Eye-brain coupling-mediated eye movement abnormalities as non-invasive biomarkers for mild cognitive impairment: A systematic review

Abstract

Background

Eye movement abnormalities have emerged as promising non-invasive candidate biomarkers for the early detection, progression monitoring, and differential diagnosis of Alzheimer's disease (AD), with preliminary clinical evidence supporting their translational potential. Current AD diagnostic methods are limited by subjectivity, high cost, and complexity—making non-invasive biomarkers critical, especially for mild cognitive impairment (MCI). The tight functional link between the eye and brain underscores eye movement abnormalities as a window into AD-related pathology.

Objective

This systematic review summarizes AD-associated multi-modal eye movement dysfunctions (focusing on saccades, fixation, and smooth pursuit), clarifies their pathological mechanisms, clinical value, and translational feasibility, providing a basis for constructing an AD biomarker system.

Methods

Literature searches were conducted in PubMed, Web of Science, and Google Scholar using (“Alzheimer's disease” OR “mild cognitive impairment” AND “eye movements” OR “saccades” OR “smooth pursuit” OR “fixation”) with a search cutoff date of September 30, 2025. Studies deemed irrelevant or lacking sufficient data were excluded.

Results

AD/MCI patients exhibit eye movement abnormalities: prolonged saccadic latency, increased antisaccade errors, reduced fixation stability, and attenuated smooth pursuit gain, which are closely linked to core AD pathologies, detectable in AD/MCI stages, and can, to a certain extent, distinguish AD from Parkinson's disease and frontotemporal dementia, and have value for MCI-to-AD conversion prediction.

Conclusions

Eye movement abnormalities hold promise as non-invasive biomarkers for AD, with potential for preclinical screening and differential diagnosis. To advance translation, future research should prioritize AI-driven multi-modal integration, standardized detection protocols, portable device development, and preclinical longitudinal validation.

Keywords

Alzheimer's disease eye movement fixation mild cognitive impairment saccades smooth pursuit

Introduction

Alzheimer's disease (AD) is a neurodegenerative disease characterized by progressive memory loss and cognitive decline. Its typical neuropathological features include senile plaques formed by amyloid-β (Aβ) deposition, neurofibrillary tangles (NFTs) caused by hyperphosphorylation of tau protein, and significant loss of cholinergic neurons in the basal forebrain.¹ At present, the clinical diagnosis of AD still mainly relies on neuropsychological scale assessment combined with cerebrospinal fluid (CSF) detection or molecular imaging techniques, such as Aβ positron emission tomography (PET), tau-PET, magnetic resonance imaging (MRI) and other imaging techniques. However, these approaches exhibit distinct limitations when applied to early screening and popularization efforts, primarily due to strong subjectivity, insufficient specificity, and the practical challenges of high cost and complex operation.^2,3 Therefore, the identification of objective, non-invasive biomarkers with high feasibility for widespread implementation has emerged as a critical research focus in the field of AD.

The eye and brain are highly analogous with respect to developmental origin, functional characteristics, cellular composition, neurotransmitter receptors, blood supply sources, and angiogenesis.^4,5 These similarities suggest that aberrant eye movements could have an inherent link to abnormalities in brain network function, underscoring the eye's potential as a window into central nervous system pathology.^6,7 It is reported that the regulation of eyeball movement depends on a fine neural network composed of frontal eye movement area (FEF),⁸ parietal eye fields (PEF),⁹ superior colliculus (SC),¹⁰ cerebellum and brainstem nuclei.¹¹ It is worth noting that this network is the key area involved in the early stage of AD pathology. Studies have shown that AD-related Aβ and tau pathological deposition, cholinergic system degeneration, and neural network connection breakage can directly affect the efficiency and accuracy of eye movement control.^12–15 Therefore, AD patients may have quantifiable eye movement abnormalities before clinical symptoms appear, and eye movement tracking technology is expected to become a powerful tool for early identification and course monitoring of AD.

Compared with traditional technologies such as neuroimaging, eye movement detection has the advantages of non-invasive, short time-consuming and low cost, and is more suitable for large-scale population screening and longitudinal monitoring. Existing studies have initially revealed that AD patients have characteristic changes in many types of eye movement tasks, including saccade¹⁶ (such as prolonged latency, increased error rate), fixation stability¹⁷ (such as increased drift), smooth pursuit¹⁸ (such as decreased gain), and exploratory strategies.¹⁹ However, most of the existing reviews focus on a single type of eye movement abnormalities, and there is a lack of integrated analysis of multi-modal eye movement indicators and a systematic exposition of their association with the pathological progression of AD.

This review aims to systematically review the research progress of AD-related eye movement dysfunction, focusing on the abnormal characteristics of various eye movement modes such as saccade, fixation, and smooth pursuit and their evolution with the progression of the disease, and to explore the neurobiological mechanisms such as Aβ/tau pathology, cholinergic defects, and neural network disruption. In addition, this paper will also compare the differences in eye movement patterns between AD and other neurodegenerative diseases (such as Parkinson's disease (PD) and frontotemporal dementia (FTD)) to evaluate the value of eye movement parameters in differential diagnosis, and provide a theoretical basis and application framework for the construction of AD biomarker system based on eye movement tracking. Notably, this review emphasizes MCI as a critical transitional stage, highlighting the unique value of eye movement parameters in predicting AD conversion and guiding early intervention.

To systematically retrieve literature on eye movements linked to AD, core search terms were defined as (“Alzheimer's disease” OR “Mild Cognitive Impairment”) AND (“eye movements” OR “saccades” OR “smooth pursuit” OR “fixation”). Searches were done in PubMed, Web of Science, and Google Scholar with a retrieval cutoff date of September 30, 2025. Duplicates were removed via EndNote; initial screening (titles/abstracts) excluded irrelevant studies. Full-text screening excluded inaccessible texts, insufficient data, or mismatched designs, yielding eligible studies for later use (Figure 1).

Figure 1.

Schematic diagram of literature retrieval and screening process.

