Eye-tracking technology has emerged as a powerful non-invasive digital biomarker for detecting cognitive impairment in Alzheimer's disease (AD). Recent research demonstrates that novel eye-tracking digital markers can outperform traditional plasma biomarkers in detecting cognitive impairment, offering a minimally invasive, accessible screening approach that can be deployed in clinical settings, research environments, and even at-home screening programs["@eyetracking2025"][@eyetracking2024].
Eye-tracking technology has emerged as a powerful non-invasive digital biomarker for detecting cognitive impairment in Alzheimer's disease (AD). Recent research demonstrates that novel eye-tracking digital markers can outperform traditional plasma biomarkers in detecting cognitive impairment, offering a minimally invasive, accessible screening approach that can be deployed in clinical settings, research environments, and even at-home screening programs["@eyetracking2025"][@eyetracking2024].
The development of eye-tracking as a biomarker represents a significant advancement in the field of digital health for neurodegeneration. Unlike traditional diagnostic approaches that require invasive procedures such as lumbar puncture for cerebrospinal fluid biomarkers or expensive imaging such as PET scans, eye-tracking offers a practical alternative that maintains high sensitivity and specificity while dramatically reducing barriers to implementation["@digital2024"].
Eye movements are controlled by a distributed neural network that involves multiple brain regions, making them sensitive to changes in neural circuitry that occur in Alzheimer's disease[@eyetracking2024a]. The key neural structures involved in oculomotor control include:
Frontal Eye Fields (FEF): Located in the prefrontal cortex, the FEF is responsible for initiating voluntary saccades and is particularly vulnerable to early AD-related changes. Neurofibrillary tangle formation in the FEF correlates with impaired saccadic performance, making it an early indicator of pathological changes.
Superior Colliculus: This midbrain structure serves as a sensorimotor interface for eye movements and receives input from multiple cortical areas. The superior colliculus is involved in both reflexive and voluntary eye movements, and its function can be affected by neurotransmitter changes in AD.
Pulvinar Nucleus: Part of the thalamus, the pulvinar plays a critical role in attention modulation and visual processing. Research has shown that pulvinar atrophy occurs in AD, contributing to attention deficits and abnormal eye movement patterns.
Basal Ganglia: The basal ganglia, particularly the caudate nucleus, is involved in the control of voluntary eye movements and task-switching. Dopaminergic degeneration in AD affects basal ganglia function, leading to characteristic oculomotor abnormalities.
Cerebellum: Though traditionally associated with motor coordination, the cerebellum also participates in eye movement control, particularly in smooth pursuit and vestibulo-ocular reflex adaptation. Cerebellar involvement in AD may contribute to pursuit abnormalities.
Multiple studies have documented specific oculomotor deficits in AD that can serve as biomarkers:
Saccadic Dysfunction: Patients with AD and mild cognitive impairment (MCI) demonstrate increased saccadic latency, reduced peak velocity, and impaired accuracy compared to healthy controls[@boccuni2018]. These abnormalities are particularly pronounced in pro-saccade (reflexive) and anti-saccade (volitional) tasks that require executive control.
Fixation Instability: AD patients show reduced fixation stability, with shorter fixation durations and increased saccadic intrusions during visual exploration tasks[@hutton2023]. This instability correlates with attention deficits and visual processing abnormalities characteristic of the disease.
Smooth Pursuit Impairments: Smooth pursuit eye movements, which are used to track moving objects, are significantly impaired in AD[@lee2023]. These deficits are thought to reflect both cortical and subcortical involvement, particularly in the frontal and parietal regions.
Pupillary Abnormalities: Pupil response to cognitive tasks is altered in AD, with reduced pupillary dilation during challenging cognitive paradigms[@goldberg2022]. This may reflect cholinergic dysfunction and can serve as an indicator of disease severity.
Scanpath Alterations: Visual exploration patterns are markedly different in AD compared to healthy controls, with reduced complexity and altered strategies for scene viewing[@nelson2024]. These patterns may reflect hippocampal and prefrontal dysfunction.
Modern eye-tracking systems use various technologies to measure eye position and movement:
Video-Oculography (VOG): The most common approach, using infrared cameras to track pupil position and corneal reflections. Modern systems achieve sub-degree accuracy and can operate in both laboratory and portable settings.
