Deep long short-term memory structures model temporal dependencies improving cognitive workload estimation

Hefron, Ryan G.; Borghetti, Brett J.; Christensen, James C.; Kabban, Christine M. Schubert

doi:10.1016/j.patrec.2017.05.020

Cited by 114 publications

(56 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We use the same hyperparameters for RF training as Schiratti et al 38 , which is 100 estimators and a maximum depth of 4. A variety of commonly used features 39 were extracted from each EEG clip to support the analysis; these include time domain features such mean, variance, skewness, kurtosis, total signal area, peak-to-peak, number of zerocrossings, and decorrelation times; frequency domain features such as total energy spectrum, energy percentage across fundamental rhythmic bands (extracted using the Discrete Fourier Transform), and the coefficients from the Discrete Wavelet Transform; brain connectivity features such as the maximal absolute cross-correlation value to measure similarity between electrodes; and local and global electrode graph measures [38][39][40][41] .…”

Section: Discussionmentioning

confidence: 99%

Weak supervision as an efficient approach for automated seizure detection in electroencephalography

et al. 2020

View full text Add to dashboard Cite

Automated seizure detection from electroencephalography (EEG) would improve the quality of patient care while reducing medical costs, but achieving reliably high performance across patients has proven difficult. Convolutional Neural Networks (CNNs) show promise in addressing this problem, but they are limited by a lack of large labeled training datasets. We propose using imperfect but plentiful archived annotations to train CNNs for automated, real-time EEG seizure detection across patients. While these weak annotations indicate possible seizures with precision scores as low as 0.37, they are commonly produced in large volumes within existing clinical workflows by a mixed group of technicians, fellows, students, and board-certified epileptologists. We find that CNNs trained using such weak annotations achieve Area Under the Receiver Operating Characteristic curve (AUROC) values of 0.93 and 0.94 for pediatric and adult seizure onset detection, respectively. Compared to currently deployed clinical software, our model provides a 31% increase (18 points) in F1-score for pediatric patients and a 17% increase (11 points) for adult patients. These results demonstrate that weak annotations, which are sustainably collected via existing clinical workflows, can be leveraged to produce clinically useful seizure detection models.

show abstract

Section: Discussionmentioning

confidence: 99%

Weak supervision as an efficient approach for automated seizure detection in electroencephalography

et al. 2020

View full text Add to dashboard Cite

show abstract

“…The application areas where MF detection using EEG has been applied are driving, mental load tasks [51], [63], [68], [69], [82], [83], [88], safety-critical tasks [59], [62], [77], train piloting [71], and aircraft piloting [72]. Fig.…”

Section: Discussion and Future Trendsmentioning

confidence: 99%

“…Wavelet packet transform (WPT) is the simplest wavelet transform, disregarding boundary treatments in the original signal [85]. Discrete wavelet transforms (DWT) are most commonly used due to their more robust performance [86]- [88].…”

Section: Time-frequency Domain Methods 1)mentioning

confidence: 99%

“…An MLP classifier was used in several works with good results [80], [84], [92] although more complex and robust neural network model was also used. Deep-Long short-term memory (LSTM) was used to construct a dynamic classifier to account for the fact that MF has a very important temporal factor, since it builds up over time [88]. Channel-wise, a convolutional neural network was used to automatically extract a complex feature from timeseries data [90].…”

Section: Time-frequency Domain Methods 1)mentioning

confidence: 99%

See 1 more Smart Citation

Using EEG for Mental Fatigue Assessment: A Comprehensive Look Into the Current State of the Art

Monteiro

Skourup²,

Zhang

2019

IEEE Trans. Human-Mach. Syst.

View full text Add to dashboard Cite

This paper provides a brief survey of recent developments on the use of electroencephalogram (EEG) sensors for detecting mental fatigue (MF) in human operators during tasks involving human-machine interaction. This research topic has received much attention since there is a consensus among experts on the increasing relation between human failure and accidents in safety-critical tasks. MF is one of the most influential aspects leading to human failure and the most reliable way to assess it is using operator's physiological data, especially EEG. In the past few decades, hundreds of publications have explored the use of EEG alone or together with other objective and subjective measures for assessing MF, drowsiness, and tiredness in human operators. With recent improvements in data preprocessing, feature extraction, and classification algorithms, the monitoring and mitigation of MF in real time has become a reality. This trend is mainly due to the increasing use of machine learning techniques. This paper provides a comprehensive look at the current state of the art in the field of MF detection using EEG, identifying the currently used technique, algorithms, and methods and possible trends and promising areas for further research. The paper is concluded by suggesting a kernel partial least squares discrete-output linear regression (KPLS-DLR) based model as an all-around good option for an MF assessment system.

show abstract

“…We apply several competitive methods on our dataset for comparison, the methods and [12]. LSTM: The deep long short-term memory (LSTM) architecture for binary classification in [30], which consists of two LSTM layers and a sigmoid activation function. CFCNN: Our method.…”

Section: A Overall Performancementioning

confidence: 99%

A Coincidence-Filtering-Based Approach for CNNs in EEG-Based Recognition

Gao

Yang

et al. 2020

IEEE Trans. Ind. Inf.

View full text Add to dashboard Cite

Electroencephalogram (EEG), obtained by wearable devices, can realize effective human health monitoring. Traditional methods based on artificially-designed features have achieved valid results in EEG-based recognition, and numerous studies start to apply deep learning techniques in this area. In this paper, we propose a coincidence filtering-based method to build a connection between artificial features-based methods and convolutional neural networks (CNNs), and design CNNs through simulating the information extraction pattern of artificial features-based methods. Based on this method, we propose a novel, simple, and effective CNNs structure for EEG-based classification. We implement two experiments to obtain EEG data, and perform experiments based on the two health monitoring tasks. The results illustrate that the proposed network can achieve a prominent average accuracy on the emotion recognition and fatigue driving detection task. Due to its generality, the proposed framework design of CNNs is expected to be useful for broader applications in health monitoring areas.

show abstract

Deep long short-term memory structures model temporal dependencies improving cognitive workload estimation

Cited by 114 publications

References 19 publications

Weak supervision as an efficient approach for automated seizure detection in electroencephalography

Weak supervision as an efficient approach for automated seizure detection in electroencephalography

Using EEG for Mental Fatigue Assessment: A Comprehensive Look Into the Current State of the Art

A Coincidence-Filtering-Based Approach for CNNs in EEG-Based Recognition

Contact Info

Product

Resources

About