Novel Method for Rapid Assessment of Cognitive Impairment Using High-Performance Eye-Tracking Technology

Abstract

A rapid increase in the number of patients with dementia has emerged as a global health challenge. Accumulating evidence suggests that early diagnosis and timely intervention can delay cognitive decline. The diagnosis of dementia is commonly performed using neuropsychological tests, such as the Mini-Mental State Examination (MMSE), administered by trained examiners. While these traditional neuropsychological tests are valid and reliable, they are neither simple nor sufficiently short as routine screening tools for dementia. Here, we developed a brief cognitive assessment utilizing an eye-tracking technology. The subject views a series of short (178 s) task movies and pictures displayed on a monitor while their gaze points are recorded by the eye-tracking device, and the cognitive scores are determined from the gaze plots data. The cognitive scores were measured by both an eye tracking-based assessment and neuropsychological tests in 80 participants, including 27 cognitively healthy controls (HC), 26 patients with mild cognitive impairment (MCI), and 27 patients with dementia. The eye tracking-based cognitive scores correlated well with the scores from the neuropsychological tests, and they showed a good diagnostic performance in detecting patients with MCI and dementia. Rapid cognitive assessment using eye-tracking technology can enable quantitative scoring and the sensitive detection of cognitive impairment.

Figure 1

Rapid cognitive assessment using an eye-tracking system and tasks. (A) The eye tracking system used in the study. The gaze point of the subject was recorded using infrared light sources and cameras located below the monitor. (B) A series of ten task movies and pictures (178 s in total) is displayed on the monitor, and the subject views them. Four representative task movies and pictures are shown. Each task assesses eye movement, deductive reasoning (odd-one-out task), visuospatial function, and working memory (from left to right). Images (gold coin, fishes, and kettle) are obtained from JVC KENWOOD Corporation with permission. (C) An example of a working memory task and representative gaze plots with a duration-based heatmap obtained from a control subject. Gaze plots represent the location and time spent looking at the objects. A cue object (double pentagon) is presented for 10 s (encoding), followed by three distinct objects with the same one as the cue object (right bottom on the monitor, target). The subject is asked to remember and gaze at the target object. Fixation duration within the region of interest (ROI) set on the target object was used as a measure of the cognitive score. Full details of the procedure are described in the Supplemental Information.

Figure 2

Correlation between MMSE scores and cognitive function scores, as assessed by the eye-tracking system. (A) Cognitive scores assessed by eye tracking system showed a strong positive correlation with the MMSE scores. p < 0.00001, Spearman’s rank test, n = 80. (B) Subjects were divided into three MMSE categories based on the severity; Low, 0–20/30 (severely impaired, n = 24); Middle, 21–26/30 (mildly impaired, n = 22); High, 27–30/30 (no apparent impairment, n = 34). Subjects in Low and Middle MMSE score categories showed lower cognitive scores assessed by eye tracking system. **p < 0.01, ANOVA followed by the Tukey–Kramer multiple comparisons test. Error bars represent standard errors.

Figure 3

Diagnostic performance of the eye tracking-based cognitive assessment. (A) Cognitive scores assessed by the eye tracking system in clinical subgroups: HC (n = 27), MCI (n = 26), and dementia (n = 27). *p < 0.05, **p < 0.01, ANOVA followed by the Tukey–Kramer multiple comparisons test. Error bars represent SEM. (B,C) ROC curve analysis of diagnostic performance of the eye tracking-based cognitive assessment (blue) and the MMSE (red) for discriminating MCI from HC (B) and any type of cognitive impairment [MCI + dementia] from HC (C). The area under the ROC curve (AUC) analysis was used to compare the diagnostic performance of the eye-tracking cognitive test and the MMSE. HC, Healthy controls; MCI, mild cognitive impairment; ROC, receiver operating characteristic.

Figure 4

The cognitive scores assessed by the eye tracking system correlates with ADAS-Cog, FAB, and CDR scores. (A) The correlation between ADAS-Cog scores and the cognitive scores assessed by the eye tracking system. p < 0.0005, Spearman’s rank test, n = 34. (B) The correlation between FAB scores and the cognitive scores assessed by the eye-tracking system. p < 0.0005, Spearman’s rank test, n = 41. (C) The correlation between CDR scores and the cognitive scores assessed by the eye tracking system. **p < 0.01, ANOVA followed by the Tukey–Kramer multiple comparisons test, n = 43. Error bars represent SEM. ADAS-Cog, Alzheimer’s Disease Assessment Scale-cognitive subscale; FAB, Frontal Assessment Battery; CDR, Clinical Dementia Rating.