans accès aux étiquettes.

Context. Electroencephalography (EEG) is a complex signal and can require several years of training, as well as advanced signal processing and feature extraction methodologies to be correctly interpreted. Recently, deep learning (DL) has shown great promise in helping make sense of EEG signals due to its capacity to learn good feature representations from raw data. Whether DL truly presents advantages as compared to more traditional EEG processing approaches, however, remains an open question. Objective. In this work, we review 154 papers that apply DL to EEG, published between January 2010 and July 2018, and spanning different application domains such as epilepsy, sleep, brain–computer interfacing, and cognitive and affective monitoring. We extract trends and highlight interesting approaches from this large body of literature in order to inform future research and formulate recommendations. Methods. Major databases spanning the fields of science and engineering were queried to identify relevant studies published in scientific journals, conferences, and electronic preprint repositories. Various data items were extracted for each study pertaining to (1) the data, (2) the preprocessing methodology, (3) the DL design choices, (4) the results, and (5) the reproducibility of the experiments. These items were then analyzed one by one to uncover trends. Results. Our analysis reveals that the amount of EEG data used across studies varies from less than ten minutes to thousands of hours, while the number of samples seen during training by a network varies from a few dozens to several millions, depending on how epochs are extracted. Interestingly, we saw that more than half the studies used publicly available data and that there has also been a clear shift from intra-subject to inter-subject approaches over the last few years. About of the studies used convolutional neural networks (CNNs), while used recurrent neural networks (RNNs), most often with a total of 3–10 layers. Moreover, almost one-half of the studies trained their models on raw or preprocessed EEG time series. Finally, the median gain in accuracy of DL approaches over traditional baselines was across all relevant studies. More importantly, however, we noticed studies often suffer from poor reproducibility: a majority of papers would be hard or impossible to reproduce given the unavailability of their data and code. Significance. To help the community progress and share work more effectively, we provide a list of recommendations for future studies and emphasize the need for more reproducible research. We also make our summary table of DL and EEG papers available and invite authors of published work to contribute to it directly. A planned follow-up to this work will be an online public benchmarking portal listing reproducible results.

With the growing complexity of various text-to-speech systems, it is becoming more important to understand the underlying perceptual and judgement processes that drive user Quality-of-Experience (QoE) perception. Typical QoE assessment techniques, such as listening tests with self-report ratings, are useful but provide limited insight into these underlying processes. Recent advances in neuroimaging and physiological monitoring technologies, however, have opened new doors and allowed us to better understand and measure QoE perception. In this paper, we explore the use of two neuroimaging techniques, namely electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), to better understand neuronal and cerebral haemodynamic changes resultant from synthesized speech of varying quality. Neural correlates of several QoE dimensions were derived and validated on the publicly available PhySyQX database. Fusion of EEG, fNIRS, and fNIRS-derived physiological parameters, combined with conventional features extracted from the synthesized speech signal showed to accurately represent several QoE dimensions, including those related to listener affective states. It is hoped that these findings will help researchers build better instrumental QoE models that incorporate technological, contextual, and human influence factors.

Context. In recent years, hybrid brain-computer interfaces (hBCIs) have proven to be a promising path towards practical brain-computer interfacing. These hybrid interfaces capitalize on the concurrent recording of various physiological signals, or of the elicitation of more than one mental process, to increase the number of possible input commands and achieve more flexible and robust systems. Although hBCIs have previously been reviewed in some articles, a more recent, and complete survey of the literature is missing to lay the foundations for further research. Objective. This work aims at systematically reviewing recent articles on the topic of non-invasive hBCIs, to comprehensively identify the current trends, limitations and challenges that these studies report. Methods. Three major databases covering the fields of science and engineering were queried. From these and others sources, 55 journal articles from 2008 to November 2014 were selected and analyzed. Results. Twenty-two items were investigated to o...

Based on recent electroencephalography (EEG) and near-infrared spectroscopy (NIRS) studies that showed that tasks such as motor imagery and mental arithmetic induce specific neural response patterns, we propose a hybrid brain-computer interface (hBCI) paradigm in which EEG and NIRS data are fused to improve binary classification performance. We recorded simultaneous NIRS-EEG data from nine participants performing seven mental tasks (word generation, mental rotation, subtraction, singing and navigation, and motor and face imagery). Classifiers were trained for each possible pair of tasks using (1) EEG features alone, (2) NIRS features alone, and (3) EEG and NIRS features combined, to identify the best task pairs and assess the usefulness of a multimodal approach. The NIRS-EEG approach led to an average increase in peak kappa of 0.03 when using features extracted from one-second windows (equivalent to an increase of 1.5% in classification accuracy for balanced classes). The increase was much stronger (0.20, corresponding to an 10% accuracy increase) when focusing on time windows of high NIRS performance. The EEG and NIRS analyses further unveiled relevant brain regions and important feature types. This work provides a basis for future NIRS-EEG hBCI studies aiming to improve classification performance toward more efficient and flexible BCIs.