Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jul 2, 2026

Closed-Loop Neurostimulation for Biomarker-Driven, Personalized Treatment of Major Depressive Disorder
05:19

Closed-Loop Neurostimulation for Biomarker-Driven, Personalized Treatment of Major Depressive Disorder

Published on: July 7, 2023

Multimodal behavioral phenotyping for depressive-spectrum classification and severity estimation using eye tracking,

Xiang-Ting Chen1, Min Huang1

  • 1Department of General Medicine, The Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, Nanjing Medical University, Suzhou, Jiangsu, China.

Frontiers in Psychiatry
|July 1, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Preexisting human antibodies neutralize recently emerged H7N9 influenza strains.

The Journal of clinical investigation·2015
Same author

Inhibitory effects of B-cell lymphoma 2 on the vasculogenic mimicry of hypoxic human glioma cells.

Experimental and therapeutic medicine·2015
Same author

Simultaneous determination of seven taxoids in rat plasma by UPLC-MS/MS and pharmacokinetic study after oral administration of Taxus yunnanensis extracts.

Journal of pharmaceutical and biomedical analysis·2015
Same author

Core muscle strength and endurance measures in ambulatory persons with multiple sclerosis: validity and reliability.

International journal of rehabilitation research. Internationale Zeitschrift fur Rehabilitationsforschung. Revue internationale de recherches de readaptation·2015
Same author

The kinase MST4 limits inflammatory responses through direct phosphorylation of the adaptor TRAF6.

Nature immunology·2015
Same author

Obesity and risk of thyroid cancer: evidence from a meta-analysis of 21 observational studies.

Medical science monitor : international medical journal of experimental and clinical research·2015

This study introduces a multimodal AI framework for accurate depression assessment, classifying severity and spectrum using eye tracking, facial behavior, and language. It enhances clinical evaluation, especially for borderline depression cases.

Area of Science:

  • Computational psychiatry
  • Multimodal machine learning
  • Digital phenotyping

Background:

  • Current depression assessment relies heavily on subjective reports and clinician judgment.
  • Objective tools for stratifying depressive disorders and estimating symptom severity are limited.
  • Existing digital depression detection often uses binary classification and struggles with missing data and calibration.

Purpose of the Study:

  • To develop a quality-aware multimodal framework for classifying normal control (NC), subthreshold depression (SD), and major depressive disorder (MDD).
  • To predict the severity of depressive symptoms using the 17-item Hamilton Depression Rating Scale (HAMD-17).
  • To address limitations in existing models, including handling missing data, calibration, and combining classification with severity estimation.
Keywords:
eye trackingfacial behaviormajor depressive disordermultimodal machine learningsubthreshold depression

Related Experiment Videos

Last Updated: Jul 2, 2026

Closed-Loop Neurostimulation for Biomarker-Driven, Personalized Treatment of Major Depressive Disorder
05:19

Closed-Loop Neurostimulation for Biomarker-Driven, Personalized Treatment of Major Depressive Disorder

Published on: July 7, 2023

Main Methods:

  • Integrated eye tracking, facial behavior, and transcript-derived language data.
  • Employed modality-specific encoders, quality-aware gated fusion, and joint classification-regression learning.
  • Incorporated Transformer-based cross-modal interaction and uncertainty-based dynamic task weighting for improved performance and interpretability.

Main Results:

  • The developed framework achieved high accuracy (approaching 0.90) in classifying depressive states.
  • Demonstrated improved calibration and reduced misclassification, particularly near the subthreshold depression boundary.
  • Facial features were the dominant signal, complemented by eye tracking and language data, with stable modality reweighting.

Conclusions:

  • The framework offers a robust solution for depression spectrum classification and severity estimation, outperforming previous models.
  • It effectively handles missing modalities and provides calibrated, interpretable predictions.
  • This tool can augment clinical assessments, especially for nuanced cases like subthreshold depression.