A multimodal digital phenotyping dataset for depression and anxiety assessment under free-living conditions

ABSTRACT: Despite affecting hundreds of millions of people globally, depression and anxiety remain understudied through digital phenotyping in resource-constrained settings, where limited clinical capacity highlights the need for scalable monitoring. In this work, we present Neurai-VN, a multimodal dataset from a Vietnamese population integrating passive sensing and active assessments across multiple temporal scales. Data were collected from 100 Vietnamese adults recruited from the general population over two weeks. Participants were grouped into four mutually exclusive diagnostic groups based on clinical assessments: depressive disorders, anxiety disorders, healthy controls, and other psychiatric conditions. The dataset contains (1) wearable signals and smartphone-derived data captured under free-living conditions; (2) clinical assessment data, including DSM-5-based psychiatric diagnoses and symptom severity ratings; and (3) longitudinal self-reports, including PHQ-9, GAD-7, daily symptom reports, and mood logs. Overall, the released dataset contains 1,730 participant-day records from 13 passive sensing modalities and 3,642 self-report records. The Neurai-VN dataset provides a resource for reproducible computational analyses and machine learning research on multimodal digital phenotyping for mental health-related outcomes.

Latest Update: Version 1.0.2