Review Papers Claim by Claim

One line summary

The Open access dataset integrating EEG and fNIRS during Stroop tasks provides simultaneous 34 channel EEG at 1000Hz and 20 channel fNIRS at 100Hz from 21 healthy adults performing an event related Stroop paradigm, and is explicitly intended to enable multimodal algorithm development and neurovascular coupling studies Dataset paper description

Formal review and critique Open access dataset integrating EEG and fNIRS during Stroop tasks

Paper The authors provide a publicly available multimodal dataset that records high temporal resolution EEG and high sample rate fNIRS concurrently during an event related Stroop paradigm in 21 healthy adults, and present basic validation analyses showing expected EEG event related potentials and frontal fNIRS activations consistent with a Stroop effect Dataset core details

What the dataset contains and methods quality

• Simultaneous EEG and fNIRS: EEG 34 channels sampled at 1000Hz and fNIRS 20 channels sampled at 100Hz over frontal cortex; event markers for Stroop trials are included enabling event locked analyses Dataset channels and rates
• Participants: 21 healthy adults 9 females 12 males mean age 23.0 ± 2.3 years as the paper reports; demographic balance is modest but sample size is small for between-subject inference beyond methods development Participant demographics
• Task design: event-related Stroop trials with congruent and incongruent conditions enabling classic conflict related contrasts; authors validated data by demonstrating ERP and hemodynamic signatures of the Stroop effect (they do not deeply explore intersubject variability or trial-by-trial coupling) ERP and fNIRS validation statements

Strengths and why this dataset matters

1. High temporal resolution complementarity: EEG provides millisecond scale electrical dynamics whereas fNIRS offers hemodynamic measures and resistance to some motion artifacts enabling joint analysis of neurovascular coupling and multimodal algorithm development fNIRS features and multimodal suitability
2. Open access and preprocessing readiness: Authors include event structure and baseline validations making the dataset suitable for benchmarking multimodal preprocessing algorithms and machine learning model development Dataset public use intent
3. Frontal coverage matches Stroop effect loci: fNIRS whole frontal coverage is appropriate for a prefrontal cognitive conflict task, enabling expected activation contrasts fNIRS frontal coverage

Limitations and blindspots

• Limited sample size and demographic balance: 21 participants is acceptable for methods/dataset papers but limits normative or clinical generalization and estimation of intersubject variability; many multimodal modelling approaches require larger subject pools for robust generalization Sample size and demographics
• Spatial coverage restricted to frontal cortex: dataset is ideal for Stroop but cannot support whole cortex spatial mapping, source localization studies outside frontal regions, or analyses requiring occipital/parietal channels unless combined with other datasets or individualized head models Limited spatial coverage
• fNIRS systemic confounds and preprocessing: fNIRS signals contain systemic physiology components (cardiac respiration Mayer waves) and motion artifacts which must be aggressively addressed; the broader fNIRS literature warns about false positives from systemic signals and recommends short separation channels, physiological regressors and careful detrending fNIRS systemic confounds
• Limited ERP-fNIRS coupling analyses in the paper: the authors present ERP results and activation maps but do not deeply quantify trial-by-trial neurovascular coupling metrics such as latency correlations, crossmodal single trial correlations, or GLM deconvolution results; users will have to perform these advanced analyses themselves Validation but not deep coupling analysis

Reproducibility, data availability and technical notes

Open data is provided with event structures which supports reproducible analyses; however, best practice would be for authors to include example preprocessing scripts (EEGLAB HDF5 or BIDS format) and recommended fNIRS preprocessing pipelines (short separation, physiological regressors, motion correction) to maximize reuse and comparability with other multimodal datasets Open data and reuse intent

Context from other multimodal work: previous open multimodal EEG-fNIRS datasets exist for cognitive tasks (eg N-back Stroop visual working memory) and more recent work provides near whole head EEG-fNIRS and multimodal wearable patches; integrating across these resources improves external validity of algorithms trained on a single frontal dataset Visuomotor multimodal dataset context Co-located patch validation

Practical recommendations for using the dataset

1. Preprocess fNIRS with short separation regression or physiological regressors and apply motion correction and detrending before GLM or image reconstruction to reduce systemic confounds fNIRS preprocessing guidance
2. Align EEG and fNIRS in time carefully and consider deconvolution or model-based HRF estimation to allow trial-by-trial coupling analyses; cross-correlation of ERP peaks and HbO/HbR peaks may reveal neurovascular latency relationships but requires robust baseline correction ERP to Hb coupling example
3. For ML benchmarking use subjectwise cross validation and consider augmentation or transfer learning from larger EEG or fNIRS corpora to reduce overfitting given modest N=21 sample size Multimodal ML caveats and solutions

Quick visual summary (channels and sample rates)

Conclusions and final evaluation

The dataset represents a useful, clearly described resource for multimodal EEG-fNIRS methods development focused on frontal cognitive control tasks. It is not a large normative cohort nor a whole-head mapping resource, so its ideal uses are algorithm benchmarking, neurovascular coupling methods development, and educational replication studies Final dataset utility statement

Key citations used

• Open access EEG fNIRS Stroop dataset
• fNIRS review and caveats
• Co-located patch performance
• Multimodal ML examples and benefits
• Multimodal visuomotor dataset context