Electro-Oculography (EOG): Uses electrodes placed around the eyes to measure electrical potentials generated by eye movements. EOG offers good temporal resolution but lower spatial accuracy than VOG.
Search Coil Systems: Use contact lenses with embedded coils to measure eye position in magnetic fields. These provide very high accuracy but are invasive and rarely used clinically.
Portable Devices: Recent advances have led to lightweight, portable eye-tracking systems that can be used outside laboratory settings, enabling at-home monitoring and large-scale screening applications[@portable2025][@kim2024].
The following metrics have demonstrated utility as AD biomarkers[@digital2024]:
| Metric | Description | AD-Related Change |
|--------|-------------|-------------------|
| Saccadic Latency | Time to initiate eye movement | Increased |
| Peak Velocity | Maximum speed of saccades | Decreased |
| Fixation Duration | Time spent on specific points | Decreased |
| Saccadic Rate | Number of saccades per time | Increased |
| Pupil Dilation | Change in pupil size | Reduced |
| Smooth Pursuit Gain | Accuracy of tracking | Decreased |
| Anti-saccade Errors | Failure to suppress reflexive saccades | Increased |
Classical Metrics: Traditional analysis focuses on basic parameters such as latency, velocity, accuracy, and fixation statistics. These metrics have been validated in numerous studies and show consistent differences between AD patients and controls.
Machine Learning: Advanced approaches use machine learning algorithms to identify complex patterns in eye movement data that may be more sensitive to early changes than individual metrics[@chen2020][@ullah2023]. These models can achieve high classification accuracy for AD detection.
Deep Learning: Recent work has explored using deep neural networks to learn features directly from raw eye-tracking data, achieving state-of-the-art performance in some studies[@smith2024]. These approaches may capture subtle patterns not apparent to classical analysis.
Eye-tracking shows promise for early AD detection, potentially identifying individuals in the preclinical or prodromal stages before significant cognitive decline[@johnson2023]. The non-invasive nature and relatively low cost make it suitable for population-level screening.
Preclinical Detection: Studies have demonstrated that eye-tracking abnormalities can be detected in cognitively normal individuals with biomarker evidence of AD pathology (amyloid positivity, elevated p-tau), suggesting that oculomotor changes precede clinical symptoms.
MCI Differentiation: Eye-tracking can help differentiate between MCI due to AD and MCI due to other causes, with specific patterns associated with AD-type pathology.
At-Risk Populations: Eye-tracking has been studied in populations at increased risk for AD, including those with family history, genetic risk factors (APOE4 carriers), and Down syndrome[@zhang2023].
Eye-tracking patterns may help distinguish AD from other neurodegenerative conditions[@oyama2021]:
AD vs. Frontotemporal Dementia: Different oculomotor profiles may reflect the distinct regional patterns of neurodegeneration in these conditions.
AD vs. Dementia with Lewy Bodies: Specific differences in saccadic performance and pupillary response may help differentiate these conditions.
AD vs. Vascular Cognitive Impairment: Eye-tracking can help identify the additive contribution of AD pathology in patients with vascular cognitive changes.
Eye-tracking offers potential for monitoring disease progression and treatment response[@pozzilli2024]:
Progression Tracking: Longitudinal studies show that oculomotor abnormalities worsen over time, potentially correlating with clinical progression.
Treatment Response: Eye-tracking may serve as a biomarker in clinical trials, measuring effects of disease-modifying therapies on underlying neurobiological changes.
At-Home Monitoring: Portable devices enable continuous monitoring in home settings, providing rich data on functional status beyond clinical visits[@mack2022].
Eye-tracking metrics correlate with disease severity as measured by standard neuropsychological tests, potentially providing objective measures of functional impairment beyond subjective clinical ratings[@fernandez2021].
Eye-tracking-based detection achieves promising diagnostic performance:
AD Detection: Sensitivity typically ranges from 80-90%, with specificity from 75-85% in research settings. Performance varies based on the specific tasks and metrics used.
MCI Detection: Performance is lower for MCI than for dementia, reflecting the subtler oculomotor abnormalities in early disease stages.
Preclinical Detection: Detection of preclinical AD remains challenging, though some studies report promising results.
