Artificial Intelligence-Based Depression Detection

Area of Science:

• Ophthalmology
• Psychiatry
• Artificial Intelligence

Background:

• Depression poses significant safety risks in critical professions like aviation and driving.
• Current depression detection relies on subjective self-reporting, lacking speed and objectivity.
• Objective biomarkers are needed for early detection to prevent accidents.

Purpose of the Study:

• To develop and evaluate an AI-based system for real-time, objective depression detection.
• To integrate iris identification with pupillometric and eye movement analysis.
• To provide a preventive safety measure for transportation sectors.

Main Methods:

• An AI system combining iris identification and a CNN-LSTM network for depression detection.
• Analysis of pupillometric (pupil dilation) and eye movement (saccadic velocity, fixation time) biomarkers.
• Evaluation using data from 242 participants, comparing responses to neutral, positive, and negative stimuli.

Main Results:

• Iris identification module achieved an Equal Error Rate < 0.5%.
• Depression detection model (CNN-LSTM) showed 89% accuracy and 0.94 AUC.
• Depressed individuals exhibited distinct pupillometric and eye movement patterns compared to controls.

Conclusions:

• AI-driven analysis of iris and eye movement biomarkers can reliably screen for depression.
• The system offers a potential objective, real-time solution for enhancing safety in transportation.
• This technology could significantly reduce accidents attributed to depression-related human error.