Eye movement dysfunctions in AD patients

AD-related eye movement impairments include saccades, smooth pursuit, and fixation.²⁰ These abnormalities are closely associated with the stages of AD. The NIA-AA 2011 Clinical Diagnostic Criteria, combined with the DSM-5 staging system, were adopted to categorize AD into three stages: preclinical AD, MCI due to AD, and AD dementia, with the latter further subdivided into mild, moderate, and severe sub-stages in some studies.^21–23

Characteristics and pathological mechanisms of saccadic dysfunction in ad patients

Saccadic movement is the fastest type of eye movement in the human body, whose primary function is to rapidly shift the foveal fixation to the target region.²⁴ Its role is to enable the macular region—the retina's area of highest visual and color sensitivity—to rapidly capture targets, while regulating the neural network consisting of the FEF, PEF, SC, and brainstem eye movement nuclei.²⁵ This eye movement is closely linked to the brain's attentional mechanisms and serves as one of the most sensitive indicators for screening cognitive impairment.^26–28 Saccadic dysfunction in patients with AD is characterized by multiple types, early onset, and gradual progression.²⁶ It encompasses prosaccades, antisaccades, and microsaccades, with all these abnormalities exhibiting a distinct exacerbation pattern as the disease advances.

Prosaccade dysfunction in AD patients

Prosaccade denotes a type of saccadic eye movement wherein the eyes rapidly shift from a fixation point to a suddenly appearing peripheral target, aligning their direction with the target's location.²⁹ The prosaccade task typically requires participants to first maintain fixation on a central point. Upon presentation of a peripheral target stimulus, they must execute an immediate saccade toward the target, with no deliberate inhibition or redirection.²³ This eye movement is regarded as a fundamental, automatic eye movement response, commonly utilized in clinical and neuroscientific research to evaluate the integrity of visual attention, sensorimotor integration, and eye movement control.³⁰ It has been reported that prosaccadic abnormalities progressively worsen with the clinical progression of AD, with distinct disease-discriminating features observed at each stage.^27,31,32 In the preclinical stage of AD, patients typically exhibit no objective cognitive impairment,²⁸ with their prosaccadic function remaining largely intact.²⁶ There are reports indicating that upon progressing to the MCI stage of AD, saccadic abnormalities may emerge, including prolonged saccadic latency and an increased incidence of saccadic undershoot.³³ Clinical observations and relevant research have shown that as AD progresses to the dementia stage, the previously identified saccadic abnormalities are further exacerbated.²⁸ Specifically, this stage is characterized by a significant prolongation of saccadic latency, a marked reduction in peak velocity, and an increased incidence of both saccadic undershoot and saccadic overshoot^2,3,32,34 (Table 1), with the indicator mainly used for assessing disease severity rather than early prediction.

Table 1.

Prosaccade abnormalities in AD.

Group		Prosaccade latency	Prosaccade accuracy	vs. healthy controls	Key pathological correlations	Reference
Preclinical		→	→	-	Mild FEF tau pathology	^31,38
MCI		→ (p > 0.05)	→ (p > 0.05, but error rate maybe ↑, p < 0.05)		Prefrontal cortex (mild), parietal lobe (mild)	^30–32 ^,40
Dementia	Mild	↑↑ (p < 0.01)	→ (p > 0.05, but CV↑, p < 0.05)	360.7 ms vs. 313.5 ms (latency)100% vs. 100% (accuracy)	Parietal attentional network (early involvement)	^31,32,37
Dementia	Moderate-severe	↑↑↑ (p < 0.001)	↓↓ (p < 0.01)	411.9 ms vs. 313.5 ms (latency)median 80–100% vs. 100% (accuracy)	Parietal network (worsened), prefrontal network begins to involve	^31,32,37

→: Normal/No significant change.

↑/↓: Mild increase/decrease.

↑↑/↓↓: Significant increase/decrease.

↑↑↑/↓↓↓: Severe increase/decrease.

AD: Alzheimer's disease; MCI: mild cognitive impairment; CV: coefficient of variation; FEF: frontal eye movement area.

Compared with other neurodegenerative diseases, the prosaccade abnormalities of AD are mainly characterized by progressive and significant prolongation of latency, reflecting impaired information processing speed.²⁷ PD patients usually have normal latency but significant undershoot due to dopaminergic damage in the basal ganglia; FTD patients may show normal latency but more severe undershoot. These differences provide a reference for clinical identification.^35,36

The pathological basis of prosaccade abnormalities is related to the structural damage of PEF and SC. In the course of AD, pathological changes in these brain regions, including tau protein deposition,³⁷ Aβ deposition,³⁸ cholinergic system degeneration³⁹ and decreased synaptic density，are closely involved in prosaccade dysfunction. Specifically, tau deposition in PEF impairs spatial information processing and sensorimotor integration, directly leading to increased saccadic undershoot and variability in amplitude.³⁸ Additionally, neuroinflammation may further participate in and accelerate this pathological process.

Prosaccade detection can be used as a potential auxiliary biomarker for AD diagnosis, but its application has certain limitations. The sensitivity of preclinical stage is insufficient, and the specificity of MCI stage is limited, which overlaps with PD, FTD and other diseases. In the stage of mild to moderate dementia, the combined application of multiple indicators may help to assess the severity of the disease, but it still needs to be combined with multi-modal indicators such as CSF detection and imaging for comprehensive judgment.

Antisaccade dysfunction in AD patients

The antisaccade task requires participants to suppress the instinctive impulse to look at a sudden visual stimulus (prosaccade) and instead perform an eye movement pointing in the opposite direction. Successful completion of this task requires complete inhibition control, working memory, and attentional flexibility.⁴¹ Studies³¹ have shown that AD patients show increased error rate (i.e., difficulty in inhibiting the impulse of reflexive stimuli) and decreased correction ability (i.e., difficulty in autonomous correction after finding errors) in this task, and these defects are aggravated with disease progression. In the preclinical AD stage, patients may show a slight increase in error rate in complex tasks, while simple task performance is often normal, and this mild abnormality has no statistical significance for predicting AD risk. As the disease progressed to the MCI stage, the error rate further showed a moderate increase, the correction rate decreased, and the correction latency prolonged.^28,40 This stage's antisaccade error rate is a core predictive marker for AD conversion: when combined with prosaccade latency, the predictive efficacy is superior, which is far more valuable than the preclinical stage's trivial changes and the dementia stage's severe error ratethat only reflects the loss of inhibitory control function. In the AD dementia stage, the abnormality is more significant, the error rate continues to increase, the correction ability decreases significantly, and finally the patient may completely lose the reverse saccade ability⁴² (Table 2).

