Perceptual-phase eye-tracking during figure copying is associated with amyloid-β PET positivity in mild cognitive impairment

Abstract

Background

Amyloid-β (Aβ) pathology is a core feature of Alzheimer's disease (AD) and may be present in mild cognitive impairment (MCI). Accessible, non-invasive markers of early Aβ-related cognitive change are needed.

Objective

To investigate whether perceptual-phase eye-tracking metrics during figure copying are associated with Aβ-PET positivity in MCI.

Methods

In this cross-sectional study, 59 patients with MCI underwent ¹⁸F-flutemetamol PET and were classified as Aβ-PET positive (n = 34) or Aβ-PET negative (n = 25). Participants completed a simplified Rey-Osterrieth Complex Figure Test while wearing eye-tracking glasses. The display was then divided into perceptual and working areas of interest (AOIs), and metrics were extracted for each AOI. A matched-pair sensitivity analysis (n = 28) was performed to reduce differences in age, education, and Mini-Mental State Examination scores. Binary logistic regression was used to examine the association between Aβ-PET positivity and fixation duration in the perceptual AOI.

Results

The Aβ-PET positive group showed significantly longer fixation duration in the perceptual AOI (p = 0.027, r = 0.296), whereas no significant group differences were observed in the working AOI. Exploratory analyses showed higher fixation count, visit duration, visit count, and saccade count in the perceptual AOI of the Aβ-PET-positive group. Longer fixation duration in the perceptual AOI was independently associated with Aβ positivity (OR per second = 1.18, 95% CI 1.03–1.36, p = 0.021).

Conclusions

Aβ-PET positivity in MCI was associated with greater perceptual-phase gaze inefficiency during figure copying. Perceptual-phase eye-tracking metrics may serve as potential markers of amyloid-related visuospatial alteration in MCI.

Keywords

Alzheimer's disease amyloid-β peptides eye-tracking technology mild cognitive impairment positron emission tomography

Introduction

Visuospatial function, one of the cognitive domains that progressively deteriorate in Alzheimer's disease (AD), is defined as the ability to perceive, integrate, and interpret spatial relationships and visual forms. Importantly, this function often declines early during the transition from mild cognitive impairment (MCI) to AD.¹ Longitudinal evidence further indicates that visuospatial deficits become evident approximately three years prior to the clinical diagnosis of AD, highlighting their potential as a sensitive early indicator of underlying AD pathology, even preceding substantial memory impairments.² Moreover, previous studies have shown that visuospatial deficits are not only characteristic of early AD-related cognitive decline but also closely linked to amyloid-β (Aβ) pathology.^3–6 Despite this, visuospatial function remains relatively under-assessed compared with other cognitive domains, likely because conventional tests rely primarily on paper-and-pencil methods.

Compared to conventional paper-and-pencil visuospatial tests, eye-tracking has gained attention as a potential tool for assessing visuospatial dysfunction in AD.^7,8 Recent studies have demonstrated that eye-tracking can differentiate individuals with MCI or AD from healthy older adults by analyzing visuospatial processing during daily activities,^9–11 evaluating visual search performance,¹² and characterizing gaze patterns elicited by short movies and static images.¹³ Moreover, eye-tracking metrics such as fixation duration, saccadic movements, and smooth-pursuit accuracy have been correlated with established cognitive measures, including Mini-Mental State Examination (MMSE) scores,¹³ standard visuospatial test scores,¹⁴ and cortical thickness¹⁵ along the AD continuum. However, whether biological markers such as Aβ positivity are associated with eye-tracking metrics at comparable levels of cognitive impairment has not been fully established.

Given that eye-tracking metrics can capture visuospatial dysfunction, examining their association with Aβ pathology may provide insight into the mechanisms underlying visuospatial impairment. In our previous study integrating an eye-tracking approach with a simplified version of the Rey-Osterrieth Complex Figure Test (sRCFT),^16,17 Aβ-positive AD patients demonstrated longer fixation durations and increased number of fixations, saccades, and gaze switches compared to healthy controls, suggesting higher cognitive effort during complex visuospatial tasks. In addition, an eye-tracking study in patients with AD reported medium-to-large effects in visual attention measures.¹⁸ Building on these findings, we hypothesized that Aβ-positive individuals with MCI would require greater cognitive effort to compensate for visuospatial deficits than Aβ-negative individuals with MCI, as reflected primarily by longer fixation duration in the perceptual area of interest (AOI).

Therefore, we investigated whether Aβ-positive and Aβ-negative individuals with MCI differ in phase-specific eye-tracking metrics during the sRCFT, with a particular focus on the perceptual phase as a potential locus of early visuospatial alteration.

Methods

Study design

This cross-sectional study investigated the associations between eye-tracking metrics and Aβ-PET positivity in individuals with MCI at the Memory Disorder Clinic, Jeonbuk National University Hospital, between July 2021 and February 2024. During the study period, consecutive patients attending the Memory Disorder Clinic were screened for eligibility. Of these, 66 individuals who fulfilled the modified Petersen criteria for MCI and met the study inclusion and exclusion criteria were enrolled in the study. A priori power analysis using G*Power 3.1.9.7 indicated that at least 25 participants per group were required to achieve 80% power at α = 0.05, assuming a large effect size (d = 0.80). This assumption was based on a previous eye-tracking study in patients with AD that reported medium-to-large effects in visual attention measures.¹⁸

Participant recruitment

Participants with MCI met the modified criteria proposed by Petersen and colleagues.¹⁹ The inclusion criteria were: (1) subjective cognitive complaint by the patient or caregiver, (2) objective cognitive impairment below the 16^th percentile of education- and age-matched norms in at least one of four domains (memory, visuospatial, language, or frontal-executive function) on the Seoul Neuropsychological Screening Battery,^20,21 (3) normal daily functioning based on clinical judgment and the Seoul-Instrumental Activities of Daily Living (S-IADL) scale,²² (4) no dementia, and (5) aged 60–89 years. Exclusion criteria were: (1) structural brain lesions on MRI (e.g., infarction, hemorrhage, tumor, hydrocephalus, and severe WMH defined as a Fazekas score of 3) or other neurological disorders that could affect cognitive function, (2) abnormal laboratory results, (3) moderate or severe vision loss (visual acuity < 0.3), (4) MMSE score < 19, and (5) education below the 6th grade.

