Depression Detection

Depression detection studies how to identify depressive symptoms and estimate their severity from sources such as clinical interviews, social media, speech, video, EEG, and brain imaging. Common approaches include transformer and large language models, multimodal fusion, temporal and graph learning, and interpretable symptom or acoustic features. These tools are explored for screening, monitoring change over time, and supporting clinical assessment, but are not substitutes for diagnosis. A central challenge is whether models generalize across people, settings, and sites—and whether their predictions reflect meaningful symptoms rather than demographic bias, interview artifacts, or mismatches in assessment measures.