Table 2.

Antisaccade abnormalities in AD.

Stage		Antisaccade accuracy	Error rate	Antisaccade latency	Mean difference vs. HC	Key pathological correlations	Reference
Preclinical		→ (simple tasks) / ↓ (complex)	→	→	-	Mild FEF tau pathology	^37,47
MCI		↓↓ (p < 0.01)	↓↓↓ (p = 0.001)	↑↑ (p < 0.01, d = 0.77)	25% vs. 53.33% (accuracy)419 ms vs. 338 ms (latency)	Prefrontal cortex (DLPFC), ACC	^{30,31,40,42,48}
Dementia	Mild	↓↓↓ (p < 0.001, d = 1.14)	↓↓↓ (p = 0.001)	↑↑↑ (p < 0.001, d = 1.52)	26.39% vs. 56.09% (accuracy)72.17% vs. HC: 97.01% (error correction rate)473.55 ms vs. HC: 372.22 ms (latency)	Prefrontal network (DLPFC, ACC) expanding to parietal connections	^31,37,49
Dementia	Moderate-severe	↓↓↓ (Extremely low, p < 0.001)	↓↓↓ (p < 0.001)	↑↑↑ (p < 0.001)	median 0% vs. 23.6% (accuracy) median 0% vs. HC: 61.3% (error correction rate)	Widespread prefrontal-parietal networks, anterior temporal regions	^16,31,48

AD: Alzheimer's disease; MCI: mild cognitive impairment; HC: healthy controls; FEF: frontal eye movement area; DLPFC: dorsolateral prefrontal cortex; ACC: anterior cingulate cortex.

Compared with other neurodegenerative diseases, the abnormal pattern of AD in the antisaccade task may be of differential significance. Studies have shown that the increase of error rate in PD patients is usually mild, and the correction ability is relatively retained. FTD patients may show more serious error rate and correction ability defects, reflecting different pathological mechanisms and neural basis.⁴³

The dysfunction of antisaccade is related to the neuropathological changes of AD, mainly involving the damage of FEF and cholinergic system.^27,44 The pathological deposition of tau in the FEF region and the loss of cholinergic neurons in the basal forebrain may be related to the decrease of inhibitory control function,⁴⁵ but the specific pathophysiological mechanism still needs further study. Neuroimaging studies have linked prolonged corrected antisaccade latency to decreased metabolism in the right dorsolateral prefrontal cortex (DLPFC), a region critical for working memory and task monitoring during antisaccade execution.⁴⁶

The antisaccade task stands as a promising biomarker for AD and a sensitive behavioral measure to identify progressive deficits in inhibitory control and error monitoring in affected individuals. Key behavioral signatures include markedly elevated error rates, particularly uncorrected errors—compromised error correction capacity, and prolonged latency of correct responses, with performance worsening sequentially from the MCI stage to AD dementia. Critically, antisaccade impairment emerges in the prodromal phase, conferring predictive value for MCI-to-AD conversion. For differential diagnosis, a defining feature distinguishes AD from bvFTD: AD patients exhibit near-complete loss of error correction, whereas bvFTD patients retain relatively intact correction abilities. Its clinical utility is nonetheless limited in moderate-to-severe dementia due to inadequate patient cooperation, necessitating future studies to validate its real-world applicability and explore combinatorial use with other biomarkers.

Fixation dysfunction in AD patients

The fixation movement disorder of AD patients can occur in the early stage of the disease and deteriorate systematically with the progression of the disease. Fixation is not the complete stillness of the eyeball, but an active control process involving physiological drift, tremor and microsaccades.^16,50,51 Its core function is to maintain the stability of visual images in the macular fovea of the retina to support higher-order cognitive processing. This process depends on coordinated regulation of neural networks including the parietal cortex, SC, brainstem integrators, and prefrontal cortex, with the prefrontal cortex critical for suppressing irrelevant saccades via cognitive control.^31,35,51,52

In the preclinical stage, although AD patients have no significant clinical symptoms, there have been subtle changes in fixation stability, manifested as a slight increase in drift velocity and an increase in the frequency of small square wave beats. As the disease progresses to the MCI stage, the fixation instability is further aggravated, the variability of fixation position is significantly increased, and the fixation maintenance ability begins to decline.¹⁶ Patients show increased saccade intrusion rate and shortened fixation duration in tasks that require ignoring distractors. In the stage of AD dementia, the fixation function is seriously impaired, and patients often cannot maintain stable fixation, showing a purposeless “wandering fixation” mode, and the visual positioning function is basically lost¹⁶Table 3.

Table 3.

Fixation abnormalities and neural mechanisms in AD.

AD stage	Fixation accuracy	Fixation duration	Microsaccades direction (deviation from horizontal)	Mean difference vs. HC	Key pathological correlations	Reference
Preclinical	→	→	→	-	Mild FEF tau pathology	^16,47
MCI	↓ (p < 0.05)	↓ (p < 0.05)	↑ (p < 0.05)	(Median) 91.10% vs. 92.67%(Median) 2665.30 ms vs. 2193.31 ms(SD) 36.2 ± 11.7° vs. 27.6 ± 9.2°	Damage to the cerebral hippocampus, possibly reflecting declining function of the fronto-parietal network	^16,55
Dementia	↓↓↓ (p < 0.001)	↓↓↓ (p < 0.001)	↑ (p < 0.05)	(Median) 67.02% vs. 92.67%(Median) 9911.57 ms vs. 2193.31 ms(SD) 37.1 ± 10.7° vs. 27.6 ± 9.2°	Medial temporal and attention-visuospatial network dysfunction, involving occipital eye fields, MT, DLPFC, FEF, and brainstem.	^16,55

AD: Alzheimer's disease; MCI: mild cognitive impairment; HC: healthy controls; FEF: frontal eye movement area; DLPFC: dorsolateral prefrontal cortex; MT: middle temporal area.

