Characterizing the Relationship Between Eye Movement Parameters and Cognitive Functions in Non-demented Parkinson's Disease Patients with Eye Tracking

Summary

Here, we present a protocol to study the relationship between the eye movement parameters and cognitive functions in non-demented Parkinson's disease patients. The experiment used an eye tracker to measure the saccadic amplitude and fixation duration in a visual search task. The correlation with performance in multi-domain cognitive tasks was subsequently measured.

Abstract

Cognitive impairment is a common phenomenon in Parkinson’s disease that has implications on the prognosis. A simple, noninvasive and objective proxy measurement of cognitive function in Parkinson’s disease will be helpful in detecting early cognitive decline. As a physiological metric, eye movement parameter is not confounded by the subject's attributes and intelligence and can function as a proxy marker if it correlates with cognitive functions. To this end, this study explored the relationship between the eye movement parameters and performance in cognitive tests in multiple domains. In the experiment, a visual search task with eye tracking was set up, where subjects were asked to look for a number embedded in an array of alphabets scattered randomly on a computer screen. The differentiation between the number and the alphabet is an overlearned task such that the confounding effect of cognitive ability on the eye movement parameters is minimized. The average saccadic amplitude and fixation duration were captured and calculated during the visual search task. The cognitive assessment battery covered domains of frontal-executive functions, attention, verbal and visual memory. It was found that prolonged fixation duration was associated with poorer performance in verbal fluency, visual and verbal memory, allowing further exploration on the use of eye movement parameters as proxy markers for cognitive function in Parkinson’s disease patients. The experimental paradigm has been found to be highly tolerable in our group of Parkinson's disease patients and could be applied transdiagnostically to other disease entities for similar research questions.

Introduction

Parkinson’s disease is classically a motor disorder; yet, the disease is also associated with cognitive deficits, and progression into dementia is common¹. The pathophysiology of cognitive impairment in Parkinson's disease is not well understood. It is thought to be related to alpha-synuclein deposition in the cortical area based on Braak's staging². It was also proposed that a dual syndrome of degeneration of the dopaminergic and the cholinergic system leads to different cognitive deficits with prognostic implication³. More research is needed to further elucidate the exact mechanisms involved in cognitive impairment in Parkinson's disease. On the clinical aspect, the presence of cognitive impairment has a significant impact on prognosis^4,5. Assessment of cognitive function in clinical practice is, therefore, essential. However, a lengthy cognitive assessment is limited by patients’ mental and motor conditions. Therefore, a noninvasive and simple measurement that can reflect the disease's burden on cognitive function is needed.

The eye movement abnormalities are widely described detectable signs of Parkinson's disease from its early stages⁶, yet the pathophysiology is even less well-characterized than that of cognitive impairment. The generation of eye movement is through a transformation of the visual sensory input, subserved by an intertwined cortical and subcortical network, into signals to the oculomotor nuclei in the brainstem for effect⁷. Involvement of Parkinson's disease pathologies in these networks may lead to observable eye movement abnormalities. There is, perhaps overlapping of neuroanatomical structures that govern the control of eye movement and cognitive function. Furthermore, there have been studies examining the relationship between saccadic eye movement and cognitive function in other neurodegenerative disorders⁸. On such grounds, it is worthwhile to explore the use of eye movement parameters as a proxy marker of cognitive functions in Parkinson's disease. One cross-sectional study⁹ showed that reduced saccadic amplitude and longer fixation duration was associated with the severity of global cognitive impairment in Parkinson's disease. However, there is a lack of data on the correlation between eye movement parameters and specific cognitive domains. The significance and need of measurement of specific cognitive domains, rather than a general cognitive state, is that individual cognitive domain informs differential prognostic information in Parkinson's disease³ and they are subserved by different neural networks. The aim of this study is to explore the specific relationship between eye movement metrics and different cognitive functions. This is the first step to establish a foundation on which the development of biomarkers of cognitive decline in Parkinson's disease using eye tracking technology could be built.