All participants underwent diagnostic assessments including clinical interview, neuropsychological testing, MRI, and PET. Laboratory tests included a complete blood count, blood chemistry, thyroid function tests, vitamin B12 and folate levels, and syphilis serology. Aβ positivity was determined using 18F-flutemetamol PET based on a visual assessment,²³ and the final classification was established by multidisciplinary consensus among neurologists, neuropsychologists, radiologists, and nuclear medicine physicians.

All procedures were approved by the Jeonbuk National University Hospital Institutional Review Board (CUH2020-12-058), and written informed consent was obtained from all participants.

Neuropsychological assessments

All participants completed the standardized Seoul Neuropsychological Screening Battery,^20,21 which comprises multiple cognitive assessments. Attention and working memory were evaluated using the forward and backward digit span tests. Language abilities were assessed with the Korean version of the Boston Naming Test (K-BNT),²⁴ while calculation skills were measured through three problems each in addition, subtraction, multiplication, and division. Visuospatial function was examined using the Rey-Osterrieth Complex Figure Test (RCFT).²⁵ Memory function was assessed through immediate and delayed recall tasks from both the Seoul Verbal Learning Test (SVLT)²⁶ and RCFT. Frontal-executive function was evaluated using the phonemic and semantic tasks of the Controlled Oral Word Association Test (COWAT)²⁷ and the Stroop word/color reading test.²⁸ Additionally, general cognitive status and depressive symptoms were measured using the MMSE, Clinical Dementia Rating (CDR) with CDR sum of boxes, the Geriatric Depression Scale (GDS), and the Seoul-Instrumental Activities of Daily Living (S-IADL) scale to measure instrumental daily functioning.²²

Experimental apparatus

We utilized the Tobii Pro Glasses 2 (Tobii Pro, Stockholm, Sweden), a wearable eye-tracking device designed for real-world research applications. This system records binocular eye movements at a sampling rate of 50 Hz and features a scene camera with a video resolution of 1920 × 1080 pixels at 25 frames per second (fps) in H.264 format. The 50 Hz sampling rate provides sufficient resolution for fixation-based analyses; however, it does not allow for fine-grained characterization of saccadic activity. Weighing approximately 45 grams, the glasses are lightweight and non-intrusive, allowing participants to engage naturally in tasks. For the figure-copying task, we employed a 12-inch digital tablet (Samsung Galaxy Book 12, Samsung Electronics, Suwon, Korea) with a resolution of 2160 × 1440 pixels, paired with a stylus (nib diameter = 0.7 mm, pressure sensitivity = 4096 levels, Samsung Electronics, Suwon, Republic of Korea) to accurately capture participants’ drawing performance.

Experimental design

Experimental setup and eye-tracker calibration

To minimize interference from direct sunlight and enhance the quality of eye-tracking data collection, our experiment was conducted in a traditional neuropsychological testing environment with soft, evenly distributed lighting. The eye-tracker recording unit was connected to a laptop equipped with Tobii Pro Glasses Controller software. The eye-tracking glasses were then securely fitted onto the participant, with careful adjustments made to ensure that the four integrated cameras could capture clear and accurate eye movement data. For calibration, a white calibration card (7 cm × 10 cm) with a black ring and a central black dot was positioned approximately 80 cm in front of the participant's eyes. Participants were instructed to focus on the black dot until the Tobii Pro Glasses 2 Controller confirmed successful calibration.

Simplified Rey-Osterrieth complex figure test

In our previous study, we developed a simplified version of the original Rey-Osterrieth Complex Figure Test and established its validity by comparing it with the original RCFT.^16,17 The sRCFT consists of four global components (a large rectangle and diagonal, horizontal, and vertical crosses) and four local components (a square, double circles, three triangles, and four arrows). Scoring was based on both accuracy and placement, following the standardized scoring criteria proposed by Meyers. The raw scores ranged from 0.0 to 16.0.¹⁶

In this study, participants copied the sRCFT while wearing eye-tracking glasses. The display screen was divided into two regions: the upper half was designated as the “perceptual space,” where the sample figure of the sRCFT was presented, and the lower half was defined as the “working space,” where participants reproduced the figure (Figure 1). Based on our previous studies, the perceptual space is considered to primarily reflect visuoperceptual processing, whereas the working space is regarded as indicative of visuoconstructional processes.¹⁷

Figure 1.

Representative gaze plots of an Aβ-PET positive (A) and an Aβ-PET negative (B) MCI participant during the sRCFT copying task.

Before the task began, participants were given the following instructions:

“When the task starts, a sample figure will appear in the upper half of the tablet. Please copy the figure using this stylus, utilizing the empty space below as much as possible. The stylus does not have an eraser, and you will not be able to modify your drawing. If you make a mistake, forget about it and keep going. Take your time and let me know when you have finished.” The completion time for the sRCFT was defined as the interval between the end of the verbal instruction and the moment the participant verbally indicated that they had finished the drawing.

