The DOI matches its Crossref record; 52 of the 53 references it cites were found in a registry or an index; 24 of the 47 sources its claims cite were checked for support.
A Cascade xDAWN EEGNet Structure for Unified Visual-Evoked Related Potential Detection
Open accessAbstract
Visual-based brain-computer interface (BCI) enables people to communicate with others by spelling words from the brain and helps professionals recognize targets in large numbers of images. P300 signals evoked by different types of stimuli, such as words or images, may vary significantly in terms of both amplitude and latency. A unified approach is required to detect variable P300 signals, which facilitates BCI applications, as well as deepens the understanding of the P300 generation mechanism. In this study, our proposed approach involves a cascade network structure that combines xDAWN and classical EEGNet techniques. This network is designed to classify target and non-target stimuli in both P300 speller and rapid serial visual presentation (RSVP) paradigms. The proposed approach is capable of recognizing more symbols with fewer repetitions (up to 5 rounds) compared to other models while possessing a better information transfer rate (ITR) as demonstrated on Dataset II (17.22 bits/min in the second repetition round) of BCI Competition III. Additionally, our approach has the highest unweighted average recall (UAR) performance for both 5 Hz ( 0.8134±0.0259 ) and 20 Hz ( 0.6527±0.0321 ) RSVP. The results show that the cascade network structure has better performance between both the P300 Speller and RSVP paradigms, manifesting that such a cascade structure is robust enough for dealing with P300-related signals (source code is available at https://github.com/embneural/Cascade-xDAWN-EEGNet-for-ERP-Detection).
Overview
CoChat Summary
TL;DR
The paper proposes a cascade xDAWN-EEGNet model for unified detection of visual-evoked related potentials in P300 speller and RSVP paradigms. It reports improved ITR and symbol recognition on BCI Competition III dataset II and the highest UAR on 5 Hz and 20 Hz RSVP compared with several baselines.
Key findings
- On BCI Competition III dataset II, the proposed model reached an ITR of 17.22 bits/min in the second repetition round and recognized more symbols with fewer repetitions (up to 5 rounds) than other models; paired tests showed better symbol recognition than most models (p<0.05) except DeepConvNet (p>0.05) on dataset II.
- On RSVP, the proposed method achieved the highest UAR: 0.8134±0.0259 at 5 Hz and 0.6527±0.0321 at 20 Hz; at 5 Hz, EEGNet reached 0.7823±0.0201 (p<0.05), and improvements over other models were p<0.0001 for both 5 Hz and 20 Hz.
- xDAWN spatial filtering enhanced the P300 pattern; optimal settings were 8 xDAWN filters and Mixup alpha 0.3 for the P300 speller, and 4 xDAWN filters and Mixup alpha 0.4 for RSVP. For dataset II, 8 filters improved both subjects; for dataset IIb, xDAWN filters decreased the symbol recognition rate.
- For RSVP, increasing xDAWN filters improved 5 Hz detection accuracy and larger Mixup alpha led to larger UAR; for 20 Hz, only 4 xDAWN filters gave a slight improvement over EEGNet, otherwise performance decreased, though Mixup still significantly improved UAR.
Methods
- Datasets: P300 speller data from BCI Competition II dataset IIb (one participant) and BCI Competition III dataset II (subjects A and B), recorded with BCI2000 using 64 electrodes at 240 Hz; RSVP data with 200 visual objects from 16 adults (5 females; age 18–38 years), 5 Hz and 20 Hz sessions, 64 electrodes, recorded at 1000 Hz and downsampled to 250 Hz.
- Preprocessing and model: P300 trials were cropped 800 ms after stimulus onset, detrended, filtered with a 30 Hz low-pass Chebyshev zero-phase filter, and downsampled by half to 96 time samples; RSVP trials used 0–1000 ms, detrending, the same filter, and 250 time samples. xDAWN spatial filters were applied before EEGNet, followed by z-score normalization.
- Training and evaluation: PyTorch on a GeForce RTX 2080 Ti, batch size 64, Adam with initial learning rate 0.001 and exponential decay 0.96, focal loss with weights and gamma=2, and Mixup augmentation; P300 trained for 80 epochs, RSVP evaluated with 3-fold cross-validation. Metrics were symbol recognition accuracy and ITR for P300, and UAR for RSVP; paired t-tests used 30 pairs for dataset II, 15 pairs for IIb, and n=16 for RSVP comparisons against SWFP, ESVMs, DeepConvNet, ShallowConvNet, BN3, CNN-RG-MINMAX-CNN, ST-CapsNet, DCPM, and EEGNet.
Limitations
- Stated: for dataset IIb, xDAWN filters decreased the symbol recognition rate, so the number of xDAWN filters must be carefully selected according to the specific dataset and task.
- Stated: DeepConvNet showed significant variance across subjects, attributed to its large model parameters and sensitivity to hyperparameters such as the gamma value in focal loss, leading to convergence difficulty; SWFP performed worst among all models.
