Eye Tracking for Predicting ADHD

Eye Tracking for Predicting
ADHD
Gavindya Jayawardena
PhD Student
Neuro Information Retrieval and Data Science Lab
Department of Computer Science
Old Dominion University
@Gavindya2, @WebSciDL
Eye Tracker use in Computer Science Research Work

Overview
● Eye Tracking
● Eye Movements
● Eye Tracker Demonstration
● Research Work
○ Predicting ADHD using Eye Gaze Metrics
○ Results
○ Publications
● WEKA Demonstration
2Eye Tracker use in Computer Science Research Work

About Me
● 1st Year PhD Student @ODU
● GPA - 4.0/4.0
● Joined ODU Spring 2019
● Dominion Scholar
● Research Assistant @ODUCS @WSDL
● http://www.cs.odu.edu/~gavindya
● BSc Degree @UOM, Sri Lanka
● GPA - 3.85/4.20
● Originally from Sri Lanka

My Research Area
● Research in Eye tracking for
targeted population
- Current work: Predicting ADHD
using Eye Tracking

What is Eye Tracking?
● Measuring the gaze point (where one is
looking)
Source- https://help.tobii.com/hc/en-us/articles/115003295025
-Eye-tracking-in-gaming-how-does-it-work-
● Measuring the motion of an eye
Source - Eye-Tracking-Krankenkassen-Vergleich-Google-
Search-Heat-Map.jpg

Human Eye

Basic Elements of Eye Movement
• Fixations
• High acuity vision
• Eye is stable in regard to the object of interest
• Saccades
• No vision
• Move eyes between eye fixations very rapidly
• Smooth pursuits
• Eyes follow an object

Why is Eye Tracking Important?
● We move eyes to bring a particular portion of the visible field of view into high
resolution
● We divert our attention to that point so that we can concentrate
● Eye movements follow along the path of attention

Eye Trackers
● A device for measuring eye positions and eye movement.
Tobii 4C
90 Hz
$169.00
PupilLabs
200 Hz
$2740.00
Gazepoint
150 Hz
$1,995.00

Eye Tracking Applications
Gaming

Market ResearchUsability Research
Source - https://imotions.com/blog/top-8-applications-eye-tracking-research/

Medical Research
Source - https://www.tobiipro.com/imagevault/publishedmedia/t2pps4lriuxsx2jv8sn2/TobiiPro-Spectrum-FoU-banners_BiometricEEG-3_1.jpg

Demo of Eye Trackers
13

Tobii 4C Eye Tracker
● Used for - Gaming
● Sampling Rate - 90 Hz
● Price - $169 in Amazon

PupilLabs Core Eye Tracker
● Used for - Research
● Sampling Rate
○ World camera: 120 Hz
○ Eye cameras: 200 Hz
● Price - $2740
Source - https://pupil-labs.com/products/core/configure/

What Do We Do @NIRDS Lab?
● What do we have?
○ Eye Trackers
○ Wearable EEG Caps
○ Empatica E4
● What do we really do?
○ Study Eye Tracking Data of ADHD and Non-ADHD adults
○ Study EEG Signals of Autistic and healthy kids
● What kind of Research?
○ Predict a diagnosis of ADHD using Eye Tracking Data
○ Analyze relationship of brain activities and ASD using EEG
Emotiv EEG Empatica E4

Predicting ADHD using Eye Gaze Metrics Indexing
Working Memory Capacity
In Collaboration with Dr. Anne Michalek from the
Department of Communication Disorders & Special
Education, Old Dominion University
17
Eye Tracking Research

Attention-Deficit/Hyperactivity Disorder (ADHD)
● Three distinct behavioral symptoms
○ Difficulty in paying attention
○ Difficulty in controlling impulsive behaviours
○ Being overactive
● Researchers have recognized that ADHD persists through adulthood, with an
estimation of 10.2% in 2016 in the U.S.

Eye Tracker Specification
● Tobii Pro X2-60 computer
screen-based eye tracker
● Tobii Studio analysis software
● Sampling rate - 60 Hz
(approximately once every 16.23
milliseconds)
● Each participant was calibrated
using Tobii's standard calibration
methods
Tobii X2-60 Eye Tracker
Tobii Studio analysis software

Participants
● Adult participants:
○ 7 participants diagnosed with ADHD by medical practitioners and confirmed through verified
documentation (6 F, 1 M)
○ 7 participants without a diagnosis (4 F, 3 M)
● All of the participants fulfilled the following inclusion criteria:
○ Between 18 and 65 years of age
○ Spoke English as their first language
○ No vision impairments
○ No history of psychotic symptoms
○ No documented learning / reading disabilities

The Experiment
● Purpose
○ To differentiate between ADHD and Non-ADHD
● Hypothesis
○ There is a significant difference in attention level between ADHD and
Non-ADHD subjects
● Measured Using
○ Working Memory Capacity (WMC)

Working Memory Capacity (WMC)
● Hold and manipulate information
simultaneously
● Greater WMC ⇒
■ More attention
■ Less distraction

ADHD and WMC
● Researchers have shown that adults with ADHD have reduced WMC
compared to their peers.
● Differences in WMC is because of differences in attention control
● Differences in WMC affects performance during cognitively demanding tasks
○ Example - In Education
■ Kids, young adults, and adults suffer because of the bad grades
■ Can we use Eye Tracking to identify people in classroom setting?

