NIMHANS, IIT Kharagpur and LGBRIMH Partner to Develop AI-Based Depression Screening Tool

The National Institute of Mental Health and Neuro Sciences (NIMHANS), IIT Kharagpur, and the Lokopriya Gopinath Bordoloi Regional Institute of Mental Health (LGBRIMH) have officially launched a collaborative research initiative aimed at improving mental health diagnostics. The project, titled Human-in-the-loop Evaluation of Assisted Depression Screening (HEADS), seeks to develop an AI-powered system capable of screening for depression across multiple Indian languages.

Funded by the Wellcome Charitable Foundation, this two-year research endeavor focuses on making mental health assessments more inclusive and accessible. By enabling clinical interviews to be conducted in Kannada, Hindi, Bengali, Assamese, and English, the project aims to overcome language barriers that often hinder the accurate reporting of psychological distress. The initiative was formalized through the signing of two Memoranda of Understanding (MoUs) between the participating institutions.

Project Scope and Technical Implementation

The HEADS project is designed to create an open-source AI system that assists clinicians by transcribing and translating clinical interviews. The technology will generate a comprehensive summary of the conversation, provide a preliminary depression assessment, and offer an estimate of the condition's severity. Crucially, the system will also provide the reasoning behind its findings and a confidence score to guide the clinician. Despite these advanced capabilities, the project emphasizes that the AI will not replace human judgment; clinicians will retain full authority to review, correct, and finalize all assessments.

IIT Kharagpur is spearheading the AI engineering component, which includes the development and rigorous testing of the models. A key part of their mandate is to evaluate the system for safety and potential bias, ensuring that the technology is reliable for clinical environments. This technical focus addresses findings from pilot studies, which indicated that existing commercial speech recognition and language models often fail to capture the nuances, idioms, and specific expressions used by individuals to describe mental health struggles in their native languages.

Clinical Data Collection and Future Objectives

The clinical phase of the project involves extensive data collection across the partner institutions. NIMHANS will manage data gathering in Kannada, Hindi, and English, while LGBRIMH will focus on Assamese, Bengali, and English. Over the 24-month duration of the project, the researchers plan to interview 4,500 participants, including 4,000 patients and 500 healthy volunteers, to ensure the AI is trained on a diverse and representative dataset.

The ultimate goal of the HEADS project is to facilitate the earlier recognition of depression. By developing tools that function effectively in regional languages, the institutions hope to empower non-specialist health workers to conduct screenings, thereby expanding the reach of mental health services to populations who may find it difficult to communicate their distress in English. The project represents a significant step toward integrating advanced technology into India's public health infrastructure to improve patient outcomes.