- Evident from the text: P300 speller evaluation relies on very few subjects (dataset II: two subjects; IIb: one subject), and the RSVP task had no dedicated test set, using 3-fold cross-validation, which may limit generalizability.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
23 Valid · 0 Not valid · 11 Not checked · 71 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“However, deep learning models often require a large number of samples for better learning, as they lack domain knowledge about the data[25].”
What the sources say
- Supportedcites [25] source
The Introduction (page not provided) describes deep networks as heavily dependent on large amounts of labeled data and says incorporating prior knowledge can enable learning with fewer labeled examples. The Discussion adds that networks do not always infer and generalize domain knowledge from training data. (LLM, conf=0.91)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“However, this averaging method is time-consuming and inefficient [14].”
What the sources say
- Supportedcites [14] source
The Introduction (page not available in our copy) explicitly says that collecting, aggregating, and averaging P300 data from multiple trials is time-consuming and inefficient. (LLM, conf=0.95)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Moreover, a multi-feature predictor that includes multiple RSVP features has demonstrated that it can accurately predict P300 speller performance, outperforming single-feature predictors [21].”
What the sources say
- Supportedcites [21] source
The Methods section describes a predictor combining multiple RSVP features. The Conclusion reports that these multi-feature predictors predicted offline P300 speller performance better than any single-feature predictor. (LLM, conf=0.93)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Additionally, our method has the highest unweighted average recall (UAR) performance for both 5 Hz (0.8134±0.0259) and 20 Hz (0.6527±0.0321) RSVP.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The proposed method is capable of recognizing more symbols with fewer repetitions (up to 5 rounds) compared to other models while demonstrating a better information transfer rate (ITR) on dataset II (achieved 17.22 bits/min in the second repetition round) of BCI Competition III.”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
35 more claims
Create a free account to see all 105 claims and every source verdict.
- Probably not a substantial claim. 35
- It may state two separate claims as one. 29
- The quote only partly states this claim. 15
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A. Dataset Description
“the RSVP dataset [35] employed a stimulus set of 200 visual objects”
§ REFERENCES
“- [35] A. G. Howard et al. , 'MobileNets,' arXiv preprint arXiv:1704.04861 , 2017.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ C. Performance of ITR For P300 Speller
“The ITRs of each model on datasets II and IIb were plotted in Fig. 3 to visually compare the speed of symbol spelling.”
§ A. Performance of Symbol Recognition For P300 Speller
“Fig. 3 Effect of different number of xDAWN filters on the symbol recognition rate. Subplots (a), (b), and (c) represent the symbol recognition rates of subjects A and B in dataset II and dataset IIb, respectively.”
Citations
Cites
PyTorch: An Imperative Style, High-Performance Deep Learning Library
Paszke, A., Gross, S., Massa, F., Lerer, A., Bradbury, J., Chanan, G., Killeen, T., Lin, Z., Gimelshein, N., Antiga, L., et al. (2019) · arXiv · doi:10.48550/arxiv.1912.01703 · Cited by 31,473
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015
Focal Loss for Dense Object Detection
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollar, P. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2018.2858826 · Cited by 9,556
Brain–computer interfaces for communication and control
Wolpaw, J. R., Birbaumer, N., McFarland, D. J., Pfurtscheller, G., & Vaughan, T. M. (2002) · Clinical Neurophysiology · doi:10.1016/s1388-2457(02)00057-3 · Cited by 7,927 · cited 1×
“B CIs are systems that enable people to communicate with their surroundings using brain waves [1], [2].”
Backgroundmixup: Beyond Empirical Risk Minimization
Zhang, H., Cisse, M., Dauphin, Y. N., & Lopez-Paz, D. (2017) · arXiv · doi:10.48550/arxiv.1710.09412 · Cited by 5,831
EEGNet: a compact convolutional neural network for EEG-based brain-computer interfaces
Lawhern, V. J., Solon, A. J., Waytowich, N. R., Gordon, S. M., Hung, C. P., & Lance, B. J. (2018) · Journal of Neural Engineering · doi:10.1088/1741-2552/aace8c · Cited by 5,115 · cited 2×
“Several convolutional neural networks (CNN), such as EEGNet, DeepConvNet, Shallow-ConvNet, and BN3, have been developed for detecting ERP signals [5], [13], [23], [24].”
BackgroundMethodEEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces
Lawhern, V. J., Solon, A. J., Waytowich, N. R., Gordon, S. M., Hung, C. P., & Lance, B. J. (2016) · arXiv · doi:10.48550/arxiv.1611.08024 · Cited by 5,105
Deep learning with convolutional neural networks for EEG decoding and visualization
Schirrmeister, R. T., Springenberg, J. T., Fiederer, L. D. J., Glasstetter, M., Eggensperger, K., Tangermann, M., Hutter, F., Burgard, W., & Ball, T. (2017) · Human Brain Mapping · doi:10.1002/hbm.23730 · Cited by 3,751
A spelling device for the paralysed
Birbaumer, N., Ghanayim, N., Hinterberger, T., Iversen, I., Kotchoubey, B., Kübler, A., Perelmouter, J., Taub, E., & Flor, H. (1999) · Nature · doi:10.1038/18581 · Cited by 1,577 · cited 1×
“They are particularly helpful for individuals with disabilities, such as those with amyotrophic lateral sclerosis (ALS), who may rely on a BCI to interact with others [3], [4].”
BackgroundSingle-trial analysis and classification of ERP components — A tutorial
Blankertz, B., Lemm, S., Treder, M., Haufe, S., & Müller, K.-R. (2011) · NeuroImage · doi:10.1016/j.neuroimage.2010.06.048 · Cited by 1,138 · cited 2×
“Traditional machine learning methods, such as support vector machines (SVM), discriminant analysis, and common spatial pattern algorithms, were first applied [42], [43], [46], [49], [50].”
Method
Cited by
Ji, Z., Li, S., Zhang, H., Chen, C., Xu, Q., Li, J., & Wang, H. (2025) · Brain-Apparatus Communication A Journal of Bacomics · doi:10.1080/27706710.2025.2489396 · Cited by 17
Self-citationChen, H., Chan, I. Y. S., Dong, Z., & Samuel, T.-A. (2025) · Advanced Engineering Informatics · doi:10.1016/j.aei.2025.103555 · Cited by 15
A framework for Interpretable deep learning in cross-subject detection of event-related potentials
Jalilpour, S., & Müller-Putz, G. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.109642 · Cited by 11
Lightweight deep learning models for EEG decoding: a review
Li, Y., Chen, E., Xiao, X., Xu, M., & Ming, D. (2025) · Journal of Neural Engineering · doi:10.1088/1741-2552/ae2717 · Cited by 12
“In 2024, Wang et al [78] introduced a cascaded architecture that integrates the xDAWN spatial filtering algorithm with the EEGNet framework for classification tasks in P300 and rapid serial visual presentation (RSVP) paradigms.”
InfluentialPark, S., Ha, J., & Kim, L. (2024) · Frontiers in Physiology · doi:10.3389/fphys.2024.1325784 · Cited by 5
“…number of stimuli in a short time, can elicit strong EPs caused by a stimulus paradigm consisting of targets and nontargets (Acqualagna and Blankertz, 2013), and has been utilized in many visual-based brain-computer interface studies (Zhang H. et al., 2022; Wang et al., 2023b; Wang et al., 2024).”
Method
Metadata
- Authors
- Wang, H., Wang, Z., Sun, Y., Yuan, Z., Xu, T., & Li, J.
- Affiliations
- University of Essex (GB)
- Corresponding author
- Wang, Hongtao
- Venue
- IEEE Transactions on Neural Systems and Rehabilitation Engineering
- Published
- 2024
- Volume
- 32
- Pages
- 2270-2280
- ISSN
- 1534-4320 (ISSN-L)
- 1558-0210
- OpenAlex
- W4399727965
- Cited by
- 20
- References
- 51
- Citations by year
- 2026: 6 · 2025: 11 · 2024: 3
- FWCI
- 4.84
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- IEEE Transactions on Neural Systems and Rehabilitation Engineering (publishedVersion, cc-by)
- PubMed (publishedVersion)
- Open Access at Essex (University of Essex) (submittedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Funders
- Special Projects in Key Fields Supported by the Technology Development Project of Guangdong Province (2020ZDZX3018)
- Wuyi University and Hong Kong and Macao Joint Research Project (2019WGALH16)
- Projects for International Scientific and Technological Cooperation of Guangdong Province (2023A0505050144)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Physical Sciences › Computer Science › Human-Computer Interaction › Gaze Tracking and Assistive Technology
- Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
- Keywords
- P300 detection, visual evoked potentials, cascaded neural networks, EEGNet, P300 speller, rapid serial visual presentation, information transfer rate
- MeSH terms
- Algorithms, Communication Devices for People with Disabilities, Electroencephalography, Evoked Potentials, Visual, Humans, Male, Photic Stimulation, Reproducibility of Results, Neural Networks, Computer, Event-Related Potentials, P300, Brain-Computer Interfaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A Cascade xDAWN EEGNet Structure for Unified Visual-Evoked Related Potential Detection. https://app.cochat.ai/papers/10.1109/tnsre.2024.3415474
- The paper
- Wang, H., Wang, Z., Sun, Y., Yuan, Z., Xu, T., & Li, J. (2024). A Cascade xDAWN EEGNet Structure for Unified Visual-Evoked Related Potential Detection. IEEE Transactions on Neural Systems and Rehabilitation Engineering. https://doi.org/10.1109/tnsre.2024.3415474
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.