← All work
Academic project · IIT Gandhinagar · M.Sc. thesis, 2019

Can the way we look at happy and sad images help screen for depression?

My master’s thesis in cognitive science. I tested a computer-based system that records where people look and how their pulse changes while they view emotional images, and compared people diagnosed with depression with healthy controls.

My roleM.Sc. researcher, thesis
Timeline2018 – 2019
LabIRACS Lab, IIT Gandhinagar
MethodsGaze tracking, pulse sensing, BDI-II, PHQ-9, HADS
3 of 3
Measures that differed between groups: reaction time, fixation, pulse
p < .05
For all three differences
4
Image categories: happy and sad faces, happy and sad situations
01

The questionCould gaze add to a questionnaire?

Depression screening usually relies on self-report questionnaires. They are useful, but they depend on what people are willing or able to say. Research suggests people with depression look at emotional information differently, so gaze might add an objective signal.

The hypothesis

People with depression show different eye movements to emotional images, and a computer-based test can pick those differences up.

02

What I didBuilt the study around a gaze-tracking setup

1Designed the stimuliFour categories of emotional images. We built our own set, because existing image databases were western-oriented.
2Ran the sessionsParticipants viewed images side by side around a central fixation point, while a camera tracked gaze and a finger sensor recorded pulse.
3Added questionnairesBDI-II, PHQ-9 and HADS, to compare gaze with standard pen-and-paper screening.
4Analysed the signalsReaction time, fixation duration and pulse rate, compared between groups in MATLAB and SPSS.
The study setup: chin-rest, camera and two monitors
The setup: chin-rest, camera and monitor.
Calibration screen showing two eyes on a green background
Calibration screen, before each session.
03

What I foundSmall sample, clear signals

Seven people diagnosed with depression took part, alongside a larger group of healthy controls.

Slower responses

Participants with depression took longer to respond to the images (p < .01).

Less stable gaze

Their fixations were shorter, suggesting poorer fixation stability (p < .05).

Higher pulse rate

Pulse rate during the task differed between the groups (p < .05).

Feasible as a screen

Combining gaze with a questionnaire was practical to run and gave a fuller picture than either alone.

What it taught me

Behaviour can say things people don’t. Pairing what someone reports with what they do is a habit I still bring to every study, now with users instead of patients.

Experimental designEye trackingPhysiological measuresStandardised questionnairesStatistical analysisSensitive research ethics

Academic thesis, supervised by Dr. Uttama Lahiri. Participant details are not shown, and the emotional images used in the study are left out because they show real people.