The experimental paradigm presented is composed of 2 major parts: the cognitive assessment and the eye tracking task. The cognitive assessment battery encompassed a range of cognitive functions, including attention and working memory, executive function, language, verbal memory and visuospatial function. The choice of these 5 cognitive domains is based on the Movement Disorder Society Task Force Guidelines for the mild cognitive impairment in Parkinson's disease¹⁰, and a set of locally available cognitive tests were selected to build the assessment battery. In a previous similar eye tracking study on Parkinson's disease cognition mentioned⁹, the author extracted the eye movement parameters while the subjects were engaged in visual cognitive tasks, where the parameters may potentially be influenced by the subject's cognitive ability. As this study aimed to assess the correlation between the eye movement parameters and different cognitive domains, the potential confounding effect of cognitive abilities on the eye parameters must be addressed. In this regards, a visual search task, adapted from another eye tracking study on Alzheimer’s disease¹¹, was employed to capture the eye movement parameters of the subjects. During the task, subjects had to search for a single number on a computer screen among multiple alphabet distracters. This task would elicit the alternate use of saccadic eye movement and visual fixation, the abnormalities of which are described widely in Parkinson's disease. The identification and differentiation of number and alphabet is an overlearned task where the demand for cognitive functions is only minimal and would, therefore, be suitable to answer the research question of this study. A computer program was developed based on the specifications and design as stated by Rösler et al.¹¹. in their original study to be run within the in-built software of our eye tracker. An in-house algorithm for classification and analysis of the eye tracking data was also developed for this study.

Protocol

This research project was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee (CREC Ref. No.: 2015.263).

1. Participants Recruitment and Baseline Assessment

1. Recruit Parkinson's disease patients aged less than or equal to 70 from a neurology specialist clinic with the diagnosis made based on the United Kingdom Parkinson's Disease Society (UKPDS) Brain Bank Diagnostic Criteria¹².
   1. Exclude subjects with psychiatric illnesses, ophthalmological diseases that would impair eye movement, or other neurological disorders. Also, exclude cases using anticholinergics as they are known to affect cognitive performance and eye movement.
2. Recruit healthy controls on a 1:1 basis matched by sex, age, and education.
3. Obtain informed consent from the subject.
4. Conduct a clinical diagnostic interview with the subject and, if available, their relatives, to exclude dementia and screen for cognitive impairment with Mini-Mental State Examination (MMSE)¹³ and Montreal Cognitive Assessment (MoCA)¹⁴. Exclude dementia cases from the study or if the subject’s scores of either MMSE or MoCA is <22/30.
5. Assess the visual acuity with a Snellen chart. Exclude the subject if the visual acuity is less than 20/40.
6. Assess motor severity and staging of Parkinson's disease using the Unified Parkinson's Disease Rating Scale (UPDRS) Part II & III¹⁵ and Modified Hoehn and Yahr (H&Y) Staging¹⁶, respectively. Also, obtain information about the current medications taken by the subject.
7. Assess the depressive mood state by the Beck Depression Inventory-II (BDI-II)¹⁷.

2. Experimental Setup

1. Conduct the experiment in a quiet room with an adequate light source.
2. Conduct the experiment for Parkinson's disease subjects when they are on medication with optimal motor function.
3. Prepare the setup that consists of a screen-based eye tracker, a computer, a mouse, a standard keyboard, a chin rest, and cognitive assessment tools (Table of Materials).
4. Use an eye tracker with a sampling rate of at least 300 Hz.
5. Place the chin rest 60 cm in front of the eye tracker screen.

