LAGUN: Learning from the Analysis of Generative approaches and Unstructured Narratives for mental health understanding (LAGUN)

TITLE OF THE COORDINATED PROJECT (ACRONYM): understanding Mental health across socIally relevaNt Diseases through explainable InSIGHTs from Multisource Text Data (MIND-SIGHT) This project focuses on the early detection, characterisation and longitudinal analysis of mental health, as well as on the study of psychological disorders associated with diseases of high social relevance, including rare disorders, HIV and cancer. These areas of clinical interest require advanced methods capable of integrating diverse information sources. Electronic Health Records (EHRs) will serve as the primary data source, complemented by social media content, patient-generated information, and additional modalities such as text or audio recordings. This combination will enable the capture of early signals, the modelling of clinical trajectories and a deeper understanding of the psychological impact of these conditions. The proposal is framed within the thematic area “Information and Communication Technologies”, within the sub-area “Computer Science and Information Technology”, and with the overarching thematic priority of “Health” in alignment with the priorities of the State Plan for Scientific, Technical and Innovation Research 2024-2027 related to digital health, trustworthy artificial intelligence and data-driven biomedical innovation. The project addresses the growing need for preventive and personalised healthcare by focusing on diseases of high social relevance, including rare disorders, HIV and cancer, and incorporating mental health as a core analytical dimension. Early identification of risks associated with these conditions is essential to improve prognosis, reduce healthcare burden and support patient well-being. The scientific contribution of the project builds on three interconnected pillars: advancing explainable and interpretable NLP methods capable of providing transparent and clinically meaningful justifications; developing and evaluating synthetic biomedical data to overcome limitations in resource availability while preserving privacy; and integrating mental-health indicators extracted from clinical narratives, social media and other heterogeneous sources to better characterise patients’ trajectories. Through this combination, the proposal aims to deliver trustworthy, equitable and socially relevant technologies that contribute to the digital transformation of healthcare and to the strategic objectives defined in the national research agenda. TITLE OF THE SUB-PROJECT (ACRONYM): Learning from the Analysis of Generative approaches and Unstructured Narratives for mental health understanding (LAGUN) TEAM: HITZ MOTIVATION: LAGUN sub-project focuses on disruptive e-Health technologies to contribute to the creation of a European health data space to foster targeted research on health conditions particularly related to mental health. Current research applied to either Spanish or Basque clinical domain struggles to promote NLP -based tool adaptation to clinical domain and also reproducible research and this is mainly due to the lack of available clinical data. The project pursues three main motivations in corresponding pillars. A pillar of this project is to develop artificial though realistic patient histories to boost clinical NLP. A second pillar rests on the ability to communicate health conditions to patients in lay language in a concise though correct and clear manner. The third pillar involves the ability to monitor health conditions to promote preventive medicine in terms of early detection mechanisms. SCIENTIFIC OBJECTIVES (HITZ SUBPROJECT): SOH1: Development of core NLP tools for medical text analysis with focus on mental health and Spanish, Basque and English as target languages SOH2: Processing medical unstructured (EHRs and social media) text in Spanish or English and synthetic lay-text generation in Basque SOH3: Patient centered NLP: Agents to aid patients to clarify inquiries on their own health history SOH4: Clinician centered NLP: Early detection mental health risks with special focus on suicide