The neural correlates of this impairment are linked to distinct pathological processes. The degradation of basic fixation stability is primarily associated with posterior cortical dysfunction, particularly in parietal regions essential for spatial processing^31,53—exacerbated by tau and Aβ deposition that weakens position signal coding. Concurrently, the failure to actively suppress distracting glances implicates a breakdown of inhibitory control mechanisms governed by the prefrontal cortex.^29,46 Neuroimaging evidence corroborates altered activation within these frontoparietal networks during fixation tasks in AD.¹⁶

Although fixation parameters provide promising objective biomarkers for the early identification of AD, their clinical translation remains hampered by inherent challenges. Notably, fixation indices are susceptible to confounding factors such as aging-related physiological changes and comorbid eye disorders, which may compromise diagnostic specificity. To address this limitation, multi-modal integration strategies, including combining gaze metrics with CSF biomarkers or structural MRI data, are warranted to enhance diagnostic accuracy. Future research should focus on standardizing detection protocols to ensure cross-center reproducibility, establishing large-scale longitudinal databases to clarify the trajectory of fixation abnormalities throughout AD progression, and exploring the utility of fixation parameters for efficacy assessment in clinical trials of anti-AD therapeutics. These efforts will facilitate the successful translation of this non-invasive detection approach into routine clinical practice.

Microsaccades, small reflexive eye movements during fixation, may also demonstrate altered properties in AD. Studies have shown that the microsaccade parameters of AD patients may change with disease progression, including abnormalities in frequency, amplitude stability, and direction control^16,18 (Table 3).

A comparative perspective aids differential diagnosis. The eye movement profile of AD, characterized by progressive fixation instability, contrasts with patterns seen in other disorders. PD, for instance, often features saccadic hypometria alongside movement-related fixation issues.^29,43,54 Notably, patients with semantic dementia, a frontotemporal lobar degeneration subtype, frequently retain normal eye movement function, providing a key distinguishing feature from AD.^46,53

The pathophysiology underlying microsaccadic changes in AD is not fully elucidated but likely involves the degradation of cortical networks responsible for predictive visual control. Pathology in visual and posterior parietal association cortices may disrupt the fine-tuning of these movements.^31,53

As a potential biomarker, microsaccades are still in the research phase. In early AD, their sensitivity is insufficient, and they are susceptible to confounders (such as eye diseases, mental disorders), which limits ability as an independent diagnostic marker. Future research should focus on standardized detection protocols and explore combinatorial applications with CSF or imaging biomarkers to improve effectiveness in early AD diagnosis and disease monitoring.

Smooth pursuit dysfunction in AD patients

Smooth pursuit eye movements (SPEM), which allow the eyes to closely follow a moving target, are frequently impaired in patients with AD.^31,53 These abnormalities emerge early in disease progression, correlate closely with cognitive decline, and have drawn attention for their potential role in AD's early assessment, though their diagnostic specificity is not uniformly high across studies.^31,46,53

The core deficits in AD involve a reduction in pursuit gain and an increase in tracking error.⁵³ Empirical data show that AD patients exhibit significantly greater horizontal pursuit error compared to healthy controls (e.g., an increase of 2.92°).⁴⁶ Prolonged vertical pursuit latency has also been documented.⁵³ These abnormalities worsen as the disease advances. In the preclinical stage, no definitive, specific SPEM changes have been validated-subtle, non-diagnostic alterations may occur but lack predictive value for AD conversion.^53,55 By the MCI stage, subtle deficits emerge: gain may decrease modestly, and tracking error begins to rise, though these changes are often less pronounced than saccadic or fixation abnormalities.^31,46 The AD dementia stage brings severe SPEM dysfunction: tracking error increases substantially, gain further declines, and initial acceleration is severely compromised.^46,53 This progressive pattern reflects the spread of neurodegenerative changes along visuomotor pathways, particularly involving posterior cortical and cerebellar-brainstem networks^31,53 (Table 4).

Table 4.

SPEM changes in AD progression.

AD stage	Start latency	Gain	Tracking error	vs. HC	Pathological association	Reference
Preclinical	→	→	→ (p > 0.05)	-	Early pathology in V5 area	³¹
MCI	→ (p > 0.05)	→ (p > 0.05)	→ (p > 0.05)	-	Likely minimal involvement of pursuit networks	^31,35
Dementia	↑ (p < 0.05)	↓↓ (p < 0.01)	↑↑↑ (p < 0.001)	+2.92° in Horizontal SP+1.43° in Vertical SP	Posterior cortical networks (MT/V5, parietal), FEF, Cerebellum	^35,46,53

AD: Alzheimer's disease; MCI: mild cognitive impairment; SPEM: smooth pursuit eye movements; HC: healthy controls; FEF: frontal eye movement area.

Notably, SPEM impairment patterns differ across neurodegenerative diseases, conferring limited differential diagnostic value.^46,53,54 PD patients exhibit reduced SPEM gain and increased saccadic intrusions, distinct from AD's prominent tracking error elevation.⁵⁴ bvFTD is characterized by decreased horizontal tracking gain (0.19 lower than controls, p = 0.031) but minimal tracking error increase, whereas AD shows marked error elevation across horizontal and vertical axes.^46,53 These differences may assist in clinical differentiation but are not definitive in isolation⁴⁶ (Table 5).

Table 5.

Comparison of eye movement features in AD, PD, and FTD.