Eye-tracking data acquisition and preprocessing

Eye-tracking data were processed using Tobii Pro Lab software (version 1.181; Tobii AB, Stockholm, Sweden) for semantic gaze mapping, which aligned gaze coordinates with scene video recordings. Participants’ sRCFT snapshots were imported, and gaze data were automatically mapped and manually verified by the same experimenter for accuracy. Fixation, visit, and saccade metrics were extracted from two predefined areas of interest (AOIs): the perceptual and working spaces (Figure 1).

For quality control, recordings with > 10% tracking loss (fraction of samples with missing x–y coordinates) or metrics exceeding ±3 SD from the group mean were excluded. Before analysis, fixation duration in the perceptual AOI was designated as the primary endpoint based on our prior work¹⁷ and the mechanistic focus of the present study on visuoperceptual processing, while all other metrics were considered exploratory.

Fixations, visits, and saccades were identified using the I-VT (Velocity-Threshold Identification) algorithm: fixations as gaze points with velocities below 30°/s, saccades as sequences exceeding this threshold, and visits as the time interval from the first to the last fixation within an AOI. Two fixation metrics (fixation duration and fixation count), two visit metrics (visit duration and visit count), and one saccade metric (saccade count) were extracted for each AOI.

For exploratory clinical interpretation, fixation duration in the perceptual AOI was additionally categorized into tertiles (low, middle, and high), and Aβ-PET positivity and selected neuropsychological measures were compared across tertile-defined subgroups.

Statistical analysis

Group differences in clinical characteristics, cognitive profiles, and eye-tracking metrics were analyzed using independent-samples t-test or Mann–Whitney U test for continuous variables and Chi-square test for categorical variables. To control for potential confounding effects of age, education, and MMSE score on eye-tracking metrics, a matched subgroup analysis (n = 28) was performed using paired t-tests or Wilcoxon signed-rank tests for continuous variables in the matched-pair analysis, and McNemar tests for categorical variables as appropriate. For the sensitivity analysis, Aβ-PET positive and Aβ-PET negative participants were matched 1:1 within ±2 years for age, within ±2 years for education, and within ±1 point on the MMSE score, with additional consideration of sex balance whenever possible. Matching was performed before comparing eye-tracking outcomes. Because of the limited sample size and the difficulty of identifying closely comparable pairs across multiple variables, only 14 matched pairs could be constructed from the full analytic sample.

To examine the association between eye-tracking metrics and Aβ-PET positivity, a logistic regression was performed with Aβ-PET positivity (0 = negative, 1 = positive) as the dependent variable and fixation duration in the perceptual AOI as the primary predictor. In the primary model, age, sex, education, MMSE score, and task-completion time were entered as covariates based on their clinical relevance and potential confounding effects. To assess the robustness of the findings, an additional sensitivity logistic regression analysis was performed, including APOE ε4 carrier status, SVLT delayed recall, and CDR sum of boxes, which significantly differed between groups and represented key disease-related domains. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Model fit was evaluated using the Hosmer–Lemeshow test and Nagelkerke R², and multicollinearity was assessed via variance inflation factors (< 5).

For exploratory clinical interpretation, participants were additionally categorized into tertiles according to fixation duration in the perceptual AOI. Trends in Aβ-PET positivity across tertiles were evaluated using Pearson chi-square test and linear-by-linear association test, whereas continuous neuropsychological measures were compared using one-way analysis of variance or the Kruskal-Wallis test, as appropriate.

All statistical analyses were performed using IBM SPSS version 29.0 (IBM Corp., Armonk, NY). Radar plots illustrating group differences in key eye-tracking parameters were generated using OriginPro, version 2025b (OriginLab Corporation, Northampton, MA, USA). Statistical significance was set at two-tailed p < 0.05.

Results

Clinical characteristics and cognitive performance

Of the 66 individuals initially screened, seven were excluded from the final statistical analysis. Three participants were excluded due to inadequate eye-tracking calibration caused by narrow palpebral fissures, two were excluded because of excessive tracking loss (>10%) based on quality-control criteria, and two withdrew from the study for personal reasons. The final analytic sample consisted of 59 participants, including 34 Aβ-positive and 25 Aβ-negative individuals.

No significant group differences were observed in age, sex distribution, and years of education. The prevalence of APOE ε4 carriers was significantly higher in the Aβ-positive MCI individuals (88.24%) than in the Aβ-negative MCI individuals (32%) (p < 0.001). In cognitive assessments, the Aβ-positive MCI individuals showed poorer performance on memory-related tests, including the Seoul Verbal Learning Test (SVLT) immediate recall (p = 0.01), delayed recall (p = 0.005), and recognition (p = 0.003), as well as RCFT immediate recall (p < 0.001), delayed recall (p < 0.001), and recognition (p = 0.002) scores. They also scored lower on the COWAT supermarket category (p = 0.004). Moreover, the MMSE scores were lower (p < 0.001), whereas CDR sum of boxes (p < 0.001) and S-IADL (p < 0.001) scores were higher in the Aβ-positive MCI individuals. Scores on tests of attention, language, and visuospatial function, including the sRCFT copy score, did not differ between the groups. Demographic and cognitive characteristics are summarized in Table 1.

Table 1.

Clinical and cognitive characteristics in all participants.

