ABSTRACT
Population-level modeling can define quantitative measures of individual aging by applying machine learning to large volumes of brain images. These measures of brain age, obtained from the general population, helped characterize disease severity in neurological populations, improving estimates of diagnosis or prognosis. Magnetoencephalography (MEG) and Electroencephalography (EEG) have the potential to further generalize this approach towards prevention and public health by enabling assessments of brain health at large scales in socioeconomically diverse environments. However, more research is needed to define methods that can handle the complexity and diversity of M/EEG signals across diverse real-world contexts. To catalyse this effort, here we propose reusable benchmarks of competing machine learning approaches for brain age modeling. We benchmarked popular classical machine learning pipelines and deep learning architectures previously used for pathology decoding or brain age estimation in 4 international M/EEG cohorts from diverse countries and cultural contexts, including recordings from more than 2500 participants. Our benchmarks were built on top of the M/EEG adaptations of the BIDS standard, providing tools that can be applied with minimal modification on any M/EEG dataset provided in the BIDS format. Our results suggest that, regardless of whether classical machine learning or deep learning was used, the highest performance was reached by pipelines and architectures involving spatially aware representations of the M/EEG signals, leading to R2 scores between 0.60-0.74. Hand-crafted features paired with random forest regression provided robust benchmarks even in situations in which other approaches failed. Taken together, this set of benchmarks, accompanied by open-source software and high-level Python scripts, can serve as a starting point and quantitative reference for future efforts at developing M/EEG-based measures of brain aging. The generality of the approach renders this benchmark reusable for other related objectives such as modeling specific cognitive variables or clinical endpoints.
Subject(s)
Benchmarking , Brain-Computer Interfaces , Algorithms , Brain , Brain Mapping/methods , Electroencephalography/methods , HumansABSTRACT
Building machine learning models using EEG recorded outside of the laboratory setting requires methods robust to noisy data and randomly missing channels. This need is particularly great when working with sparse EEG montages (1-6 channels), often encountered in consumer-grade or mobile EEG devices. Neither classical machine learning models nor deep neural networks trained end-to-end on EEG are typically designed or tested for robustness to corruption, and especially to randomly missing channels. While some studies have proposed strategies for using data with missing channels, these approaches are not practical when sparse montages are used and computing power is limited (e.g., wearables, cell phones). To tackle this problem, we propose dynamic spatial filtering (DSF), a multi-head attention module that can be plugged in before the first layer of a neural network to handle missing EEG channels by learning to focus on good channels and to ignore bad ones. We tested DSF on public EEG data encompassing â¼4000 recordings with simulated channel corruption and on a private dataset of â¼100 at-home recordings of mobile EEG with natural corruption. Our proposed approach achieves the same performance as baseline models when no noise is applied, but outperforms baselines by as much as 29.4% accuracy when significant channel corruption is present. Moreover, DSF outputs are interpretable, making it possible to monitor the effective channel importance in real-time. This approach has the potential to enable the analysis of EEG in challenging settings where channel corruption hampers the reading of brain signals.
Subject(s)
Brain-Computer Interfaces , Electroencephalography , Algorithms , Brain , Electroencephalography/methods , Humans , Machine Learning , Neural Networks, ComputerABSTRACT
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 SSL approaches on EEG data. Our results suggest that self-supervision may pave the way to a wider use of deep learning models on EEG data.
Subject(s)
Electroencephalography , Neural Networks, Computer , Humans , Research Design , Sleep Stages , Supervised Machine LearningABSTRACT
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 [Formula: see text] of the studies used convolutional neural networks (CNNs), while [Formula: see text] 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 [Formula: see text] 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.
Subject(s)
Brain-Computer Interfaces/trends , Brain/physiology , Deep Learning/trends , Electroencephalography/trends , Databases, Factual/trends , Electroencephalography/methods , HumansABSTRACT
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.