Feature		AD	PD	FTD	Reference
Core pathology	Key protein deposition	Aβ deposition, P-tau NFTs	α-synuclein aggregates (Lewy bodies/neurites)	Mainly TDP-43 protein; Partially tau	^{29,31,39,43,53}
Core pathology	Critical brain region damage	Onset in the medial temporal lobe (hippocampus) → Parietal attention network → Prefrontal executive network	Basal ganglia output dysfunction (affecting SC) and impairment of the prefrontal-striatal cognitive circuit	Primary severe atrophy of the prefrontal cortex (especially DLPFC)	^{26,29,31,43,46,48,53}
Prosaccade latency		↑↑↑ (p < 0.001, CV ↑↑↑, p < 0.001)	↑↑↑ (p < 0.001)	→ (p > 0.05)	^31,37,43,46
Prosaccade accuracy		↓↓ (p < 0.01)	↓↓↓ (p < 0.001)	-	⁴³
Antisaccade accuracy		↓↓↓ (Extremely low, p < 0.001)	↓↓↓ (p < 0.001)	↓ (p < 0.05)	⁴⁶
Antisaccade error rate		↑↑↑ (p < 0.001)	↑↑ (p < 0.01)	-	^29,46
Antisaccade error correction rate		↓↓↓ (p < 0.001)	↑↑ (p < 0.01)	-	⁴⁶
Antisaccade latency		↑↑↑ (p < 0.001)	prolonged (inconsistent results)	-	^29,46
Fixation		deviation duration↑↑↑ (p < 0.001)time and fixations↑	fixation time ↑ (p < 0.05), exploration range ↓ (p < 0.05)	→ Overall fixation stability better than AD	^16,27,46,54
Microsaccades		abnormal direction ↑ (p < 0.05)No differences in dynamic parameters	amplitude ↑↑↑ (p < 0.001) peak velocity ↑↑ (p < 0.005)	→ (Normal, p > 0.05)	^{46,51,53,56,57}
SPEM		tracking error ↑↑↑ (p < 0.001)gain ↓↓ (p < 0.05)	reduced gain	horizontal tracking gain and acceleration ↓ (p < 0.05);Overall impairment milder than AD	^46,53,54
Key points for differentiation		prosaccade latency prolonged, antisaccade error rate increased	Saccadic hypometria, increased errors in antisaccade	severe antisaccade impairment occurs early, but error correction ability is relatively preserved	^{29,31,43,46,48}

AD: Alzheimer's disease; PD: Parkinson's disease; FTD: frontotemporal dementia; Aβ: amyloid-β; NFTs: neurofibrillary tangles; SC: superior colliculus; DLPFC: dorsolateral prefrontal cortex; CV: coefficient of variation; SPEM: smooth pursuit eye movements.

The neural control of SPEM depends on a network encompassing visual cortical areas (e.g., MT/V5), the frontal eye fields, cerebellum, and brainstem nuclei.^29,53 In AD, the observed decline in pursuit gain and acceleration is associated with dysfunction within this network, particularly involving frontal eye fields, medial temporal visual areas, and the cerebellum.⁵³ Widespread cortical neurodegeneration likely disrupts the neural circuits responsible for visual motion prediction and velocity matching.³¹

As a potential auxiliary biomarker, SPEM assessment offers a promising direction for early screening and differential diagnosis in AD. It is suitable for large-scale population screening and could enhance diagnostic accuracy when integrated into a multimodal strategy alongside other biomarkers. However, current evidence is insufficient to support its use as a standalone diagnostic marker in routine clinical practice. Future work must prioritize the development of standardized testing protocols, the establishment of normative data from large samples, and longitudinal studies to clarify its prognostic value.

MCI-related potential eye movement biomarkers

MCI serves as a critical transitional stage between normal cognition and AD dementia, characterized by distinct eye movement abnormalities.^26,30 These abnormalities exhibit predictive value for AD conversion, rendering them promising candidate non-invasive biomarkers to guide early intervention.^31,40,48 The clinically relevant candidate eye movement biomarker panel for MCI centers on three core metrics: antisaccade error rate, prosaccade latency, and fixation stability-their combination yields superior diagnostic performance compared to single neuropsychological scales such as the Mini-Mental State Examination.^40,46,48

Eye movement abnormalities in MCI reflect functional decline rather than overt loss of function.^31,40 Key manifestations include prolonged prosaccade latency (p = 0.024),³⁰ increased antisaccade error rate (p < 0.001),⁴⁰ and elevated lateral fixation deviation (MCI group 16.00 times vs. healthy controls 13.00 times, p < 0.05).¹⁶ Core eye movement functions-such as saccade execution and basic error correction-remain relatively preserved in MCI.^30,31 In contrast, AD dementia is marked by severe functional disruption: loss of saccade direction control, inability to sustain effective smooth pursuit, and near-complete impairment of error correction.^16,31 This shift from functional decline to overt dysfunction mirrors the gradual accumulation of AD-related pathology, enabling eye movement metrics to support both cross-sectional diagnosis and longitudinal disease monitoring^31,40

The pathological specificity of MCI stage eye movement abnormalities enhances their clinical utility.^31,51,53 These changes are closely linked to early tau deposition in the prefrontal cortex (particularly the DLPFC) and parietal cortex,^31,48 as well as cholinergic system dysfunction.³⁸ Eye movement abnormalities in MCI are distinguishable from benign age-related physiological changes. For example, MCI patients exhibit oblique microsaccade direction (36.2 ± 11.7° deviation from horizontal) while healthy controls show horizontal dominance (27.6 ± 9.2°, p = 0.011)⁵¹-age-related changes lack such directional bias and pathological relevance.^51,55 This distinction allows MCI eye movement parameters to identify individuals at high risk of progressing to AD, facilitating targeted early interventions.^40,48 Multimodal integration of eye movement metrics optimizes diagnostic performance: antisaccade error rate offers high sensitivity,⁴⁰ prosaccade latency provides good specificity, and their combination yields an AUC of 0.77 (Gap paradigm)⁴⁰ which outperforms single eye movement indicators or traditional neuropsychological tests⁴⁰ (Table 6).

Table 6.

Summary of eye movement parameters in MCI.

Indicator	Core findings	Pathological correlation	Reference
Prosaccade	latency ↑↑ (p < 0.01)accuracy→ (p > 0.05, but CV ↑, p < 0.05)	Parietal/occipital visuospatial attention network	^28,30,31,40
Antisaccade	error rate ↑↑↑ (p < 0.001)	Prefrontal executive control network (DLPFC)	^30,40
Fixation	Directional bias ↑ (p < 0.05)lateral fixation deviation↑ (p < 0.05)	Posterior parietal cortex and prefrontal attentional control network	^16,51
Combined panel	antisaccade error rate + prosaccade latency + fixation stability	Reflects synergistic impairment of multi-cognitive domain networks (executive control, attention, memory)	^26,30,40

MCI: mild cognitive impairment; DLPFC: dorsolateral prefrontal cortex; CV: coefficient of variation.