	Amyloid-β positive (n = 34)	Amyloid-β negative (n = 25)	p
Age	71.8 ± 4.82	68.9 ± 6.53	0.073
Sex F:M	21:13	13:12	0.453
Education (IQR)	9 (6, 12)	9 (7, 14)	0.249
APOE ε4 carrier	30/34 (88.24%)	8/25 (32%)	<0.001
Attention
Forward digit span	5.55 ± 1.34	6.04 ± 1.76	0.242
Backward digit span (IQR)	3 (3, 4)	4 (3, 4)	0.449
Language
K-BNT	41.39 ± 7.02	45.46 ± 8.37	0.055
Calculation (IQR)	11 (9.5, 12)	11 (10, 12)	0.694
Visuospatial function
RCFT: copying (IQR)	32 (28.25, 34)	32 (28.25, 34)	0.858
sRCFT: copying (IQR)	11 (10, 13)	12 (11, 15)	0.693
Memory
SVLT: immediate recall	15.32 ± 4.04	18.63 ± 5.07	0.010
SVLT: delayed recall (IQR)	1 (0, 3)	3 (1.75, 6.25)	0.005
SVLT: recognition	18.13 ± 2.97	20.5 ± 2.55	0.003
RCFT: immediate recall (IQR)	3.5 (0.5, 5.5)	9 (5, 14.5)	<0.001
RCFT: delayed recall (IQR)	1.5 (0, 5.5)	8.5 (5.38, 15)	<0.001
RCFT: recognition (IQR)	17 (16, 19)	19 (18, 20)	0.002
Frontal/executive functions
COWAT animal	12.03 ± 4.24	14.42 ± 4.92	0.059
COWAT supermarket	12.65 ± 5.01	17.33 ± 6.48	0.004
COWAT phonemic	18.67 ± 8.24	21.29 ± 12.31	0.376
Stroop test: color	65.19 ± 27.32	74.04 ± 23.48	0.211
MMSE (IQR)	25 (23.25, 26)	28 (26.75, 29)	<0.001
CDR (IQR)	0.5 (0.5, 0.5)	0.5 (0.5, 0.5)	0.401
CDR sum of boxes (IQR)	2 (1.5, 2.5)	1 (1, 1.5)	<0.001
GDS	5.03 ± 3.45	6.5 ± 3.79	0.130
SIADL (IQR)	3 (2, 5)	1.5 (0.75, 3)	<0.001

Continuous variables are presented as mean ± SD or median (IQR), as appropriate. IQR: interquartile range; K-BNT: Korean Boston Naming Test; RCFT: Rey-Osterrieth Complex Figure Test; SVLT: Seoul Verbal Learning Test; COWAT: Controlled Oral Word Association Test; MMSE: Mini-Mental State Examination; CDR: Clinical Dementia Rating; GDS: Geriatric Depression Scale; SIADL: Seoul Instrumental Activities of Daily Living.

To control for potential confounding effects of age, education, and general cognition (MMSE), we conducted a sensitivity analysis using a matched-pair design, yielding 14 matched pairs (14 Aβ-positive and 14 Aβ-negative). There were no significant differences in global cognitive or functional measures, but significant differences remained in RCFT immediate recall (p = 0.049) and RCFT delayed recall (p = 0.024) scores (Table 2).

Table 2.

Clinical and cognitive characteristics in the matched-pair analysis.

	Amyloid-β positive(n = 14)	Amyloid-β negative(n = 14)	p
Age (IQR)	69 (67.25, 71)	70.5 (64, 75.5)	0.645
Sex (F:M)	8:6	10:4	0.695
Education (IQR)	12 (9.75, 12)	9 (7.5, 12)	0.289
APOE ε4 carrier	11/14 (78.57%)	6/14 (42.86%)	0.336
Attention
Forward digit span	6.07 ± 1.54	5.79 ± 1.58	0.632
Backward digit span (IQR)	3 (3, 4)	3.5 (3, 4)	0.982
Language
K-BNT	42.93 ± 7.89	42.71 ± 8.26	0.945
Calculation (IQR)	12 (10.25, 12)	11 (10.25, 12)	0.541
Visuospatial function
RCFT: copying (IQR)	32 (29, 32.75)	30.5 (27.75, 33.75)	0.839
sRCFT: copying (IQR)	11 (10.25, 12.75)	11 (10.63, 12)	0.734
Memory
SVLT: immediate recall (IQR)	15.5 (13.5, 18)	17 (13, 18.75)	0.804
SVLT: delayed recall	2.36 ± 1.91	2.79 ± 2.12	0.578
SVLT: recognition	18.07 ± 3.17	19.86 ± 2.35	0.103
RCFT: immediate recall (IQR)	4 (1, 7.5)	7.75 (6.13, 12.38)	0.049
RCFT: delayed recall	4.77 ± 4.39	9.64 ± 6.84	0.024
RCFT: recognition (IQR)	18 (16, 19.75)	19 (18.25, 20)	0.265
Frontal/executive functions
COWAT animal	13.36 ± 3.00	13 ± 3.80	0.785
COWAT supermarket	14.71 ± 3.58	17.07 ± 5.27	0.178
COWAT phonemic (IQR)	21.5 (17.5, 24.5)	18 (14.5, 31)	0.635
Stroop test: color	78.64 ± 27.09	69.43 ± 20.52	0.32
MMSE (IQR)	26 (26, 27)	27 (25, 27)	0.982
CDR (IQR)	0.5 (0.5, 0.5)	0.5 (0.5, 0.5)	0.999
CDR sum of boxes (IQR)	1.5 (1.5, 1.5)	1.5 (1, 1.5)	0.062
GDS	4.64 ± 2.24	5.86 ± 3.86	0.318
SIADL (IQR)	2 (1, 3)	2 (1, 3.75)	0.376

Continuous variables are presented as mean ± SD or median (IQR), as appropriate. IQR: interquartile range; K-BNT: Korean Boston Naming Test; RCFT: Rey-Osterrieth Complex Figure Test; SVLT: Seoul Verbal Learning Test; COWAT: Controlled Oral Word Association Test; MMSE: Mini-Mental State Examination; CDR: Clinical Dementia Rating; GDS: Geriatric Depression Scale (short form); SIADL: Seoul Instrumental Activities of Daily Living.

