Prediction of Response Time and Vigilance Score in a Sustained Attention Task from Pre-trial Phase Synchrony using...

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Objective assessment of sustained attention is crucial for real-time performance monitoring.
• Existing methods may disrupt user behavior or require specialized hardware like eye trackers.
• Understanding the neural correlates of vigilance variations is key to developing non-invasive monitoring tools.

Purpose of the Study:

• To investigate if phasic functional connectivity patterns predict tonic performance in the Sustained Attention to Response Task (SART).
• To develop a system for real-time vigilance assessment using neural data.
• To explore the utility of deep neural networks for predicting vigilance metrics.

Main Methods:

• Utilized pre-trial phase synchrony indices (PSIs) from attention networks as features.
• Employed deep neural networks (DNNs) with Mean Squared Error (MSE) and Mean Absolute Error (MAE) loss functions.
• Conducted 4-fold cross-validations to assess prediction accuracy for cumulative vigilance score (CVS) and hit response time (HRT).

Main Results:

• DNNs with MSE loss outperformed those with MAE.
• Beta sub-band (16-20 Hz) PSIs accurately predicted average CVS (RMSE=0.043, R=0.806).
• Alpha oscillation PSIs effectively predicted mean HRT (RMSE=51.91 ms, R=0.903).

Conclusions:

• Phasic functional connectivity, specifically PSIs, can reliably predict sustained attention performance.
• The proposed DNN-based system offers an objective, non-invasive method for monitoring vigilance variations.
• This approach advances the use of neural network regression models for vigilance assessment, distinct from prior methods.