3. The Flow of the Cognitive Assessment and the Visual Search Task

1. Carry out the Chinese Categorical Verbal Fluency Test¹⁸. Instruct the subject to name as many animals as possible in a minute. Record the number of answers and perseverative error. Then repeat the same in the category of fruits and vegetables.
2. Conduct the registration part (Trial 1, 2 and 3) of the Hong Kong List Learning Test (HKLLT)¹⁹ by reading out a pre-defined 16-vocabulary word list and instruct the subject to remember them. Afterwards ask the subject to do free recall of the word list and record the answer (Trial 1).
   1. Repeat step 3.2 twice for Trial 2 and Trial 3.
3. Wait 10 min and 30 min after the registration part of the HKLLT for the 10 and 30 min delay recall.
4. Before the 10 min delayed recall of the HKLLT, perform the Pattern Recognition Memory (PRM) from Cambridge Neuropsychological Test Automated Battery (CANTAB)²⁰ (Table of Material).
   1. Using the tablet computer, present 24 visual patterns, one at a time, at the center of the screen. Instruct the subject to remember the pattern.
   2. After the presentation, in a 2-choice force discrimination paradigm, instruct the subject to choose the pattern that he/she can recognize.
5. Perform the 10 min delay recall of HKLLT by asking the subject to do free recall of the 16-vocabulary word list.
6. Before the 30 min delayed recall of the HKLLT, perform Spatial Span (SSP) from CANTAB²⁰.
   1. Use the tablet computer to show a pattern of white boxes which change in color, one by one, in variable sequences.
   2. Afterwards instruct the subject to touch the boxes in the same sequence they were presented and record the spatial span length that the subject could attain as the difficulty (number of boxes change in color) of the task increases.
7. Carry out the 30 min delay recall by asking the subject to do free recall of the 16-vocabulary word list.
   1. Conduct the recognition and discrimination part of the HKLLT by reading out another pre-defined 32-vocabulary word list, of which half of the vocabularies are from the original word list in 3.2. Instruct the subject to determine whether each vocabulary read out is from the original word list or not.
8. Allow the subject to rest quietly if they finish the tasks in 3.4 and 3.6 before the 10- and 30-min delay recall, respectively.
9. Perform the Stocking of Cambridge (SOC) from CANTAB²⁰.
   1. Using the tablet computer, present 20 scenarios of two parallel displays of 3 balls held in 3 vertical stockings, of which the arrangement of the balls in the displays varies in each scenario.
   2. Instruct the subject to determine, in each scenario, the least number of moves required to rearrange the balls in the lower display in order to copy the pattern shown in the upper display. Record the mean number of choices to correct answer.
10. Perform the Stroop Test²¹.
    1. Give the subject 3 cards consecutively; the first card contains dots printed in different colors, the second card contains Chinese characters printed in different colors while the last card has Chinese characters denoting different colors (e.g., Chinese words of "blue", "yellow", "green", or "red") but printed in a color not denoted by the name (e.g., the word "red" printed in blue ink).
    2. Ask the subject to read out the printed color of the dots/Chinese characters as quickly as possible and record the time required for each card (T1, T2, and T3).
    3. Calculate the interference index with the formula (T3-T1)/T1.
11. Proceed to the visual search task after completing the cognitive tests.
    NOTE: Do not carry out any verbal cognitive task after the registration part of HKLLT until the end of the whole HKLLT (3.7) to prevent interference effect on verbal memory performance.

4. Visual Search Task

1. Position the subject on a chair and place their chin on the chin rest with their forehead against a bar to minimize head movement. Align the subject's eyes to approximately the center of the computer screen. Begin by clicking the Start Recording button in the computer program.
2. Calibration
   1. Calibrate the eye tracker with the in-built calibration program by clicking the Start button in the calibration interface.
   2. Ask the subject to gaze at a red dot moving across the screen with 9 fixation points, while keeping the head still.
   3. Check for the quality of the calibration by viewing the calibration plot (Figure 1). Make sure that the length of the green lines, which represent the error vectors, fall within the grey circles for an acceptable quality of the calibration. Redo the calibration if there is any missing point or the green lines fall outside the grey circles. Click Accept to proceed to the visual search task.
3. Instruction
   1. Provide verbal instruction to the subject and start with 5 practice runs to familiarize the subject with the task.
   2. Instruct the subject to fixate their gaze on the central fixation cross at the beginning of each trial. Then, press Enter on the keyboard to begin a trial, at which the computer screen will display a single number and 79 distracter alphabets scattered randomly (Figure 2).
   3. Instruct the subject to look as quickly as possible for the number and then simultaneously click on the mouse and state the number aloud as soon as the number is located.
   4. Cross check if the number stated is correct or not.
   5. Administer a total of 40 trials after the 5 practice runs.
4. Design of the trial images in the visual search task
   NOTE: The program code, written in PHP, for this section can be found in Supplement File 1.
   1. Use the numbers 4, 6, 7 and 9 exclusively (Supplementary File 1 - Line 5).
      NOTE: The pilot study¹¹ showed that these numbers are most easily discriminated from the alphabets.
   2. Ensure that the location of the target number is randomized from trial to trial with the rule that it could not be in the same visual quadrant for more than three successive trials (Supplementary File 1 - Line 48-52).
   3. Do not use ambiguous alphabets such as "I" and "O" (Supplementary File 1 - Line 76-78).
   4. Set the size of the fixation cross, alphabets, and numbers at 0.85° visual angle (equivalent to around 0.9 cm on a 23 inches computer screen).
      NOTE: Numbers and alphabets are used because these are easily recognizable visual stimuli yet require foveation for identification.
   5. Allow a time lapse of 1.5 s after the investigator has pressed Enter in 4.3.2 and before the display of the central fixation cross is switched to a trial image to begin a trial (Supplementary File 2 - Line 71; 156-158).
   6. Ensure that the screen will go blank with the fixation cross reappearing as the mouse is clicked or after 10 s have elapsed since the beginning of a trial, whichever that is earlier (Supplementary File 2 - Line 72; 162-180).
   7. As the task is finished, generate a .csv file that contains the timestamps of the beginning and the end of each trial (Supplementary File 2 – line 48-59; 199-208). Use this file in the data analysis in section 5.