Eye-tracking metrics

Eye-tracking data were analyzed for the total, perceptual, and working AOIs. Metrics exceeding ±3 SD from the group mean were identified as outliers and excluded (1.03% of all data). No significant differences were observed between Aβ-positive and Aβ-negative MCI individuals in task completion time or in any of the eye-tracking metrics in the total AOI, including total fixation duration, total fixation count, total visit duration, total visit count, and saccade count.

When examining AOI-specific patterns, Aβ-positive MCI individuals showed significantly longer fixation duration in the perceptual AOI (p = 0.027, r = 0.296), the primary outcome measure. Exploratory analysis revealed higher fixation count (p = 0.018, r = 0.316), longer visit duration (p = 0.018, r = 0.316), increased visit count (p = 0.034, r = 0.285), and greater saccade count (p = 0.013, r = 0.322) compared with Aβ-negative MCI individuals in the perceptual AOI (Figure 2). In contrast, no significant differences were found in any eye-tracking metrics in the working AOI. The numeric results are summarized in Table 3.

Figure 2.

Group differences in eye-tracking metrics between Aβ-PET positive and Aβ-PET negative MCI individuals. (A–E) Aβ-positive individuals exhibited significantly higher fixation duration, fixation count, visit duration, visit count, and saccade count in the perceptual AOI. (F-G) Radar plots of the five AOI metrics in the perceptual and working AOIs. Values were generally greater in Aβ-PET positive than in Aβ-PET negative individuals with MCI in the perceptual AOI, whereas no significant group differences were observed in the working AOI. Radar plots are shown for visualization only and are based on min–max normalized group-wise median values.

Table 3.

Eye-tracking results during the sRCFT task in all participants.

	Amyloid-β positive (n = 34)	Amyloid-β negative (n = 25)	p	Effect size (r)
Task completion time (sec)	73.68 (57.48, 87.18)	64.38 (52.15, 79.09)	0.357	0.120
Total
Total fixation duration (sec)	49.24 (42.45, 68.55)	39.36 (34.80, 48.97)	0.059	0.252
Total fixation count	159 (122, 202)	140 (111.5, 167)	0.526	0.085
Total visit duration (sec)	60.87 (50.89, 79.63)	50.41 (40.76, 64.46)	0.108	0.215
Total visit count	59 (50.75, 85.5)	56 (41, 65)	0.323	0.133
Total saccade count	57 (50, 77.5)	51 (39, 68)	0.357	0.120
Perceptual AOI
Fixation duration (sec)	14.29 (11.28, 19.68)	9.43 (7.54, 14.87)	0.027	0.296
Fixation count	65.00 (43.00, 86.50)	49.00 (38.00, 58.00)	0.018	0.316
Visit duration (sec)	17.59 (13.34, 22.40)	13.21 (9.67, 16.87)	0.018	0.316
Visit count	29.50 (25.00, 38.75)	25.00 (19.00, 32.00)	0.034	0.285
Saccade count	19.00 (15.00, 32.00)	14.00 (8.00, 19.00)	0.013	0.322
Working AOI
Fixation duration (sec)	38.42 (26.44, 48.14)	29.94 (24.47, 39.68)	0.387	0.116
Fixation count	93.00 (70.00, 141.50)	86.00 (72.00, 117.00)	0.592	0.072
Visit duration (sec)	40.78 (31.38, 56.69)	38.32 (29.86, 48.37)	0.504	0.089
Visit count	33.50 (25.25, 44.75)	31.00 (24.00, 34.00)	0.112	0.215
Saccade count	38.50 (24.25, 51.75)	33.00 (26.00, 52.00)	0.724	0.046

All eye-tracking variables are presented as median (IQR) because of non-normal distribution. sRCFT: simplified Rey-Osterrieth Complex Figure Test; AOI: area of interest.

A sensitivity analysis using a matched-pair design largely supported these findings. Consistent with the main analysis, Aβ-positive MCI individuals also exhibited longer fixation duration (p = 0.018, Hedges’ g = 0.678), higher fixation count (p = 0.014, Hedges’ g = 0.709), longer visit duration (p = 0.013, Hedges’ g = 0.719), increased visit count (p = 0.037, Hedges’ g = 0.586), and greater saccade count (p = 0.009, Hedges’ g = 0.657) in the perceptual AOI No significant differences were observed in any eye-tracking metrics in the working AOI (Supplemental Table 1).

Logistic regression

A logistic regression was conducted to determine whether fixation duration in the perceptual AOI predicted Aβ-PET positivity after adjusting for demographic and cognitive covariates. The model was specified as:

logit (p) = β_{0} + β_{1} F i x D u r_{p e r c} + β_{2} Age + β_{3} Sex + β_{4} Education + β_{5} MMSE + β_{6} T a s k T i m e,

where p is the probability of Aβ-PET positivity and

{FixDur}_{perc}

denotes fixation duration (sec) in the perceptual AOI. All coefficients

(β)

were estimated using the enter method, converted to odds ratios (OR =

e^{β}

) with 95% confidence intervals (CIs), and evaluated for multicollinearity (all variance-inflation factors < 5).