Currently, the clinical translation of potential eye movement biomarkers faces challenges. Their diagnostic specificity may be influenced by age, attention, and comorbidities. Future research must focus on standardizing testing protocols, establishing normative data from large samples, and validating their predictive value for MCI-to-AD conversion through longitudinal studies.

Discussion and conclusion

Eye movement research holds great promise for early AD detection and monitoring. Yet the field faces limitations that hinder its clinical translation. These hurdles include sample selection bias, methodological heterogeneity, sparse longitudinal data, and variability in technical equipment. This systematic review pulls together evidence proving multi-modal eye movement abnormalities as reliable non-invasive biomarkers across the entire AD disease spectrum. Three key findings emerge from this synthesis. AD patients show eye movement deficits specific to disease stage. MCI features prolonged prosaccade latency, elevation in antisaccade error rate, and fixation stability decline. AD dementia worsens these anomalies: prosaccade latency becomes further prolonged, and antisaccade error rate rises to a higher level. These abnormalities tie mechanistically to Aβ/tau deposition, NFTs, and frontoparietal network disconnection. Eye movement parameters also aid differential diagnosis: AD differs from PD in progressive prosaccade latency prolongation (opposed to PD's multi-step saccade pattern) and from FTD in poor antisaccade error correction (versus FTD's retained correction capacity).

Besides pathological mechanisms involving Aβ, tau, and the cholinergic system,^38,47 several neural circuits or networks are also implicated in the abnormal eye movements in AD. Eye movements rely on an extensive, hierarchical distributed neural network,⁵⁸ a complex system in human that requires the integration of sensory, motor, attentional, and executive processes.^58–60 To date, no direct studies have identified a common or specific neural circuit impairment that underlying the diverse eye movement abnormalities observed in AD patients. Neuroimaging evidence has linked impaired smooth pursuit and prolonged saccade latency in AD patients to hypometabolism within the precuneus and middle temporal areas.⁴⁶ Additionally, poor performance on the antisaccade task in patients with MCI has been associated with reduced cortical thickness in the superior frontal gyrus, precuneus, and the angular gyrus.⁴²

Although abnormalities in eye movement parameters (e.g., antisaccade error rate) have been consistently observed in patients with MCI/AD,^{27,30,31,48,53} heterogeneity across multiple parameters persists, which limits the generalizability of these findings. Such heterogeneity is largely attributable to methodological variations in sample size, task design, patient population stratification, eye tracker models, and experimental equipment. For instance, one study reported significantly prolonged prosaccade latency in patients with MCI relative to healthy controls,³⁰ whereas another found no such difference in amnestic MCI (aMCI) in its primary analysis.³² A key source of this discrepancy may lie in interstudy differences in patient populations. Indeed, several studies have reported that antisaccade impairments are prominent in aMCI but minimal in non-amnestic MCI (naMCI).⁴⁸ This suggests that failure to stratify MCI into its clinical subgroups may lead to differential detection of latency delays, with aMCI-enriched samples being more likely to reveal such delays. Beyond patient population characteristics, experimental paradigms also play a critical role in driving these heterogeneous results. For example, the overlap task employed requires greater cortical (parietal) engagement compared with the gap task, and this heightened cortical demand may render the overlap task more sensitive to early AD-related deficits.^30,32 Most studies find no significant eye movement abnormalities in preclinical AD, while one study shows that prodromal AD patients exhibit poorer antisaccade accuracy than healthy controls in complex tasks.⁴⁸ These eye movement changes might stem from different pathological load in preclinical AD, the prodromal AD is defined as a later preclinical phase.⁶¹

Eye movement abnormalities have emerged as promising non-invasive candidate biomarkers for the early detection, progression monitoring, and differential AD diagnosis, with accumulating preclinical and preliminary clinical evidence supporting their translational potential.^26,51,53 Nevertheless, their clinical translation is hindered by interconnected challenges, including marked methodological heterogeneity, sample selection bias, sparse longitudinal data, inadequate disease specificity, high costs of specialized eye-tracking tools, and regulatory/clinical acceptance barriers.

The clinical translation and feasibility of eye movement measures need further research. Unifying equipment and task paradigms, eliminating confounding effects from aging and ocular comorbidities, and integrating with existing diagnostic workflows and multimodal biomarker frameworks can further enhance the standardization and clinical reliability of oculomotor candidate biomarkers.

Future research requires multicenter collaborations to validate results across diverse populations,^62,63 developing AI-driven multimodal fusion models for extracting pathological signals,^62,64 creating portable smart systems tailored for primary care settings,⁶² and verifying ties between eye movement abnormalities and AD-specific neural circuit dysfunction.^31,51,53

In summary, eye movement abnormalities hold promise as non-invasive biomarkers for AD, with potential for preclinical screening and differential diagnosis. To advance translation, future research should prioritize AI-driven multi-modal integration, standardized detection protocols, portable device development, and preclinical longitudinal validation.

Footnotes

Acknowledgements

Thanks for the technical support by the Core Facilities Platform, Health Science Center, Ningbo University, and the Laboratory Animal Center of Ningbo University.

ORCID iDs

Yi Li

Shujun Xu

Zhengchun Wang

Qinwen Wang

Ethical considerations

Not applicable

Consent to participate

Not applicable

Consent for publication

Not applicable

Author contribution(s)

Yi Li: Conceptualization; Investigation; Writing – original draft.

Jiaxin Feng: Investigation; Writing – original draft.

Yanhong Xie: Conceptualization; Writing – review & editing.

Qi Yao: Investigation; Writing – review & editing.

Jingbo Yu: Methodology; Writing – original draft.

Shujun Xu: Formal analysis; Writing – review & editing.

Zhengchun Wang: Funding acquisition; Resources; Supervision; Writing – review & editing.

