Adaptive transfer learning for EEG motor imagery classification with deep Convolutional Neural Network

Zhang, Kaishuo; Robinson, Neethu; Lee, Seong Whan; Guan, Cuntai

doi:10.1016/j.neunet.2020.12.013

Cited by 185 publications

(158 citation statements)

References 80 publications

Supporting

Mentioning

123

Contrasting

Order By: Relevance

“…Slightly lower classification accuracies for continuous fist and feet motor execution and imagery from the EEGMMI dataset can be attributed to the leave-one-participant-out LDA classifier training scheme as for each participant, the classifier was trained with the data from the remaining participants, whereas, for finger tapping dataset, the classifier was trained for each participant using their own data with 10 x 10 fold cross-validation. The classification accuracies using broadband LRTC are comparable to the accuracies obtained in the BCI literature (Ibáñez et al, 2014;Lew et al, 2014;Lopez-Larraz et al, 2014;Xu et al, 2014;Padfield et al, 2019;Zhang et al, 2021). Thus, broadband LRTCs can be used as features independently for application in BCI.…”

Section: Discussionsupporting

confidence: 71%

Dynamics of Long-Range Temporal Correlations in Broadband EEG During Different Motor Execution and Imagery Tasks

2021

View full text Add to dashboard Cite

Brain activity is composed of oscillatory and broadband arrhythmic components; however, there is more focus on oscillatory sensorimotor rhythms to study movement, but temporal dynamics of broadband arrhythmic electroencephalography (EEG) remain unexplored. We have previously demonstrated that broadband arrhythmic EEG contains both short- and long-range temporal correlations that change significantly during movement. In this study, we build upon our previous work to gain a deeper understanding of these changes in the long-range temporal correlation (LRTC) in broadband EEG and contrast them with the well-known LRTC in alpha oscillation amplitude typically found in the literature. We investigate and validate changes in LRTCs during five different types of movements and motor imagery tasks using two independent EEG datasets recorded with two different paradigms—our finger tapping dataset with single self-initiated asynchronous finger taps and publicly available EEG dataset containing cued continuous movement and motor imagery of fists and feet. We quantified instantaneous changes in broadband LRTCs by detrended fluctuation analysis on single trial 2 s EEG sliding windows. The broadband LRTC increased significantly (p < 0.05) during all motor tasks as compared to the resting state. In contrast, the alpha oscillation LRTC, which had to be computed on longer stitched EEG segments, decreased significantly (p < 0.05) consistently with the literature. This suggests the complementarity of underlying fast and slow neuronal scale-free dynamics during movement and motor imagery. The single trial broadband LRTC gave high average binary classification accuracy in the range of 70.54±10.03% to 76.07±6.40% for all motor execution and imagery tasks and hence can be used in brain–computer interface (BCI). Thus, we demonstrate generalizability, robustness, and reproducibility of novel motor neural correlate, the single trial broadband LRTC, during different motor execution and imagery tasks in single asynchronous and cued continuous motor-BCI paradigms and its contrasting behavior with LRTC in alpha oscillation amplitude.

show abstract

Section: Discussionsupporting

confidence: 71%

Dynamics of Long-Range Temporal Correlations in Broadband EEG During Different Motor Execution and Imagery Tasks

2021

View full text Add to dashboard Cite

show abstract

“…With recent release of large scale EEG datasets (e.g. Cho et al, 2017; Lee et al 2019), there have been more attempts on employing DL models on signals from large number of participants (e.g., Stieger et al, 2020; Zhang et al, 2021; Ko et al, 2020; Mane et al, 2020), showing the relevance and timeliness of this study in the BCI field. Although these studies report the same conclusion for superiority of the DL approach in MI-BCI classification, their methodology and approach in building the DL model is different from our study.…”

Section: Discussionmentioning

confidence: 92%

“…Mane et al (2020) and Ko et al (2020) focused on feature representations in the model; Mane et al (2020) employed Filter-Bank CNN to decompose data into multiple frequency bands and extract spatially discriminative patterns in each band, and Ko et al (2020) applied a Multi-Scale Neural Network to exploit spatio-spectral-temporal features for all BCI paradigms. Zhang et al (2021) focused on transfer learning and employed a CNN model to develop a subject-independent classifier. Therefore, while our study pursues a similar goal, it dissociates itself from past research by conducting a statistically supported subject-wise comparison between the DL and ML approaches and also providing evidence for suitability of the DL approach for inefficient BCI users.…”

Section: Discussionmentioning

confidence: 99%

“…Past research has already established the effectiveness of the DL approach, especially Convolutional Neural Network (CNN), in classification of MI-EEG (Tang et al, 2017; Gao et al, 2018; Sakhavi et al, 2015; Li et al, 2020; Dai et al, 2019; Tayeb et al, 2019; Stieger et al, 2020; Zhang et al, 2021; Ko et al, 2020; Mane et al, 2020). The advantages of CNN model include handling raw data without any feature engineering process, facilitating end-to-end learning and requiring lesser parameters than other deep neural networks (Shen et al, 2017; Albawi et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The Promise of Deep Learning for BCIs: Classification of Motor Imagery EEG using Convolutional Neural Network

Leeuwis

Alimardani

2021

Preprint

View full text Add to dashboard Cite

Motor Imagery (MI) is a mental process by which an individual rehearses body movements without actually performing physical actions. Motor Imagery Brain-Computer Interfaces (MI-BCIs) are AI-driven systems that capture brain activity patterns associated with this mental process and convert them into commands for external devices. Traditionally, MI-BCIs operate on Machine Learning (ML) algorithms, which require extensive signal processing and feature engineering to extract changes in sensorimotor rhythms (SMR). However, in recent years, Deep Learning (DL) models have gained popularity for EEG classification as they provide a solution for automatic extraction of spatio-temporal features in the signals. In this study, EEG signals from 54 subjects who performed a MI task of left- or right-hand grasp was employed to compare the performance of two MI-BCI classifiers; a ML approach vs. a DL approach. In the ML approach, Common Spatial Patterns (CSP) was used for feature extraction and then Linear Discriminant Analysis (LDA) model was employed for binary classification of the MI task. In the DL approach, a Convolutional Neural Network (CNN) model was constructed on the raw EEG signals. The mean classification accuracies achieved by the CNN and CSP+LDA models were 69.42% and 52.56%, respectively. Further analysis showed that the DL approach improved the classification accuracy for all subjects within the range of 2.37 to 28.28% and that the improvement was significantly stronger for low performers. Our findings show promise for employment of DL models in future MI-BCI systems, particularly for BCI inefficient users who are unable to produce desired sensorimotor patterns for conventional ML approaches.

show abstract

“…To effectively classify patients infected with and without COVID-19, the deep convolution network [ 59 , 60 ] includes seven convolution layers, three pooling layers, and three full connection layers. Among them,

represents the input image layer,

represents the convolution layer,

represents the maximum pooling layer,

represents the full connection layer,

represents the feature map, and

represents the output result layer.…”

Section: Deep Learning Network Based On Covid-19 Ct Imagementioning

confidence: 99%

COVID-19 deep classification network based on convolution and deconvolution local enhancement

Fang

Wang

2021

Computers in Biology and Medicine

View full text Add to dashboard Cite

Computer Tomography (CT) detection can effectively overcome the problems of traditional detection of Corona Virus Disease 2019 (COVID-19), such as lagging detection results and wrong diagnosis results, which lead to the increase of disease infection rate and prevalence rate. The novel coronavirus pneumonia is a significant difference between the positive and negative patients with asymptomatic infections. To effectively improve the accuracy of doctors' manual judgment of positive and negative COVID-19, this paper proposes a deep classification network model of the novel coronavirus pneumonia based on convolution and deconvolution local enhancement. Through convolution and deconvolution operation, the contrast between the local lesion region and the abdominal cavity of COVID-19 is enhanced. Besides, the middle-level features that can effectively distinguish the image types are obtained. By transforming the novel coronavirus detection problem into the region of interest (ROI) feature classification problem, it can effectively determine whether the feature vector in each feature channel contains the image features of COVID-19. This paper uses an open-source COVID-CT dataset provided by Petuum researchers from the University of California, San Diego, which is collected from 143 novel coronavirus pneumonia patients and the corresponding features are preserved. The complete dataset (including original image and enhanced image) contains 1460 images. Among them, 1022 (70%) and 438 (30%) are used to train and test the performance of the proposed model, respectively. The proposed model verifies the classification precision in different convolution layers and learning rates. Besides, it is compared with most state-of-the-art models. It is found that the proposed algorithm has good classification performance. The corresponding sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and precision are 0.98, 0.96, 0.98, and 0.97, respectively.

show abstract

Adaptive transfer learning for EEG motor imagery classification with deep Convolutional Neural Network

Cited by 185 publications

References 80 publications

Dynamics of Long-Range Temporal Correlations in Broadband EEG During Different Motor Execution and Imagery Tasks

Dynamics of Long-Range Temporal Correlations in Broadband EEG During Different Motor Execution and Imagery Tasks

The Promise of Deep Learning for BCIs: Classification of Motor Imagery EEG using Convolutional Neural Network

COVID-19 deep classification network based on convolution and deconvolution local enhancement

Contact Info

Product

Resources

About