5. Eye Tracking Data Processing and Analysis

1. In the Replay section of the computer program, check the Samples Percentage of the eyes during the visual search task (Figure 3). Discard the subject’s data if more than 20% missing data are observed.
   NOTE: Samples Percentage denotes the percentage of time the eyes are successfully located by the eye tracker during the visual search task.
2. Click on the Play button for the recording to check the quality of the data by eyeballing the visualized scan path video generated (Figure 4). Discard the whole subject's data if it is grossly erroneous (Figure 5).
3. Discard any trial(s) in which the subject pressed the mouse accidentally and prematurely.
4. In the Data Export section of the program, select GazePointX (ADCSpx) and GazePointY (ADCSpx) and the subject of interest (Figure 6). Click Export Data to export each subject’s data and save as a .csv file. The file contains the x and y coordinates of the subject's eyes position on the computer screen, in pixels, at every time point.
5. Use the Visual Search Analyzer and in the interface (Figure 7), select the data exported in 5.4 as the input of Eye Data and the .csv file generated in 4.4.7 as the input of Action data. Select ST DBScan as the classification algorithm and click on Run. Then, click on Summary to generate a spreadsheet file containing the mean saccade amplitude and the mean fixation duration of the subject.
6. Design of the visual search analyzer
   NOTE: The coding for the design of the analyzer could be found at https://github.com/lab-viso-limited/visual-search-analyzer. Its program code can be found in Supplementary File 3.
   1. Program the analyzer such that it extracts and analyzes only the data from the beginning to the end of the trial (i.e., from the display of the number and alphabets until the mouse is clicked or 10 s have elapsed), using the .csv file generated in 4.4.7 (Supplementary File 3 - Line 6-173).
   2. Program the analyzer such that it fills in the data loss due to eye-blinking by averaging the x and y coordinates of the gaze point immediately before and after the blinking (Supplementary File 3 - Line 176-260).
   3. Program the analyzer such that it classifies the raw data into either saccade or fixation by using the algorithm developed based on ST-DBSCAN²² (Program code in Supplementary File 4). 

Results

The full result of this study is available in the original paper published²³. Parkinson’s disease subjects (n = 67) were recruited and completed the assessment. However, 5 cases failed to complete the visual search task as they wore progressive lens incompatible with the eye tracker and their data was discarded. The mean age of the subjects was 58.9 years (SD = 7.5 years) with a male to female ratio of 1.7:1. 62 healthy age-, sex-, and education-matched controls were recruited for comparison.

...

Discussion

The protocol presented above was designed as the first part of a longitudinal study in exploring the potential clinical utility of eye movement parameters as surrogate markers for cognitive functions in Parkinson's disease. While there are studies that examine more classical eye tracking paradigms such as self-paced saccade, reflexive saccade, and anti-saccade^25,26,27, a visual search task was used in this study to measure e...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Computer	Intel
Computerized cognitive assessment tool	CANTAB	CANTAB Research Suite	Contains Pattern Recognition Memory, Spatial Span, and Stockings of Cambridge
Eye Movement Analyzer	Lab Viso Limited		https://github.com/lab-viso-limited/visual-search-analyzer
Eye tracker	Tobii	Tx300	23 inch computer screen with resolution of 1920 x 1080, Sampling rate at 300 Hz
Hong Kong List Leanrning Test	Department of Psychology, The Chinese University of Hong Kong	The Hong Kong List Learning Test (HKLLT) 2nd Edition
Stroop test	Laboratory of Neuropsychology, The University of Hong Kong	Neuropsychological Measures: Normative Data for Chinese, Second Edition (Revised)
Tobii Studio	Tobii	Tobii Studio version 3.2.2	Computer programme for running the visual search task
Visual Search Task	Lab Viso Limited		https://www.labviso.com/#products