In the primary logistic regression model adjusting for age, sex, education, MMSE, and task completion time, longer fixation duration in the perceptual AOI was independently associated with higher odds of Aβ-PET positivity (β = 0.167; OR = 1.18, 95% CI 1.03–1.36, p = 0.021). Higher MMSE scores were inversely associated with Aβ-PET positivity (β = –0.672; OR = 0.51, p = 0.003), whereas age, sex, education, and task completion time were not significant. The model showed good calibration (Hosmer–Lemeshow χ² = 5.84, p = 0.558) and explained approximately 50% of the variance in Aβ status (Nagelkerke R² = 0.495). The primary logistic regression results are presented in Table 4, and Figure 3 illustrates the association between fixation duration in the perceptual AOI and Aβ-PET positivity.

Figure 3.

Visual relationship between fixation duration in the perceptual AOI and Aβ-PET positivity during the sRCFT copying task.

Table 4.

Primary logistic regression model for Aβ-PET positivity.

Predictor	β	SE	Wald	df	p	OR	95% CI for OR (lower–upper)
Intercept	8.982	7.433	1.431	1	0.232	7274.221
Fixation duration in perceptual AOI (sec)	0.168	0.073	5.329	1	0.021	1.182	1.026–1.363
Age (yrs)	0.099	0.07	1.974	1	0.16	1.104	0.962–1.266
Sex (0 = F, 1 = M)	−0.406	0.796	0.261	1	0.61	0.666	0.140–3.172
Education (years)	0.102	0.115	0.791	1	0.374	1.107	0.885–1.386
MMSE	–0.672	0.223	9.05	1	0.003	0.511	0.330–0.791
Task-completion time (sec)	−0.023	0.018	1.771	1	0.183	0.977	0.944–1.011

β: unstandardized regression coefficient; SE: standard error; Wald: Wald χ² statistic; OR: odds ratio (= Exp[β]); CI: confidence interval; MMSE: Mini-Mental State Examination.

In a sensitivity model that additionally included APOE ε4 carrier status, SVLT delayed recall, and CDR sum of boxes, the association between fixation duration in the perceptual AOI and Aβ-PET positivity remained significant (OR = 1.234, 95% CI 1.039–1.467, p = 0.017). MMSE remained significant in this expanded model, whereas the additional variables were not independently significant (Supplemental Table 2).

Exploratory tertile-based subgroup analysis

For exploratory clinical interpretation, participants were additionally categorized into tertiles according to fixation duration in the perceptual AOI. The proportion of Aβ-PET positive individuals increased significantly across tertiles, from 31.6% in the low tertile to 65.0% in the middle tertile and 75.0% in the high tertile (Pearson χ² = 8.197, p = 0.017; linear-by-linear association p = 0.007; Supplemental Table 3). In contrast, key neuropsychological measures, including MMSE, SVLT delayed recall, CDR sum of boxes, and S-IADL, did not differ significantly across tertiles.

Discussion

In this study, we compared eye-tracking metrics during the sRCFT copy task between Aβ-PET positive and Aβ-negative MCI individuals. In the perceptual AOI, Aβ-PET positive MCI individuals showed longer fixation duration—the primary endpoint—as well as longer visit durations, higher fixation and visit counts, and greater saccade counts, whereas no significant group differences were observed in the working AOI. Logistic regression analysis further demonstrated that longer fixation duration in the perceptual AOI was significantly associated with Aβ-PET positivity, independent of demographic and cognitive covariates. Exploratory tertile-based subgroup analysis further showed that longer fixation duration in the perceptual AOI was associated with a higher proportion of Aβ-PET positivity. Together, these findings suggest that Aβ-related visuospatial alterations at the MCI stage are more pronounced in perceptual processing than in visuoconstruction.

A key finding of this study is that Aβ-PET positive MCI individuals exhibited both longer fixation duration and increased fixation count in the perceptual AOI. Successful performance on the sRCFT requires repeated encoding of complex visual elements through sustained visuoperceptual processing. In this context, a prolonged fixation duration likely reflects a greater allocation of cognitive resources to perceptual encoding, whereas an increased fixation count suggests more frequent revisiting of the same regions, consistent with less efficient visual exploration. These findings are in line with our previous work in AD, in which longer fixations and higher fixation counts were similarly interpreted as reflecting increased perceptual demand and reduced processing efficiency.¹⁷ In that study, working-AOI eye-tracking metrics did not differ significantly between individuals with late-onset AD and cognitively normal controls, whereas significant differences were observed in young-onset AD.¹⁷ Together with the present findings, this suggests that perceptual-AOI alterations may be a more sensitive marker of early amyloid-related change at the MCI stage, whereas working-AOI abnormalities may emerge later or in specific clinical subtypes.

To minimize potential confounding effects of age, education, and global cognition, we conducted a matched-pair sensitivity analysis. Notably, the analysis largely supported the primary findings from the full sample, indicating that the observed alterations in eye-tracking metrics in the perceptual AOI are unlikely to be artifacts of sample heterogeneity. The matched-pair sensitivity analysis further suggested that the group differences in eye-tracking measures were not solely explained by differences in age, education, or global cognitive status.

