Artificial Intelligence in Diagnosing Depression Through Behavioural Cues: A Diagnostic Accuracy Systematic Review and Meta-Analysis.
Yee Shyan Goh, Qi Rui See, Nopporn Vongsirimas et al.
✦ AI-curated · Sources linked
In 30 seconds
This diagnostic accuracy systematic review and meta-analysis assessed AI methods for detecting depression from behavioural cues in adults in healthcare and community settings. It included 30 studies; 20 reported that AI models had potential to detect depression. Across modalities, speech and facial expression showed better sensitivity, while text and movement showed better specificity. Heterogeneity was initially high and was lower within modality subgroups. The review considered speech, text, movement and facial expression, but the abstract does not report pooled accuracy estimates or numerical comparisons between modalities.
Key findings
- Of 30 included studies, 20 indicated potential for AI models to detect depression.
- Speech and facial expression showed better sensitivity for identifying people with depression.
- Text and movement showed better specificity for ruling out non-depressed individuals.
- Heterogeneity was initially high, with less heterogeneity within modality subgroups.
Why it matters
Depressive symptoms can be challenging to detect in general healthcare and community settings. The findings suggest that behavioural cues analysed by AI may offer a basis for objective screening, with different modalities showing strengths in sensitivity or specificity. This could inform development and evaluation of screening approaches, but the review does not establish that such systems improve diagnosis or patient outcomes.
What to keep in mind
The abstract reports high initial heterogeneity and provides no pooled accuracy estimates or numerical results. The evidence is based on included studies rather than a reported evaluation of an AI system in clinical practice.