Keskeiset käsitteet
A neuroergonomics model is proposed to effectively evaluate the performance of nuclear power plant operators under heat stress by leveraging ECG time-frequency spectrums and fNIRS prefrontal cortex network.
Tiivistelmä
The study aimed to develop a neuroergonomics model to evaluate the performance of nuclear power plant (NPP) operators under heat stress. The model utilized electrocardiogram (ECG) time-frequency spectrums and functional near-infrared spectroscopy (fNIRS) prefrontal cortex (PFC) network to extract discriminative features.
Key highlights:
- Operators' performance was categorized into three levels based on their performance in major tasks, situation awareness, workload, and working memory under different heat stress conditions.
- ECG time-frequency spectrums were obtained using short-time Fourier transform (STFT) and fed into a 4-layer convolutional neural network (CNN) backbone.
- fNIRS PFC network was constructed with activation betas as node features and functional connectivity as edge features, and then embedded using a 2-layer graph attention network (GAT) backbone.
- The CNN-learned features from ECG spectrums, the GAT-learned features from fNIRS PFC network, and the handcrafted ECG and fNIRS features were fused to achieve the final classification.
- The proposed CNN-GAT fusion model outperformed other variants, achieving 81.82% accuracy and 0.90 AUC in the three-class classification task.
- The results demonstrate that the small-world nature of the brain network can be effectively captured by GAT, leading to better discriminative feature extraction compared to simple non-linear transformation.
- The multi-physiological fusion model shows high ecological validity in detecting abnormal states of operators under extreme heat stress, providing a potential neuroergonomics application for industry 5.0 scenarios.
Tilastot
The ECG time-frequency spectrums were obtained using short-time Fourier transform (STFT) on the filtered ECG signals.
The fNIRS PFC network was constructed with activation betas in channels as node features and functional connectivity between channels as edge features.
Lainaukset
"The model fused a convolutional neural network (CNN) backbone and a graph attention network (GAT) backbone to extract discriminative features from ECG time-frequency spectrums and fNIRS prefrontal cortex (PFC) network respectively with deeper neuroscience domain knowledge, and eventually achieved 0.90 AUC."
"Inspired by the small-world nature of the brain network, the fNIRS PFC network was organized as an undirected graph and embedded by GAT. It is proven to perform better in information aggregation and delivery compared to a simple non-linear transformation."