Qinwen Wang: Funding acquisition; Resources; Supervision; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Zhejiang Key Research and Development Project (grant number 2024C03101), National Research and Development Project of China (grant number 2023YFF0724802), Ningbo Key Research and Development Project (grant number 2023Z173), Ningbo Science and Technology Plan Project (grant number 2025S168), and K.C. Wong Magna funded at Ningbo University.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

Not applicable

References

Frisoni

Aho

Brayne

, et al. Alzheimer's disease outlook: controversies and future directions. Lancet 2025; 406: 1424–1442.

Tatulian

. Challenges and hopes for Alzheimer’s disease. Drug Discov Today 2022; 27: 1027–1043.

Colligris

Perez de Lara

Colligris

, et al. Ocular manifestations of Alzheimer’s and other neurodegenerative diseases: the prospect of the eye as a tool for the early diagnosis of Alzheimer’s disease. J Ophthalmol 2018; 2018: 8538573.

Banna

Slayo

Armitage

, et al. Imaging the eye as a window to brain health: frontier approaches and future directions. J Neuroinflammation 2024; 21: 309.

Edwards

Riordan

Buchholz

, et al. Stratification of the extent of visual impairment identifies sex-specific degenerative changes in retinal structure and function during aging. J Integr Neurosci 2025; 24: 25805.

Zia Ul Haq

Zhao

Obeng Apori

, et al. Molecular biomarkers for early detection of Alzheimer's Disease and the complementary role of engineered nanomaterials: a systematic review. Int J Mol Sci 2025; 26: 9282.

Sharma

Chitranshi

Wall

, et al. Trans-synaptic degeneration in the visual pathway: neural connectivity, pathophysiology, and clinical implications in neurodegenerative disorders. Surv Ophthalmol 2022; 67: 411–426.

Seo

Bharmauria

Schütz

, et al. Multiunit frontal eye field activity codes the visuomotor transformation, but not gaze prediction or retrospective target memory, in a delayed saccade task. eNeuro 2024; 11: ENEURO.0413–23.2024.

Cretton

Schipper

Hassan

, et al. Enhancing perceptual, attentional, and working memory demands through variable practice schedules: insights from high-density EEG multi-scale analyses. Cereb Cortex 2024; 34: bhae425.

10.

Goffart

. Orienting gaze toward a visual target: neurophysiological synthesis with epistemological considerations. Vision (Basel) 2025; 9: 6.

11.

Zhang

Hou

, et al. Effectiveness of digital screening tools in detecting cognitive impairment among community-dwelling elderly in Northern China: a large cohort study. J Prev Alzheimers Dis 2025; 12: 100080.

12.

Bischof

Jaeger

Giehl

, et al. Affection of motor network regions by tau pathology across the Alzheimer's disease spectrum. eNeuro 2024; 11: ENEURO.0242–23.2023.

13.

Fonseca

Baker

Dobyns

, et al. Tau accumulation and atrophy predict amyloid independent cognitive decline in aging. Alzheimers Dement 2024; 20: 2526–2537.

14.

Murphy

Abel

Armitage

, et al. Effects of tracker location on the accuracy and precision of the Gazepoint GP3 HD for spectacle wearers. Behav Res Methods 2024; 56: 43–52.

15.

Zhao

Xie

Fonzo

, et al. Dementia subtypes defined through neuropsychiatric symptom-associated brain connectivity patterns. JAMA Netw Open 2024; 7: e2420479.

16.

Lian

Guo

, et al. Abnormal eye movements: relationship with clinical symptoms and predictive value for Alzheimer's disease. Front Aging Neurosci 2024; 16: 1471698.

17.

Leitner

Hutzler

Schuster

, et al. Eye-tracking-based visual field analysis (EFA): a reliable and precise perimetric methodology for the assessment of visual field defects. BMJ Open Ophthalmol 2021; 6: e000429.

18.

Hirota

Kato

Fukushima

, et al. Analysis of smooth pursuit eye movements in a clinical context by tracking the target and eyes. Sci Rep 2022; 12: 8501.

19.

Lisong

Jiaqi

Yang

. A logistic regression model for diagnosing Alzheimer's disease based on eye movement tracking technology. J Army Med Univ 2023; 45: 102–110.

20.

Sekar

Panouillères

MTN

Kaski

. Detecting abnormal eye movements in patients with neurodegenerative diseases - current insights. Eye Brain 2024; 16: 3–16.

21.

Sperling

Aisen

Beckett

, et al. Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 280–292.

22.

Albert

DeKosky

Dickson

, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 270–279.

23.

First

. Diagnostic and statistical manual of mental disorders, 5th edition, and clinical utility. J Nerv Ment Dis 2013; 201: 727–729.

24.

Liu

Zhang

Deng

, et al. The 3D characteristics of nystagmus in posterior semicircular canal benign paroxysmal positional vertigo. Front Neurosci 2022; 16: 988733.

25.

Crawford

Devereaux

Higham

, et al. The disengagement of visual attention in Alzheimer's disease: a longitudinal eye-tracking study. Front Aging Neurosci 2015; 7: 118.

26.

Boujelbane

Trabelsi

Salem

, et al. Eye tracking during visual paired-comparison tasks: a systematic review and meta-analysis of the diagnostic test accuracy for detecting cognitive decline. J Alzheimers Dis 2024; 99: 207–221.

27.

Kahana Levy

Lavidor

Vakil

. Prosaccade and antisaccade paradigms in persons with Alzheimer's disease: a meta-analytic review. Neuropsychol Rev 2018; 28: 16–31.

28.

Rane

Dash

Dutt

, et al. Distinctive visual tasks for characterizing mild cognitive impairment and dementia using oculomotor behavior. Front Aging Neurosci 2023; 15: 1125651.

29.

Wang

Kapoor

Fielding

, et al. Saccadic eye movements in neurological disease: cognitive mechanisms and clinical applications. J Neurol 2025; 272: 539.

30.

Opwonya

Wang

Jang

, et al. Inhibitory control of saccadic eye movements and cognitive impairment in mild cognitive impairment. Front Aging Neurosci 2022; 14: 871432.

31.

Yao

Liu

, et al. The link between eye movements and cognitive function in mild to moderate Alzheimer's disease. Exp Brain Res 2025; 243: 39.

32.

