Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 19, 2026Scientific ReportsOpen Access

Multimodal deep learning combining EEG and ECG outperforms single modalities with ~96% stress detection accuracy.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Current stress assessment methods rely on subjective reports or isolated physiological parameters, limiting accuracy and consistency.

Does an attention-based multimodal deep learning framework integrating EEG and ECG improve accuracy in detecting psychological stress compared to single-modality methods?

Population

35 subjects in neutral, tense, and positive states from WESAD and CASE datasets

Comparison

Multimodal EEG and ECG deep learning model vs EEG-only, ECG-only, or individual networks

Design

Machine learning model development and validation study

Key result

The attention-based multimodal deep learning framework integrating EEG and ECG achieved 95.7% accuracy in identifying stress states, outperforming EEG-only (82.3%) and ECG-only (85.6%) methods.

Authors

RKRakesh KumarSingapore Institute of TechnologySKSivanesan Bala KrishnanSingapore Institute of TechnologyRYRakesh Kumar YadavUniversity Of Information Technology

Discussion

Loading...

Member takes

Overview

Supports multimodal monitoring for stress in cardiovascular patients; extends single-modality approaches but should not yet change practice.

Key Points

  • This study aims to develop an objective method for stress detection using multimodal physiological data from EEG and ECG.
  • Utilized deep transfer learning with three neural network architectures: VGG16, EfficientNetB0, ResNeXt50.
  • Converted EEG data into time-frequency maps and analyzed ECG data for time-series variability.
  • Implemented an advanced fusion layer with attention weights to merge data from EEG and ECG.
  • Achieved 95.7% accuracy in stress condition identification, outperforming EEG-only (82.3%) and ECG-only (85.6%) methods.
  • Demonstrated high flexibility across various testing conditions.
  • Showed mutual enhancement of stress assessment from EEG and ECG signals.

Structured PICO

Does an attention-based multimodal deep learning framework integrating EEG and ECG improve accuracy in detecting psychological stress compared to single-modality methods?

P
Population
35 subjects from the WESAD (n=15) and CASE (n=20) datasets
I
Intervention
Attention-based multimodal deep learning framework integrating EEG and ECG using VGG16, EfficientNetB0, and ResNeXt50
C
Comparator
EEG-only methods, ECG-only methods, and individual networks
O
Outcome
Accuracy in identifying between neutral (control), tense, and positive states

Main Result

Absolute Event Rate: 95.7% vs 85.6%

A multimodal deep learning framework integrating EEG and ECG data significantly improves the accuracy of psychological stress detection compared to single-modality approaches.

Cite This Study

Kumar et al. (2026) studied Psychological stress (n=35). Attention-based multimodal deep learning framework integrating EEG and ECG vs. EEG-only and ECG-only methods was evaluated on Stress detection accuracy across neutral, tense, and positive states. The attention-based multimodal deep learning framework integrating EEG and ECG achieved 95.7% accuracy in identifying stress states, outperforming EEG-only (82.3%) and ECG-only (85.6%) methods.

synapsesocial.com/papers/69bb929b496e729e629800dehttps://doi.org/10.1038/s41598-026-44499-0
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Deep ECGNet: An Optimal Deep Learning Framework for Monitoring Mental Stress Using Ultra Short-Term ECG Signals2018 · 133 citations
  2. 2Deep Learning Based Classification and Combined Transform Based Feature Extraction Approach for Mental Stress Prediction of Human Beings Using <scp>EEG</scp>2025 · 1 citations
  3. 3Deep ECG-Respiration Network (DeepER Net) for Recognizing Mental Stress2019 · 67 citations
  4. 4Hybrid Deep Learning Approach for Stress Detection Using Decomposed EEG Signals2023 · 63 citations
  5. 5Psychological Stress Classification Using EEG and ECG: A CNN Based Multimodal Fusion Model2024 · 2 citations