An important consideration, however, is that despite matching for age, education, and MMSE, the Aβ-positive group demonstrated lower scores for RCFT immediate and delayed recall. This raises the possibility that residual confounding by visuospatial memory cannot be completely ruled out. Impaired visuospatial memory performance may increase fixation demands in the perceptual AOI. Conversely, inefficient perceptual encoding may itself lead to downstream memory deficits. These bidirectional interactions are inherently difficult to disentangle in a cross-sectional design.

Each additional second of fixation corresponded to an approximately 18% increase in the odds of Aβ-PET positivity (OR = 1.18). Importantly, this association remained significant in a sensitivity model that additionally included APOE ε4 carrier status, SVLT delayed recall, and CDR sum of boxes. In contrast to our findings, patients with clinically diagnosed AD have been reported to exhibit shorter fixations during visual target-search paradigms,^12,29 likely reflecting divergent task demands: target-search tasks emphasize selective attention to a single discrete target, whereas the copying task used in our study requires broad, sustained visuoperceptual encoding. These results suggest that individuals with Aβ pathology may allocate perceptual resources to less informative visual elements, contributing to less efficient exploration strategies. In the expanded model, MMSE remained significantly associated with Aβ-PET positivity, consistent with prior literature that links global cognition to Aβ burden.³⁰ The logistic model showed good calibration, and the moderate Nagelkerke R² suggests that fixation duration in the perceptual AOI captures a meaningful, though not exhaustive, proportion of the variance in Aβ status.

Exploratory outcomes revealed longer visit durations, higher fixation and visit counts, and more saccades exclusively in the perceptual AOI. Unlike fixations, which reflect moment-to-moment visual attention, visits reflect sequences of multiple fixations and saccades within a defined area, providing a broader measure of sustained visual engagement. Increased visit and saccade metrics in Aβ-PET positive individuals may therefore reflect disorganized scanning patterns, impaired top-down attentional control,³¹ or a compensatory response to reduced perceptual efficiency. These alterations align with known early dysfunction in frontoparietal and visual association cortices, which are particularly vulnerable to Aβ and tau pathology.³² Collectively, these findings suggest that eye-tracking metrics may provide a sensitive indicator of impaired perceptual encoding associated with Aβ pathology.

The tertile-based subgroup analysis supported the clinical interpretability of the main finding. Higher tertiles of fixation duration in the perceptual AOI were associated with a greater proportion of Aβ-PET positivity, whereas conventional neuropsychological measures did not differ across tertiles. This suggests that prolonged fixation duration in the perceptual AOI may reflect subtle Aβ-related visuoperceptual changes not captured by standard cognitive measures.

Although we sought to reduce potential bias through matched-pair and covariate-adjusted analyses, several limitations remain. The matched-pair analysis was based on only 14 pairs and therefore provides supportive rather than definitive evidence. In addition, residual confounding by visuospatial memory cannot be fully excluded, as lower RCFT memory scores persisted in the Aβ-positive group despite matching for age, education, and MMSE. Aβ positivity was determined by visual assessment and multidisciplinary consensus rather than quantitative centiloid-based thresholds, which may limit external validity and comparability across studies. The 50-Hz eye-tracking system may also have been less sensitive to fast or small-amplitude saccades; however, this is unlikely to have substantially influenced the primary findings, which focused mainly on fixation duration. Furthermore, the Aβ-PET negative MCI group was not further subclassified using biomarker-based or pathology-specific assessments, and some degree of etiologic heterogeneity may therefore have remained within this group. Finally, the single-center Korean design may limit generalizability to other populations.

In summary, Aβ-positive MCI individuals exhibited longer fixation and visit durations, higher fixation and visit counts, and greater saccade counts in the perceptual AOI, reflecting early differences in visuospatial processing associated with Aβ pathology. Although these findings do not establish diagnostic utility, they suggest that perceptual-AOI eye-tracking metrics may serve as a candidate mechanistic marker for future validation and multimodal integration. Larger and longitudinal studies with more rigorous control of baseline cognitive differences are warranted to confirm and extend these observations.

Supplemental Material

sj-docx-1-alz-10.1177_13872877261461540 - Supplemental material for Perceptual-phase eye-tracking during figure copying is associated with amyloid-β PET positivity in mild cognitive impairment

Supplemental material, sj-docx-1-alz-10.1177_13872877261461540 for Perceptual-phase eye-tracking during figure copying is associated with amyloid-β PET positivity in mild cognitive impairment by Qi Wang, Yutong Bai, Byoung-Soo Shin and Ko Woon Kim in Journal of Alzheimer's Disease

Footnotes

Acknowledgements

We express our sincere gratitude to all participants in this study.

ORCID iDs

Qi Wang

Yutong Bai

Byoung-Soo Shin

Ko Woon Kim

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of Jeonbuk National University Hospital (CUH2020-12-058). Written informed consent was obtained from all participants prior to enrollment.

Consent to participate

Not applicable

Consent for publication

Not applicable

Author contribution(s)

Qi Wang: Data curation; Writing – original draft.

Yutong Bai: Formal analysis; Writing – original draft.

Byoung-Soo Shin: Conceptualization; Writing – review & editing.

Ko Woon Kim: Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Resources; Writing – original draft; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by National University Development Project at Jeonbuk National University in 2023.

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

The datasets generated and/or analyzed during the current study will be available from the corresponding author upon reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Berente

Zsuffa

Werber

, et al. Alteration of visuospatial system as an early marker of cognitive decline: a double-center neuroimaging study. Front Aging Neurosci 2022; 14: 854368.

