-
1
A multimodal framework was developed to classify depression spectrum and assess severity using eye tracking, facial expressions, and language analysis.
-
2
The study included 186 participants completing tasks that generated data for eye tracking, facial behavior, and language analysis.
-
3
Baseline-3+ outperformed previous models, achieving nearly 0.90 accuracy in classification and lower calibration error compared to Baseline-3.
-
4
Facial features were the most significant indicators of depression, supported by eye tracking and language contributions in the analysis.
-
5
The framework aims to enhance clinical assessment of depression, particularly for subthreshold depression, addressing limitations of existing models.
