Related Experiment Video
Updated: Jun 20, 2026
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Comparative Efficacy of MultiModal AI Methods in Screening for Major Depressive Disorder: Machine Learning Model
Donghao Chen1, Pengfei Wang2,3, Xiaolong Zhang2,3
1School of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing, China.
JMIR Formative Research
|May 30, 2025
Summary
Artificial intelligence (AI) analysis of audiovisual signals shows promise for major depressive disorder (MDD) screening. The question and answering (Q&A) paradigm demonstrated higher efficacy than mental imagery description (MID) for objective MD screening.
Area of Science:
- Psychiatry
- Artificial Intelligence
- Biomedical Engineering
Background:
- Major depressive disorder (MDD) screening traditionally relies on subjective self-rated scales and clinical interviews.
- Artificial intelligence (AI) offers potential for objective psychiatric assessment using audiovisual signals.
Purpose of the Study:
- To evaluate the efficacy of different AI-driven paradigms for analyzing audiovisual signals in MDD screening.
- To compare the performance of conventional scale (CS), question and answering (Q&A), mental imagery description (MID), and video watching (VW) paradigms.
Main Methods:
- Recruited 89 participants (41 with MDD, 48 asymptomatic).
- Developed AI models analyzing facial movement, acoustic, and text features from videos.
- Utilized ablation experiments and 5-fold cross-validation with two AI methods.
Main Results:
- The Q&A paradigm showed higher sensitivity (79.06%) than MID (P=.03) in video clip analysis.
- Combining Q&A and MID improved individual-level accuracy (80.00%) compared to MID alone (P=.01).
- AI models achieved over 76.25% binary accuracy for video predictions and 74.12% for individual predictions.
Conclusions:
- The Q&A paradigm is more effective than MID for MDD screening, individually and combined.
- AI analysis of audiovisual signals across multiple paradigms shows potential as an effective tool for objective MDD screening.