Johnson

Storandt

Morris

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

MacLean

Zwiers

Cox

, et al. Cerebral amyloid angiopathy and visuospatial functioning: analysis of the clock drawing test. Alzheimers Dement 2021; 17: e053014.

Loreto

Gunning

Golemme

, et al. Evaluating cognitive profiles of patients undergoing clinical amyloid-PET imaging. Brain Commun 2021; 3: fcab035.

Jang

Kim

, et al. Linguistic, visuospatial, and kinematic writing characteristics in cognitively impaired patients with beta-amyloid deposition. Front Aging Neurosci 2023; 15: 1217746.

Seo

Kim

Park

, et al. Independent and interactive influences of the APOE genotype and beta-amyloid burden on cognitive function in mild cognitive impairment. J Korean Med Sci 2016; 31: 286–295.

Wolf

Ueda

. Contribution of eye-tracking to study cognitive impairments among clinical populations. Front Psychol 2021; 12: 590986.

Hannonen

Andberg

Kärkkäinen

, et al. Shortening of saccades as a possible easy-to-use biomarker to detect risk of Alzheimer's disease. J Alzheimers Dis 2022; 88: 609–618.

Seligman

Giovannetti

. The potential utility of eye movements in the detection and characterization of everyday functional difficulties in mild cognitive impairment. Neuropsychol Rev 2015; 25: 199–215.

10.

Readman

Polden

Gibbs

, et al. The potential of naturalistic eye movement tasks in the diagnosis of Alzheimer's disease: a review. Brain Sci 2021; 11: 1503.

11.

Akkoyun

Kocoglu

Eraslan Boz

, et al. Visual search for real-world scenes in patients with Alzheimer's disease and amnestic mild cognitive impairment. Brain Behav 2024; 14: e3567.

12.

Eraslan Boz

Koçoğlu

Akkoyun

, et al. Visual search in Alzheimer's disease and amnestic mild cognitive impairment: an eye-tracking study. Alzheimers Dement 2024; 20: 759–768.

13.

Oyama

Takeda

Ito

, et al. Novel method for rapid assessment of cognitive impairment using high-performance eye-tracking technology. Sci Rep 2019; 9: 12932.

14.

Pavisic

Firth

Parsons

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

15.

Shakespeare

Kaski

Yong

, et al. Abnormalities of fixation, saccade and pursuit in posterior cortical atrophy. Brain 2015; 138: 1976–1991.

16.

Kim

Lee

Choi

, et al. A comprehensive evaluation of the process of copying a complex figure in early- and late-onset Alzheimer disease: a quantitative analysis of digital pen data. J Med Internet Res 2020; 22: e18136.

17.

Kim

Choi

Chin

, et al. Eye-tracking metrics for figure-copying processes in early- vs. late-onset Alzheimer's disease. Front Neurol 2022; 13: 844341.

18.

Chau

Chung

Herrmann

, et al. Apathy and attentional biases in Alzheimer's disease. J Alzheimers Dis 2016; 51: 837–846.

19.

Petersen

Lopez

Armstrong

, et al. Practice guideline update summary: mild cognitive impairment: report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology. Neurology 2018; 90: 126–135.

20.

Kang

Park

Lee

, et al. The cortical neuroanatomy related to specific neuropsychological deficits in Alzheimer's continuum. Dement Neurocogn Disord 2019; 18: 77–95.

21.

Kim

Lee

, et al. Validity of the Montreal Cognitive Assessment (MoCA) index scores: a comparison with the cognitive domain scores of the Seoul Neuropsychological Screening Battery (SNSB). Dement Neurocogn Disord 2021; 20: 28–37.

22.

H-M

Kim

J-H

Kwon

E-J

, et al. A study on the reliability and validity of Seoul-Instrumental Activities of Daily Living (S-IADL). J Korean Neuropsychiatr Assoc 2004; 43: 189–199.

23.

Collij

Salvado

Shekari

, et al. Visual assessment of [(18)F]flutemetamol PET images can detect early amyloid pathology and grade its extent. Eur J Nucl Med Mol Imaging 2021; 48: 2169–2182.

24.

Kim

. Normative data on the Korean version of the Boston Naming Test. J Clin Exp Neuropsychol 1999; 21: 127–133.

25.

Cherrier

Mendez

Dave

, et al. Performance on the Rey-Osterrieth Complex Figure Test in Alzheimer disease and vascular dementia. Neuropsychiatry Neuropsychol Behav Neurol 1999; 12: 95–101.

26.

Kwak

Y-T

Cho

D-S

. Usefulness of Seoul verbal learning test in differential diagnosis of Alzheimer's disease and subcortical vascular dementia. J Korean Neurol Assoc 2004; 22: 22–28.

27.

Ross

Calhoun

Cox

, et al. The reliability and validity of qualitative scores for the Controlled Oral Word Association Test. Arch Clin Neuropsychol 2007; 22: 475–488.

28.

Jensen

Rohwer

WD Jr

. The Stroop color-word test: a review. Acta Psychol 1966; 25: 36–93.

29.

Molitor

Ally

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

30.

Farrell

Kennedy

Rodrigue

, et al. Association of longitudinal cognitive decline with amyloid burden in middle-aged and older adults: evidence for a dose-response relationship. JAMA Neurol 2017; 74: 830–838.

31.

Noiret

Carvalho

Laurent

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

32.

Ranasinghe

Cha

Iaccarino

, et al. Neurophysiological signatures in Alzheimer’s disease are distinctly associated with TAU, amyloid-β accumulation, and cognitive decline. Sci Transl Med 2020; 12: eaaz4069.