Eye-tracking compares favorably with established biomarkers in some contexts[@eyetracking2025]:
vs. Plasma Biomarkers: Recent research suggests that eye-tracking digital markers can outperform plasma biomarkers (p-tau181, p-tau217) for detecting cognitive impairment in some populations.
vs. CSF Biomarkers: Eye-tracking shows reasonable agreement with CSF biomarkers but may be less sensitive to early pathological changes.
vs. Imaging: Eye-tracking does not provide information about regional brain atrophy or amyloid/tau burden, but may complement imaging biomarkers.
Combining eye-tracking with other biomarkers may improve diagnostic accuracy:
Eye-Tracking + Plasma: Studies have shown that combining eye-tracking with plasma biomarkers improves classification beyond either modality alone[@ramirez2023].
Eye-Tracking + Genetic: Integration with genetic risk scores (APOE status) may improve risk stratification.
The specific eye-tracking tasks used influence diagnostic performance:
Pro-saccade Tasks: Simple reflexive saccades to visual targets. Sensitive to basic oculomotor function but less specific to cognitive impairment.
Anti-saccade Tasks: Require suppressing reflexive saccades and generating voluntary saccades in the opposite direction. Strongly dependent on executive function and particularly sensitive to prefrontal involvement.
Gap/Overlap Tasks: Manipulate the timing between fixation offset and target appearance to assess attention and latency. Useful for detecting subtle deficits.
Smooth Pursuit Tasks: Track moving targets. Reveals deficits in sensorimotor integration and visual processing.
Scene Viewing: Free exploration of complex images. Captures naturalistic attention patterns and visual processing strategies.
Proper standardization is critical for reliable results:
Lighting: Ambient lighting affects pupil detection and may influence results.
Head Movement: Motion artifacts can corrupt data; head stabilization improves quality.
Calibration: Regular calibration ensures accurate eye position measurement.
Task Instructions: Clear, consistent instructions are essential for valid results.
Quality control is essential:
Artifact Rejection: Blinks, head movements, and technical artifacts must be identified and excluded.
Minimum Data Requirements: Sufficient trials and samples are needed for reliable metrics.
Attention and Cooperation: Patient engagement affects performance; short, engaging tasks may improve data quality.
Disease Specificity: Eye-tracking abnormalities are not specific to AD and may occur in other conditions, requiring integration with other clinical information.
Early Detection: Sensitivity for preclinical AD remains limited; further optimization is needed.
Standardization: Lack of standardized protocols across studies limits comparability and clinical translation.
Technical Expertise: Requires specialized equipment and expertise that may not be available in all clinical settings.
Cost: While less expensive than PET imaging, eye-tracking systems still require investment.
Accessibility: Not all healthcare settings have access to eye-tracking technology.
Training: Proper administration and interpretation require training.
Patient Factors: Some patients (e.g., those with visual impairments, severe tremor) may be difficult to assess.
Improved Hardware: Smaller, more affordable, and easier-to-use eye-tracking devices are being developed[@kim2024].
Remote Monitoring: Web-based and smartphone-compatible eye-tracking may enable large-scale remote assessment[@yang2024].
AI Integration: Machine learning approaches continue to improve classification accuracy.
Standardization: Development of standardized protocols and normative data is needed.
Validation: Large-scale validation studies are required before clinical implementation.
Regulatory Approval: FDA clearance will be needed for clinical diagnostic use.
Clinical Trials: Eye-tracking may serve as a biomarker for patient selection and outcome measurement in trials.
Biomarker Development: Integration with fluid and imaging biomarkers for comprehensive characterization.
Precision Medicine: Personalized risk prediction and treatment selection based on multimodal data.
Eye-tracking represents one of several digital biomarker modalities under development for AD:
vs. Speech and Language: Speech analysis can detect linguistic changes; eye-tracking provides complementary information about visual attention and oculomotor control.
vs. Actigraphy: Movement-based biomarkers capture daily activity patterns; eye-tracking focuses on specific cognitive processes.
vs. Smartphone Sensors: Touch and typing patterns provide cognitive assessment; eye-tracking requires specialized hardware.
Eye-tracking is related to other biomarkers and topics in the NeuroWiki: