Analysis of eye movement features in patients with Alzheimer's disease based on intelligent eye movement analysis and evaluation system

Abstract

Background

The early identification of Alzheimer's disease (AD) benefits patients, so creating a simple and convenient method is crucial for diagnosing early symptoms.

Objective

To offer a potential approach for the early detection of both AD and mild cognitive impairment (MCI).

Methods

Eye movement data from 66 patients were divided into three groups, including healthy control group (HC), MCI group, and AD group. We searched for parameters that can detect MCI at an early stage and drew receiver operating characteristic (ROC) curves. The correlation between eye movement parameters and cognitive scores was analyzed.

Results

The MCI group differed from the HC group in error correction rate of antisaccade (p = 0.008) and total offset degrees (>4°) (p = 0.011) of lateral fixation. The AD group had different overlap prosaccade accuracy (p = 0.025), latency (p = 0.009) and average completion time (p = 0.015), gap prosaccade latency (p = 0.005) and average completion time (p = 0.005), antisaccade accuracy (p = 0.006), error correction rate (p < 0.001) and average saccade velocity (p = 0.035), and lateral fixation accuracy (p = 0.018), total offset degrees (>4°) (p = 0.041) compared to the HC group. The AD group differed significantly from the MCI group in accuracy (p = 0.001) and error correction rate (p = 0.044) of antisaccades, the latency (p = 0.009) and average completion time (p = 0.025) of overlap prosaccade and the latency (p = 0.038) of gap prosaccade, these parameters can serve as indicators to monitor the progress of the disease. Lateral fixation combined with antisaccade was more conducive to identifying MCI patients with the area under the ROC curve of 0.837. Most eye movement parameters had a light to moderate correlation with cognitive scores.

Conclusions

Eye movements can be used for early identification of MCI/AD patients and to monitor disease progression.

Keywords

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

Introduction

With the global population aging, dementia has emerged as a prominent contributor to disability among the elderly, resulting in significant economic burdens on individuals, families, and society at large.¹ The incidence and mortality of Alzheimer's disease (AD) are steadily rising. Early identification of AD is crucial for slowing down its progression. Mild cognitive impairment (MCI) is an intermediate stage between normal cognition and AD dementia. Currently, the diagnosis of AD primarily relies on various assessments, including neuropsychological scales, cerebrospinal fluid AD four-item examination, and positron emission tomography–computed tomography (PET-CT) scans. However, each of these approaches has its limitations that need to be considered. At present, scale screening is the most commonly used method, but its sensitivity and specificity vary, and factors such as the patient's age, education level, cultural background and others can affect the doctor's interpretation of the results. Moreover, the cost of training doctors to administer these assessments can be significant.² Developing a simple, convenient and non-invasive biomarker is crucial for diagnosing early symptoms and monitoring disease progression.

The eyes are intricately connected to the brain, as they serve as the organs closest to it. Over half of the neural pathways in our bodies are dedicated to eye movement and vision. Remarkably, every region of the brain contains neurons that are associated with eye movement.³ Kusne et al.⁴ proposed that neurodegenerative pathologies could manifest in visual brain structures earlier than in other areas. Among them, saccade, smooth pursuit and fixation have been studied most as the three most commonly used paradigms.⁵ Previous studies have shown that eye movements can detect dementia early⁶ and that tracking eye movement can serve as a non-verbal and cognitively less demanding means of assessing disease advancement in patients with cognitive impairment.⁷ In recent years, numerous studies have demonstrated the potential of eye tracking in detecting MCI/AD, where eye movement indicators serve as early neurodegenerative disease markers.^2,8–10 By combining various paradigms, this research aims to enhance our understanding of eye movement patterns in these three groups, providing an objective and real basis for identifying early MCI/AD in clinical practice and highlighting the potential application of eye movement tracking systems in MCI/AD.

Methods

Participants

During the period from September 2022 to May 2023, we conducted a study at the Second Hospital of Hebei Medical University. Our research involved 23 patients with MCI and 23 patients with AD. Additionally, we recruited 20 healthy individuals from the local community to serve as a control group. Baseline and imaging data were collected from all subjects. Eye movement parameters were assessed using the EyeKnow™ intelligent eye movement analysis evaluation system. This system measured various eye movement tasks, including overlap prosaccade, gap prosaccade, antisaccade, smooth pursuit, central fixation, and lateral fixation.

Selection criteria

Inclusion criteria: Healthy control group: (1) Normal eye movement function and ability to complete all test tasks. (2) Normal neurological examination, no history of cerebrovascular disease or other diseases affecting cognition. (3) Normal cognitive function, Mini-Mental State Examination (MMSE) ≥ 26, Clinical Dementia Rating (CDR) = 0. MCI group: (1) Age ≥ 50 years old. (2) Cognitive impairment reported by patients or informants, or detected by experienced clinicians, with objective evidence of impairment in one or multiple cognitive domains. (3) Slight impairment in complex instrumental activities of daily living but maintaining independent daily living abilities. (4) The diagnosis of dementia has not been reached. (5) CDR = 0.5. The final diagnosis of MCI was made by consensus between two neurologists specializing in cognitive impairment based on medical history, educational level and neuropsychological scores. AD group: (1) Age ≥ 50 years old. (2) Cerebrospinal fluid (CSF) AD four items support the diagnosis of AD, or plasma p-tau217 positive or meet the National Institute on Aging-Alzheimer's Association (NIA-AA) probable AD dementia criteria or the 2018 NIA-AA diagnostic framework. (3) Consciousness is clear, able to cooperate with the doctor's instructions to complete all test tasks. (4) No evidence of mass lesions or age-inappropriate periventricular and deep white matter lesions on MRI (score ≤ 2). (5) Hachinski Ischemia Score ≤ 4. (6) The medial temporal atrophy Visual rating Scale (MTA) was assessed as grade 2 or higher. Exclusion criteria: (1) Cognitive impairment attributed to other etiologies; (2) Any neurological disorders that may impact the study; (3) Individuals with significant emotional and behavioral abnormalities as well as mental and behavioral disorders. (4) Participants with color blindness, visual impairments, ptosis, or difficulties in measurement due to their failure in a preliminary calibration test preceding this trial.

All participants provided written consent to participate in the study. The research protocol of this study was approved by the ethics committee of the Second Hospital of Hebei Medical University (approval number: 2024-R039).

Experimental setup

The EyeKnow ™ intelligent eye movement analysis evaluation system from Zhongke Ruiyi is used as an eye tracker in this research. It features a dual-screen display with immersive capabilities, boasting a 4 K resolution of 3664 × 1920 pixels and a sampling rate of 120 Hz. The system also has a wide field of view of 98°. Furthermore, the system's data tracking performance is at a remarkable 100%. EyeKnow Eye-tracking System was registered with NMPA (National Medical Products Administration), China, manufactured by Beijing CAS-Ruiyi Information Technology Co, Ltd As medical devices regulation required, EyeKnow has been rigorously validated throughout the registration and approval process to ensure their feasibility, reliability, stability, precision, and low risk.

Eye movement paradigm

To ensure the accuracy of the eye tracker throughout the task, a nine-point calibration was performed before the experiment. The task begins with written instructions and guidance. The experiment consists of three paradigms: saccade, smooth pursuit, and fixation.

Saccades

Saccades are a type of eye movement that can be categorized into three paradigms: overlap prosaccade, gap prosaccade and antisaccade.

Overlap prosaccade

The central point appears first, and then 1200 ms later, the central point disappears. However, 200 ms before the disappearance of the central point, the target point randomly appears in up, down, left, or right positions. When the target point appears in any position, the patient needs to execute a quick saccade to look at it. This task is repeated ten times.

Gap prosaccade

The target point appears first in the center position, then disappears 800 ms later. After a blank period of 200 ms, the system point appears randomly in up, down, left, or right positions. When the target point appears in any position, the patient needs to execute a quick saccade to look at it. This task is repeated ten times.

Antisaccade

Subjects are asked to fixate on the central fixation point until 1000 ms later, when the target point appears randomly in the upper, lower, left, and right positions. When the target point appears, they have to quickly and accurately shift their gaze in the opposite direction of the target. This task is repeated ten times.

Smooth pursuit

This test involves smooth pursuit in the horizontal direction. Participants are required to continuously follow and gaze at the target, which moves at a speed of 10°/s for a total duration of 15 s, traversing a maximum angle of 20°.

Fixation

The tests for sustained fixation on a target including central fixation and lateral fixation.

Central fixation

Patients are required to gaze at a point that appears in the center, maintain their gaze stability, and fixate for a total duration of 10 s with a target angle of 0°.

Later Fixation

The target point is sequentially presented at five different positions: the center, 15° up, 15° right, 15° down, and 15° left. Participants are instructed to fixate on the target dot as it appeared in each position. The total fixation duration is 30 s.

Statistical analysis

SPSS 27.0 was used for data statistics and analysis. Measurement data with normal distribution were presented as average ± standard deviation ( $\bar{x} \pm s$ ) for comparison between groups, and one-way analysis of variance was used for comparison between groups. Data with non-normal distribution was presented as median (interquartile range) [M (Q1, Q3)] for comparison between groups, and Kruskal-Wallis Test was used for comparison between groups. Frequency data was presented as frequency (percentage). Spearman correlation analysis was used to test the correlation between eye movement parameters and scale scores of MMSE and Montreal Cognitive Assessment (MoCA), and canonical correlation analysis was used to test the correlation between total eye movement data and MMSE. p< 0.05 was considered statistically significant. Receiver operating characteristic curve (ROC) analysis was used to compare the diagnostic value of eye movement parameters with statistical significance in distinguishing MCI from healthy elderly people.

Results

Demographics

The baseline demographic characteristics and cognitive scores of the participants are shown in Table 1, and the years of education were lower in the AD group. There were significant differences in MMSE and MoCA scores among the three groups; There were no significant differences in age and gender among the three groups.

Table 1.

Comparison of demographic data and neuropsychological scales of the three groups.

	HC	MCI	AD	p	Post Hoc Contracts(p)
Demographic	HC	MCI	AD	p	HC vs. MCI	HC vs. AD	MCI vs. AD
Education (y)	10.50 (9.00,14.25)	12.00 (9.00,15.00)	9.00 (6.00,12.00)	0.028	1.00	0.166	0.031
Sex (M / F)	9/11	7/16	9/14	0.615	-	-	-
Age (y)	66.50 (60.25,72.00)	69.00 (64.00,76.00)	70.00 (62.00,73.00)	0.18	-	-	-
Neuropsychological
MMSE	29.00 (28.00,29.00)	26.00 (24.00,27.00)	19 .00 (15.00,20.00)	<0.001	0.007	0.000	0.000
MoCA	23.5 (22.25,25.75)	20.00 (18.00,22.00)	11.00 (9.00,15.00)	<0.001	0.005	0.000	0.000

HC: healthy control; MCI: mild cognitive impairment; AD: Alzheimer's disease

Eye movement data

Various eye movement parameters show different performance in cognitive impairment patients, after adjusting for confounding factors such as gender, age, and education, the results were shown in Figure 1.

Saccades

Table 2 presents the results of the eye movement tests. In overlap prosaccade, the AD group had a significant difference in accuracy (p = 0.025) compared to the HC group, and differed significantly from the other two groups in latency (p = 0.009) and average completion time (p = 0.015). However, overlap prosaccade did not show significant differences in distinguishing between the HC and MCI groups. In gap prosaccade, the AD group had a significantly longer latency (p = 0.005) and average completion time (p = 0.005) than the HC group, and there was a significant difference in latency (p = 0.038) between the MCI and AD groups. In antisaccade, both the AD group (p = 0.006) and MCI group (p = 0.001) had significantly lower accuracy than the HC group. The AD group had significant lower error correction rate (p < 0.05) compared to the other two groups, and the AD group had a lower average saccade velocity (p = 0.035) than the HC group and a longer average completion time (p = 0.038) than the MCI group. The MCI group had a significant difference in error correction rate of antisaccade (p = 0.008) compared to the HC group.

Table 2.

Comparison of eye movement parameters of the three groups.

Paradigms	HC	MCI	AD	p	Post Hoc Contracts (p)
Overlap prosaccade test					HC-MCI	HC-AD	MCI-AD
Accuracy (%)	100 (100,100)	100 (80, 100)	90 (62.5, 100)	0.017	0.193	0.025	0.173
Saccade latency (ms)	335.76 (311.67, 378.23)	353.72 (291.5, 385.82)	466.68 (330.42, 501.32)	0.002	0.912	0.009	0.009
Average completion time (ms)	401.55 (371.63, 515.91)	434.42 (400.90, 491.29)	523.48 (432.81, 704.7)	0.009	0.747	0.015	0.025
Average saccade velocity (°/s)	259.47 (177.07, 309.14)	261.40 (172.34, 307.42)	241.5 (128.11, 297.44)	0.595	0.841	0.342	0.445
Gap prosaccade test
Accuracy (%)	100 (100, 100)	100 (87.51, 100)	100 (70, 100)	0.339	0.523	0.174	0.416
Saccade latency (ms)	260.09 (235.07, 293.20)	269.24 (248.31, 299.05)	316.5 (276.4, 394.04)	0.003	0.263	0.005	0.038
Average completion time (ms)	320.84 (300.33, 370.68)	362.16 (308.36, 439.64)	426.95 (347.42, 556.80)	0.002	0.075	0.005	0.111
Average saccade velocity (°/s)	224.43 (184.01, 294.64)	232.16 (177.46, 274.05)	203.37 (158.65, 274.95)	0.633	0.371	0.463	0.870
Anti-saccade test
Accuracy (%)	40 (5, 69.17)	30 (0, 50)	0 (0, 14.29)	<0.001	0.358	0.006	0.001
Saccade latency (ms)	377.17 (309.74, 462.75)	391.88 (312.8, 428.22)	402.12 (285.84, 462.43)	0.784	0.810	0.499	0.659
Average completion time (ms)	455.85 (411.68, 531.91)	466.20 (395.54, 548.25)	540.34 (488.23, 640.75)	0.04	0.806	0.058	0.038
Average saccade velocity (°/s)	267.37 (208.02, 302.13)	224.14 (170.14, 303.62)	219.21 (112.82, 248.93)	0.063	0.539	0.035	0.107
Error correction rate (%)	85.72 (54.17, 100)	33.33 (20, 71.43)	0 (0, 70)	<0.001	0.008	<0.001	0.044
Smooth pursuit test
Start-up duration (ms)	545.56 (363.18, 756.65)	582.91 (389.08, 753.64)	887.62 (704.6, 2956.55)	0.011	0.142	0.082	0.310
Acceleration	47 (26.74, 72.43)	34.64 (25.34, 82.32)	22.69 (2.4, 56.35)	0.128	0.477	0.261	0.289
Smooth pursuit velocity (°/s)	21.08 (19.01, 24.93)	19.43 (11.66, 28.89)	17.98 (7.08, 26.69)	0.275	0.541	0.128	0.318
Total offset degrees (>4°)	79.22 (48.51, 191.95)	136.53 (81.68, 182.68)	103.1 (47.96, 192.18)	0.869	0.825	0.744	0.602
Central fixation test
Accuracy (%)	94.03 (85.97, 95.9)	89.99 (81.94, 94.30）	90.2 (79.97, 95.28)	0.332	0.193	0.653	0.321
Total offset degrees (>4°)	36.29 (19.72, 94.29)	53.57 (28.67, 112.46)	55.13 (20.16, 121.42)	0.535	0.283	0.506	0.664
Total offset Duration (ms)	595.58 (409.45, 1400.37)	996.78 (567.44, 1802.67)	987.09 (471.37, 1994.97)	0.332	0.193	0.653	0.321
Lateral fixation test
Accuracy (%)	91.09 (76.40, 92.48)	82.93 (62.73, 88.49)	73.91 (48.34, 88.77)	0.018	0.051	0.018	0.550
Total offset degrees (>4°)	149.49 (104.93, 236.07)	268.79 (152.37, 347.74)	227.67 (135.87, 351.55)	0.013	0.011	0.041	0.443
Total offset duration (ms)	2677.98 (2259.58, 7083.36)	5124.53 (3456.43, 11,195.49)	7832.61 (3371.01, 15,516.64)	0.018	0.051	0.018	0.550

Smooth pursuit

The three groups showed no statistically significant differences in any of the parameters, after controlling for demographic characteristics.

Fixation

There were no significant differences in central fixation; however, in lateral fixation, there were significant differences in the total offset degrees (>4°) (p = 0.011) between the MCI and HC groups; the total offset degrees (>4°) (p = 0.041), total offset duration (p = 0.018) and accuracy (p = 0.018) had significant differences between the AD and HC groups.

Figure 1.

Eye movement parameters of the three groups.

Receiver operating characteristic curve

As shown in Figure 2, ROC curve analysis was conducted for significant factors between MCI and HC group. The antisaccade error correction rate and the total offset degrees (>4°) of lateral fixation were selected for ROC analysis. The results indicated that the area under the curve (AUC) for the antisaccade error correction rate was 0.757, with a sensitivity of 0.52 and a specificity of 0.95. The AUC for the total offset degrees (>4°) of lateral fixation was 0.743, with a sensitivity of 0.74 and a specificity of 0.75. Consequently, we combined the total offset degrees (>4°) with antisaccade error correction rate to draw a ROC curve, which yielded an AUC of 0.837, with a sensitivity of 0.78 and a specificity of 0.80. Combining these two parameters can improve sensitivity and help us detect MCI patients.

Figure 2.

ROC curves comparing best eye movement parameters to differentiate MCI from HC.

Correlation

Extracting the meaningful parameters and drawing the heat map (Figure 3). Spearman correlation analysis revealed that most parameters in each eye movement paradigm had a light to moderate correlation with MMSE and MoCA scores. However, it appeared that there was no correlation between the central fixation paradigm and cognitive scale assessments. Subsequently, a typical correlation analysis revealed that MMSE was significantly associated with overall eye movement parameters (r = 0.9, p = 0.015).

Figure 3.

Correlation of eye movement data with MMSE and MoCA.

Discussion

In this research, we examined the disparities in saccades, smooth pursuit, and fixation across three groups. The aim is to explore an eye movement paradigm for early recognition of MCI, and to determine which paradigm can monitor the progression of AD. The study found notable disparities in antisaccade and lateral fixation measures between the MCI group and the HC group. Compared with the MCI group, AD group showed significant differences in the accuracy and error correction rate of antisaccade, the latency of the prosaccade, and the average completion time of the overlap saccade. Moreover, significant differences were observed in the parameters of overlap prosaccade, gap prosaccade, antisaccade and lateral fixation between the AD group and the HC group. These findings highlight the distinct characteristics exhibited by individuals with MCI and AD in comparison to their healthy counterparts.

Saccades

The saccades encompassed overlap prosaccade, gap prosaccade, and antisaccade. Our research shows no significant differences in prosaccade parameters between the MCI group and the HC group. Compared with AD patients, MCI patients have relatively preserved cognitive functions and exhibit prosaccadic performance similar to healthy elderly controls.⁹ We found that latency of overlap and gap prosaccades in AD patients were statistically significant for both control and MCI patients. Degenerative changes in posterior parietal and/or frontal lobes may explain the prolonged latency of prosaccade in AD patients.¹¹ In saccades, activity in the frontal eye field (FEF) triggers spontaneous saccades, and parietal eye fields (PEF) trigger reflex saccades.¹² The supplementary eye field (SEF) serves various functions, including controlling and supervising saccadic movements,⁹ and signals from these cortical and subcortical areas converge at the intermediate layer of the superior colliculus (SC).¹⁰ Saccade signals from both SC and directly from cortical regions transmit to the saccade generator located within reticular structures that send final commands to corresponding ocular motor neurons for movement initiation.¹¹ The cerebellum is involved in generating precise corrective feedback to ensure accurate prosaccades.¹³ Our study has demonstrated that individuals with AD exhibit prolonged latency of prosaccade, whereas those with MCI do not demonstrate this difference. Furthermore, there is no discrepancy in antisaccade latency between AD and MCI patients when compared to the HC group.^9,14 This suggests that the increased cognitive demand and difficulty associated with the antisaccade task may contribute to delayed latency across all three groups. Prior research has indicated that longer latency of antisaccade is linked to aging, while the latency of prosaccade remains unaffected.¹⁵ This highlights the significance of using prosaccadic paradigms in the follow-up of MCI patients.

In both gap and overlap prosaccades, we observed that a statistically significant difference in average completion time between the AD group and the HC group. Specifically, the attention allocation and attention span of AD patients were impaired, and the brain was unable to consciously perceive multiple pieces of information simultaneously. AD patients experienced a gradual decline in attention control, which is closely linked to the initiation of prosaccade.¹⁴ Some researchers argue that the primary cognitive function may remain intact during the initial stages of cognitive decline.¹⁶ This could explain the considerably longer time taken by the AD group to initiate saccades and complete the task, while no noticeable abnormalities were observed in the MCI group. Additionally, analysis of the data revealed that all three groups experienced a longer latency for overlap tasks compared to gap tasks. The general reduction in saccades latency in the gap saccade compared with the overlap saccade is known as the gap effect.¹⁷ In fact, our brain has a state of attentional blink between the processing of two consecutive stimuli. During this time, because our brain is still processing the previous stimulus, we appear to ignore the subsequent stimulus.¹⁸ This can be attributed to the more intricate triggering of prosaccade in the overlap paradigm, which involves the activation of additional cortical regions such as the FEF circuits.¹⁹ We found that compared to HC group, AD patients exhibit a strong gap effect, indicating that difficulty in separating visual attention is the basis for impaired overlap prosaccade in AD.

Accumulating evidence suggests that patients with AD have subtle impairments in cognitive inhibitory control in addition to working memory impairment in the early stages, which are often undetectable by traditional cognitive assessments.²⁰ Inhibitory control is one of the core components of executive function, which involves active suppression, interruption, or delay of behavior. Based on neuropsychological longitudinal investigations, it has been determined that executive function starts deteriorating 2–3 years before an AD diagnosis,²¹ and as a reliable indicator of executive dysfunction, antisaccade plays a crucial role in identifying this decline.¹⁰ Antisaccade performance involves two primary saccadic processes: inhibiting the movement towards the target and shifting gaze in the opposite direction. The neural processes underlying antisaccade behavior can be broken down into several stages. Firstly, a signal is sent from the DLPFC to the SC in order to suppress the reflexive transmission towards the target. Next, the direction of the upcoming saccade is reversed by the posterior parietal cortex and the FEF, redirecting it away from the target area and towards the opposite side of the visual field. Finally, the antisaccade away from the target are initiated by the FEF through the saccade system.²²

We found that the MCI and AD groups performed worse in error correction rate of antisaccade than the HC group. And the AD group had lower accuracy and average saccade velocity in antisaccade. Wilcockson et al.²³ suggested that the inhibition control of saccadic eye movement could potentially serve as an early biomarker for AD. We observed a notable disparity in the error correction rate of antisaccade between the two groups, supporting previous research.^10,23 Previous studies claim that the accuracy of antisaccade can differentiate between individuals with MCI and those in the HC group.^8,9 However, in our study, we did not observe a significant difference in accuracy between the MCI group and the HC group. This suggests that the error correction rate of antisaccade may hold greater sensitivity in distinguishing MCI when compared to other parameters. Crawford et al.¹⁰ proposed that AD is characterized by a notable rise in uncorrected errors. Additionally, individuals with MCI are more prone to not rectifying their errors, which can be attributed to alterations in their self-monitoring and error correction networks.²⁴ Impairments in spatial working memory play a crucial role in uncorrected errors seen in patients with MCI,¹⁰ and several studies suggest that the percentage of uncorrected antisaccade may serve as a potential biomarker for early diagnosis of MCI and mild AD.²⁵ Peltsch et al.²⁶ propose that the antisaccade task effectively detects subtle defects in MCI, particularly in the area of selective attention deficit. It is indisputable that antisaccade can identify MCI, however, which parameter of antisaccade is more sensitive for the identification of MCI is still controversial and needs further verification.

Antisaccade involves more lobes of the brain, and damage to the frontal and parietal lobes can cause abnormal antisaccade. In fact, In early stages of AD, patients may experience impaired inhibitory control due to damage to frontal lobe function.²⁷ DLPFC dysfunction in patients contributes to sending SC inhibition to the target signal transmission, which may be abnormal in the process, leading to patients with AD being unable to suppress their instinctive reactions. Kaufman et al.²⁸ propose that the DLPFC could serve as a reliable indicator for predicting the progression from MCI to AD. These deficits may become apparent prior to the detection of memory decline through clinical evaluation.²⁹ We also observed a difference in the accuracy and error correction rate of antisaccade between MCI and AD patients. We suggest that it is feasible to use the accuracy and error correction rate of antisaccade as a reactive indicators of DLPFC function impairment to monitor the progression of the disease.

Smooth pursuit

The function of smooth pursuit eye movement is to stabilize the retinal image of small moving targets. It shares neural circuits with saccadic eye movements.³⁰ Smooth pursuit extracts motion information from moving targets. This information is extracted by the lateral occipital cortex and sent to the FEF and SEF areas for tracking targets.³¹ However, in our study on smooth pursuit, after adjusting for confounding factors, the three groups showed no statistically significant differences in any of the parameters. Of course, at least one previous study has reported that Moser et al.³² discovered that individuals with mild AD demonstrate smooth pursuit function comparable to that of healthy individuals. However, Zaccara et al.³³ found that individuals with AD exhibited consistently lower average peak velocity of smooth pursuit compared to the HC group. This difference was particularly significant when the target speed exceeded 30 degrees per second, with the disparity becoming more pronounced as the speed increased. This could be attributed to the target speed set at 10 degrees per second and the inclusion of mild AD patients in our study. However, in order to gain a more comprehensive understanding, future research should consider modifying the smooth pursuit speed and incorporating participants with moderate and severe AD who are able to cooperate. This will enable the observation of potential discrepancies in smooth pursuit between these three groups.

Fixation

Fixation actually never remains completely still during gaze, but generates small (<1°) involuntary movements.³⁴ In the lateral fixation, the AD group was significantly lower than the HC group in the accuracy, and the total offset degrees (>4°) was significantly different between the MCI and HC group. This part highlights the impact of cognitive impairments on attentional abilities in individuals with AD and MCI. Specifically, the AD group demonstrated a significant increase in offset duration and a decrease in accuracy during total deviation time. This suggests that lateral fixation may provide a basis for early diagnosis of MCI. Our data further reveals significant differences in total offset degrees (>4°) and total deviation time during lateral fixation between the MCI group and HC group. Lateral fixation involves the activation of scanning-related neurons and inhibition of neurons controlled by the superior colliculus. The MCI group also exhibited impairment in attentional transfer. Okonkwo et al.³⁵ found that attentional transfer was particularly affected in MCI patients, while simple attention remained intact. However, no statistical differences in central fixation were found among the three groups. Fixation duration represents the relative attention to an object, and the longer the average fixation duration, the higher the level of attention.³⁶ Sustained attention refers to the ability to focus on a stimulus for a period of time. Sustained attention is processed by the alert network in the brain, the main regions of which include the right medial frontal lobe, the dorsolateral prefrontal cortex, and the parietal cortex. This may be related to the fact that the patients we included were milder and had a relatively simple central fixation duration of 10 s. It has been suggested that impaired attention is a major cause of changes in eye movements in patients with AD.³⁷ Previous research has indicated that AD patients’ ability to focus their attention on targets is not significantly impaired, and they perform well in tests of sustained attention under low task requirements.³⁸ However, they do exhibit impairment in transferring visual spatial attention.¹⁶ AD patients display significant differences in tasks that necessitate attention transfer.

Eye tracking technology has the potential to enhance cognitive assessments in the elderly and clinical populations by overcoming movement and physical interference. By analyzing the above data,the ROC curve was drawn by combining the total offset degrees (>4°) with the error correction rate of antisaccade, and the AUC was 0.837, with a sensitivity of 0.78 and a specificity of 0.80. Combining these two parameters significantly improves sensitivity and help us detect patients with MCI. And unlike traditional methods, eye tracking technology is not affected by ceiling and floor effects, which often hinder the accurate evaluation of cognitive performance in patients. Additionally, it enables a more effective comparison between patients and age-matched healthy individuals. By complementing other biomarkers, eye tracking technology has the potential to identify individuals at risk of developing MCI/AD in its pre-symptomatic stages and to monitor disease progression with greater precision over time.

Limitations

However, this study has the following limitations: 1. Limited Sample Size: To enhance the understanding of the relationship between eye movements and cognitive impairment in AD, the future research should focus on increasing the sample size, constructing and validating predictive models. 2. Lack of Cognitive Domain Scale: Further exploration of the relationship between eye movement parameters and specific cognitive domains would be beneficial. Future studies should consider using sub-domain scales to investigate this. 3. Sample Composition: This study mainly included mild AD patients. To gain a comprehensive understanding of eye movement characteristics at different stages of the disease, future studies should include patients with varying degrees of severity. 4. Eye movement changes can occur in various neurological diseases, and eye tracking has potential applications beyond AD. To validate this possibility, future research should incorporate different samples and include individuals with other neurological conditions.

Conclusion

This study demonstrates the potential of eye tracking as a valuable tool for clinical research on cognitive impairment caused by AD. By utilizing a combination of eye movement paradigms, we were able to gather rich information regarding the abnormal eye movement parameters observed in patients with MCI/AD. Specifically, we found that the antisaccade error correction rate and the total offset degrees (>4°) were more accurate for the early detection of patients with MCI, with an AUC of 0.837, At the same time, we believe that the latency of the prosaccade, the average completion time of overlap prosaccade, the accuracy and error correction rate of the antisaccade are meaningful in monitoring the progress of the disease. There were mild to moderate correlations between eye movement parameters and cognitive scales. Consequently, this study reinforces the argument for utilizing eye movement evaluation systems as an effective diagnostic method for cognitive impairment.

Footnotes

Acknowledgments

We thank the research participants for their contribution to the study.

ORCID iDs

Meichun Tao

Yuanyuan Du

Mei Dong

Author contributions

Meichun Tao (Data curation; Formal analysis; Investigation; Validation; Writing – review & editing); Lei Cui (Data curation; Formal analysis; Investigation); Yuanyuan Du (Data curation; Formal analysis; Investigation); Xiaotang Liu (Data curation; Formal analysis; Investigation); Can Wang (Data curation; Formal analysis; Investigation); Jing Zhao (Data curation; Formal analysis; Investigation); Huimin Qiao (Data curation; Formal analysis; Investigation); Zhenzhong Li (Conceptualization; Methodology; Validation; Writing – review & editing); Mei Dong (Conceptualization; Methodology; Validation; Writing – review & editing).

Declaration of conflicting interests

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

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Data availability

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

References

GBD 2016 Neurology Collaborators. Global, regional, and national burden of neurological disorders, 1990–2016: a systematic analysis for the global burden of disease study 2016. Lancet Neurol 2019; 18: 459–480.

, et al. The performance of saccade tasks correlates with cognitive test scores in elderly population: evidence for the usefulness of oculomotor tests in cognitive assessment. Gerontology 2023; 69: 321–335.

Shaikh

Zee

. Eye movement research in the twenty-first century-A window to the brain, mind, and more. Cerebellum 2018; 17: 252–258.

Kusne

Wolf

Townley

, et al. Visual system manifestations of Alzheimer's disease. Acta Ophthalmol 2017; 95: e668–e676.

Pavisic

Firth

Parsons

, et al. Eyetracking metrics in young onset Alzheimer's disease: A window into cognitive visual functions. Front Neurol 2017; 8: 377.

Lin

Yang

, et al. A detection model of cognitive impairment via the integrated gait and eye movement analysis from a large Chinese community cohort. Alzheimers Dement 2024; 20: 1089–1101.

Tao

Wang

Liu

, et al. Eye tracking metrics to screen and assess cognitive impairment in patients with neurological disorders. Neurol Sci 2020; 41: 1697–1704.

Kaufman

Pratt

Levine

, et al. Executive deficits detected in mild Alzheimer's disease using the antisaccade task. Brain Behav 2012; 2: 15–21.

Kahana Levy

Lavidor

Vakil

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

10.

Crawford

Higham

Mayes

, et al. The role of working memory and attentional disengagement on inhibitory control: effects of aging and Alzheimer's disease. Age (Dordr) 2013; 35: 1637–1650.

11.

Yang

Wang

, et al. Specific saccade deficits in patients with Alzheimer's disease at mild to moderate stage and in patients with amnestic mild cognitive impairment. Age (Dordr) 2013; 35: 1287–1298.

12.

Pierrot-Deseilligny

Rivaud

Gaymard

, et al. Cortical control of saccades. Ann Neurol 1995; 37: 557–567.

13.

Quaia

Lefèvre

Optican

. Model of the control of saccades by superior colliculus and cerebellum. J Neurophysiol 1999; 82: 999–1018.

14.

Noiret

Carvalho

Laurent

, et al. Saccadic eye movements and attentional control in Alzheimer's disease. Arch Clin Neuropsychol 2018; 33: 1–13.

15.

Peltsch

Hemraj

Garcia

, et al. Age-related trends in saccade characteristics among the elderly. Neurobiol Aging 2011; 32: 669–679.

16.

Ishizaki

Meguro

Nara

, et al. Impaired shifting of visuospatial attention in Alzheimer's disease as shown by the covert orienting paradigm: implications for visual construction disability. Behav Neurol 2013; 26: 121–129.

17.

Saslow

. Effects of components of displacement-step stimuli upon latency for saccadic eye movement. J Opt Soc Am 1967; 57: 1024–1029.

18.

Tang

Badcock

Visser

, et al. Training and the attentional blink: limits overcome or expectations raised. Psychon Bull Rev 2014; 21: 406–411.

19.

Vernet

Yang

Gruselle

, et al. Switching between gap and overlap pro-saccades: cost or benefit. Exp Brain Res 2009; 197: 49–58.

20.

Crawford

Higham

. Distinguishing between impairments of working memory and inhibitory control in cases of early dementia. Neuropsychologia 2016; 81: 61–67.

21.

Johnson

Storandt

Morris

, et al. Longitudinal study of the transition from healthy aging to Alzheimer disease. Arch Neurol 2009; 66: 1254–1259.

22.

Müri

Nyffeler

, et al. The role of the human dorsolateral prefrontal cortex in ocular motor behavior. Ann N Y Acad Sci 2005; 1039: 239–251.

23.

Wilcockson

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.

24.

Opwonya

Wang

Jang

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

25.

Chehrehnegar

Shati

Esmaeili

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

26.

Peltsch

Hemraj

Garcia

, et al. Saccade deficits in amnestic mild cognitive impairment resemble mild Alzheimer's disease. Eur J Neurosci 2014; 39: 2000–2013.

27.

Takeda

Terada

Sato

, et al. Wisconsin card sorting test and brain perfusion imaging in early dementia. Dement Geriatr Cogn Disord 2010; 29: 21–27.

28.

Kaufman

Pratt

Levine

, et al. Antisaccades: a probe into the dorsolateral prefrontal cortex in Alzheimer's disease. A critical review. J Alzheimers Dis 2010; 19: 781–793.

29.

Alichniewicz

Brunner

Klünemann

, et al. Neural correlates of saccadic inhibition in healthy elderly and patients with amnestic mild cognitive impairment. Front Psychol 2013; 4: 467.

30.

Krauzlis

. Recasting the smooth pursuit eye movement system. J Neurophysiol 2004; 91: 591–603.

31.

Büttner

Kremmyda

. Smooth pursuit eye movements and optokinetic nystagmus. Dev Ophthalmol 2007; 40: 76–89.

32.

Moser

Kömpf

Olschinka

. Eye movement dysfunction in dementia of the Alzheimer type. Dementia 1995; 6: 264–268.

33.

Zaccara

Gangemi

Muscas

, et al. Smooth-pursuit eye movements: alterations in Alzheimer's disease. J Neurol Sci 1992; 112: 81–89.

34.

Abadi

Gowen

. Characteristics of saccadic intrusions. Vision Res 2004; 44: 2675–2690.

35.

Okonkwo

Wadley

Ball

, et al. Dissociations in visual attention deficits among persons with mild cognitive impairment. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn 2008; 15: 492–505.

36.

Opwonya

Lee

, et al. Eye movement changes as an indicator of mild cognitive impairment. Front Neurosci 2023; 17: 1243766.

37.

Molitor

Ally

. Eye movements in Alzheimer's disease. J Alzheimers Dis 2015; 44: 1–12.

38.

Perry

Hodges

. Attention and executive deficits in Alzheimer's disease. A critical review. Brain 1999; 122: 383–404.