Paul is afraid of heights and refuses to fly on a plane. ? R
Whenever I drink the newspaper, I always get depressed. ? M
How do we measure WMC?
Reading Span (RSPAN) task
● Read a sentence and tell whether it makes sense (yes/no)
● Remember the letter at the end of the sentence
● Recall all the letters they can remember from sentences
● Based on the performance of RSPAN task, the WMC score is calculated for each
participant

Research Interests
★ Using RSPAN task as the context of study,
Identify ADHD and Non-ADHD
1. Using WMC scores
No significant difference seen in WMC scores of ADHD and Non-ADHD
2. Using Machine Learning on Eye Tracking Data
We created 3 Feature sets which includes saccades and fixations

Comparison of Eye Fixations for ADHD and Non-
ADHD Participant
26Eye Tracker use in Computer Science Predicting ADHD using Eye Gaze Metrics
Screen Capture of Tobii Studio Analysis Software during WMC Task as Generated during the Replay
Mode
ADHD Participant Non-ADHD Participant

Research Interests (contd…)
3. Using Eye Tracking data within Areas Of Interest (AOIs) based on sentences
○ We created a Feature set which includes saccades and fixations within the AOIs
of all the sentences
○ Areas Of Interest (AOIs):
■ AOI 1 - Stimulus (the whole sentence)
■ AOI 2 - Critical word (critical word when determining the coherency of the sentence)
■ AOI 3 - Determiner (the decision point with the letter to be remembered)

Machine Learning and Features
● Machine learning (ML) = Development of computer programs that can
access data and learn patterns or classifications without explicitly
programming
● ML need a dataset which consist of different feature values
● Feature = An attribute being observed
○ Fixation Duration = 100 milliseconds
○ Saccade Duration = 10 milliseconds
● Feature set = A set of all the attributes that you're interested in
○ Eg: Fixation Duration & Saccade Duration

Eye Gaze Metrics Feature Set
Fixation features
● Number of fixations
● Fixation Duration (ms)
● Average Fixation duration (ms)
● Pupil diameter of left eye (mm)
● Pupil diameter of right eye (mm)
Normal Pupils vs. Dilated Pupils
Source - https://www.allaboutvision.com/en-in/conditions/dilated-pupils/

Eye Gaze Metrics Feature Set
Saccade features
● Saccade Duration
● Saccade Amplitude
● Saccade Peak velocities

Sample Eye Gaze Metrics Feature Set
Part of Fixation Feature Set

What do we use for Machine Learning?

Evaluation of Results
● We chose standard information retrieval evaluation measures
○ Precision
○ Recall
○ Accuracy

Results (I)
Classify ADHD and Non-ADHD subjects using Machine Learning on Eye
Tracking data
Feature Set Accuracy
Fixation Features 78.48%
Saccade Features 91.14%
Both Fixation and Saccade Features 91.11%

Results (II)
Classify ADHD and Non-ADHD subjects using Machine Learning on Eye
Tracking data within Areas Of Interest (AOIs) based on sentences
Feature Set Accuracy
Sentence-based Feature set 86.20%

Publications
● 1 Book Chapter
○ Anne M. P. Michalek, *Gavindya Jayawardena, and Sampath Jayarathna. "Predicting ADHD Using Eye Gaze
Metrics Indexing Working Memory Capacity", Computational Models for Biomedical Reasoning and Problem
Solving, IGI Global, pp. 66-88. 2019
● 1 Conference Paper
○ *Gavindya Jayawardena, Anne M. P. Michalek, and Sampath Jayarathna. "Eye Tracking Area of Interest in the
Context of Working Memory Capacity Tasks", In Proceedings of the 2014 IEEE 20th International Conference
on Information Reuse and Integration (IEEE IRI 2019) (In Press)
● 1 Tech report
○ Eye Gaze Metrics and Analysis of AOI for Indexing Working Memory towards Predicting ADHD -
https://arxiv.org/abs/1906.07183
● 1 Blog Post
○ Use of Cognitive Memory to Improve the Accessibility of Digital Collections - https://ws-
dl.blogspot.com/2019/06/2019-06-19-use-of-cognitive-memory-to.html

WEKA
● Is a collection of machine learning
algorithms for data mining tasks
● Contains tools for data classification
● Facilitates to visualize how different
algorithms perform for the same data set

Demo of Weka Knowledge Flow

Questions?
Gavindya Jayawardena
Twitter Handle - @Gavindya2
Email - hjaya001@odu.edu

Thank You!

Eye Tracking for Predicting ADHD

More Related Content

What's hot (20)

Similar to Eye Tracking for Predicting ADHD (20)

Recently uploaded (20)

Eye Tracking for Predicting ADHD

Editor's Notes