Depression Severity Estimation
Overview
Depression severity estimation aims to measure symptom levels for screening, clinical decision support, and tracking change over time, often against scales such as the PHQ-8/9 or MADRS. Approaches range from statistical models of questionnaire and behavioral features to machine learning on speech, text, video, and wearable data, as well as large language models that score interviews or conversations; multimodal fusion and symptom-guided attention are common strategies. Evaluation uses clinical interview datasets and benchmarks such as DAIC-WOZ and AVEC, but performance can shift across languages, populations, interview protocols, and assessment scales. A central tension is that strong agreement with labels does not by itself establish clinical validity: calibration, evidence faithfulness, cross-context generalization, and reliable measurement of change remain open challenges.
Latest papers
36 papers2 papers in the last four weeks, down 75% on the four weeks before. 0.0% of all new papers.