Yang

Wang

, et al. Long latency and high variability in accuracy-speed of prosaccades in Alzheimer’s disease at mild to moderate stage. Dement Geriatr Cogn Dis Extra 2011; 1: 318–329.

33.

Scheltens

Blennow

Breteler

MMB

, et al. Alzheimer's disease. Lancet 2016; 388: 505–517.

34.

Kühn

Averdunk

Huckemann

, et al. Correlates of polyneuropathy in Parkinson’s disease. Ann Clin Transl Neurol 2020; 7: 1898–1907.

35.

Garbutt

Matlin

Hellmuth

, et al. Oculomotor function in frontotemporal lobar degeneration, related disorders and Alzheimer's disease. Brain 2008; 131: 1268–1281.

36.

Burrell

Hornberger

Carpenter

RHS

, et al. Saccadic abnormalities in frontotemporal dementia. Neurology 2012; 78: 1816–1823.

37.

Noiret

Carvalho

Laurent

, et al. Saccadic eye movements and attentional control in Alzheimer's disease. Arch Clin Neuropsychol 2017; 33: 1–13.

38.

Hampel

Mesulam

Cuello

, et al. The cholinergic system in the pathophysiology and treatment of Alzheimer’s disease. Brain 2018; 141: 1917–1933.

39.

Panza

Lozupone

Logroscino

, et al. A critical appraisal of amyloid-β-targeting therapies for Alzheimer disease. Nat Rev Neurol 2019; 15: 73–88.

40.

Chehrehnegar

Shati

Esmaeili

, et al. Executive function deficits in mild cognitive impairment: evidence from saccade tasks. Aging Ment Health 2022; 26: 1001–1009.

41.

Smith

Crawford

. The inhibitory effect of a recent distractor: singleton vs. multiple distractors. Exp Brain Res 2024; 242: 1745–1759.

42.

Heuer

Mirsky

Kong

, et al. Antisaccade task reflects cortical involvement in mild cognitive impairment. Neurology 2013; 81: 1235–1243.

43.

Terao

Fukuda

Yugeta

, et al. Initiation and inhibitory control of saccades with the progression of Parkinson's disease – changes in three major drives converging on the superior colliculus. Neuropsychologia 2011; 49: 1794–1806.

44.

Schützmann

Hasecke

Bachmann

, et al. Endo-lysosomal Aβ concentration and pH trigger formation of Aβ oligomers that potently induce Tau missorting. Nat Commun 2021; 12: 4634.

45.

Yamazaki

Zhao

Caulfield

, et al. Apolipoprotein E and Alzheimer disease: pathobiology and targeting strategies. Nat Rev Neurol 2019; 15: 501–518.

46.

Lage

López-García

Bejanin

, et al. Distinctive oculomotor behaviors in Alzheimer's disease and frontotemporal dementia. Front Aging Neurosci 2020; 12: 603790.

47.

Sperling

Aisen

Beckett

, et al. Toward defining the preclinical stages of Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 280–292.

48.

Wilcockson

TDW

Mardanbegi

Xia

, et al. Abnormalities of saccadic eye movements in dementia due to Alzheimer's disease and mild cognitive impairment. Aging (Albany NY) 2019; 11: 5389–5398.

49.

Crawford

Higham

Renvoize

, et al. Inhibitory control of saccadic eye movements and cognitive impairment in Alzheimer’s disease. Biol Psychiatry 2005; 57: 1052–1060.

50.

Martinez-Conde

Macknik

Hubel

. The role of fixational eye movements in visual perception. Nat Rev Neurosci 2004; 5: 229–240.

51.

Kapoula

Yang

Otero-Millan

, et al. Distinctive features of microsaccades in Alzheimer's disease and in mild cognitive impairment. Age (Dordr) 2014; 36: 535–543.

52.

Munoz

Everling

. Look away: the anti-saccade task and the voluntary control of eye movement. Nat Rev Neurosci 2004; 5: 218–228.

53.

Garbutt

Matlin

Hellmuth

, et al. Oculomotor function in frontotemporal lobar degeneration, related disorders and Alzheimer's disease. Brain 2008; 131: 1268–1281.

54.

Gibbs

Huxley

Readman

, et al. Naturalistic eye movement tasks in Parkinson's disease: a systematic review. J Parkinsons Dis 2024; 14: 1369–1386.

55.

Lee

. The development of an Alzheimer's diagnostic sensor and algorithm using microsaccades biomarkers. Annu Int Conf IEEE Eng Med Biol Soc 2024; 2024: 1–4.

56.

Farber

Jiang

Michiels

, et al. Microsaccade activity during visuospatial working memory in early-stage Parkinson's disease. J Eye Mov Res 2025; 18: 46.

57.

Zacharia

Kaski

Bouthour

, et al. Effects of deep brain stimulation frequency on eye movements and cognitive control. NPJ Parkinsons Dis 2023; 9: 50.

58.

Coiner

Pan

Bennett

, et al. Functional neuroanatomy of the human eye movement network: a review and atlas. Brain Struct Funct 2019; 224: 2603–2617.

59.

McDowell

Dyckman

Austin

, et al. Neurophysiology and neuroanatomy of reflexive and volitional saccades: evidence from studies of humans. Brain Cogn 2008; 68: 255–270.

60.

Hanes

Wurtz

. Interaction of the frontal eye field and superior colliculus for saccade generation. J Neurophysiol 2001; 85: 804–815.

61.

Watermeyer

Calia

. Neuropsychological assessment in preclinical and prodromal Alzheimer disease: a global perspective. J Glob Health 2019; 9: 010317.

62.

Zhang

Pan

, et al. A portable and efficient dementia screening tool using eye tracking machine learning and virtual reality. NPJ Digital Medicine 2024; 7: 219.

63.

Giangrande

Delatour

Andreasson

, et al. Harmonization and standardization of biofluid-based biomarker measurements for AT(N) classification in Alzheimer's disease. Alzheimers Dement (Amst) 2023; 15: e12465.

64.

Patterson

Bounds

Wagner

, et al. Oculomics: a crusade against the four horsemen of chronic disease. Ophthalmol Ther 2024; 13: 1427–1451.