# Combining EEG and eye-tracking for cognitive and physiological states monitoring: a systematic review

## TL;DR

This paper reviews how combining EEG and eye-tracking can better detect mental states like fatigue and stress, improving safety in high-stakes environments.

## Contribution

The study systematically reviews concurrent EEG and eye-tracking research to identify shared and distinct physiological patterns in perception-related states.

## Key findings

- EEG theta, alpha, and beta activity and ET metrics like PERCLOS are commonly used to monitor cognitive states.
- Multimodal EEG and ET data improve classification accuracy of fatigue, stress, and drowsiness compared to single modality.
- Fatigue and drowsiness share physiological patterns but can be distinguished using combined EEG and ET features.

## Abstract

Monitoring situational awareness is critical in highly demanding environments where sustained attention and vigilance are essential for safety and performance. Electroencephalography (EEG) and eye-tracking (ET) provide complementary insights into the perceptual layer of situational awareness, capturing neural and ocular signatures of information processing, attention, and fatigue. However, studies have typically examined perception-related conditions such as workload, fatigue, stress, and drowsiness in isolation, limiting understanding of their shared and distinct physiological patterns. This systematic review synthesizes findings from studies that recorded EEG and ET concurrently to investigate perception-related conditions. Following the PRISMA 2020 statement, five databases were searched, and 47 studies met the inclusion criteria. The most frequently reported EEG features included theta, alpha, and beta activity, while ET metrics commonly involved fixation patterns, pupil diameter, blink dynamics, and percentage of eyes closed (PERCLOS). Across studies, fatigue, mental workload, and stress exhibited overlapping physiological signatures, although multimodal data helped differentiate these closely related states. Drowsiness and vigilance decrement appeared along a shared continuum, with microsleeps showing distinct physiological profiles. Classification models generally achieved higher accuracy when integrating EEG and ET features than when using either modality alone. This review highlights the potential of concurrent EEG and ET monitoring for improving the detection of perception-related conditions and for disambiguating closely related states. These findings also support the need for standardized multimodal protocols and real-time multimodal classification models to strengthen cognitive-state monitoring, operational performance, and error prevention in high-risk domains.