Description du projet
L’analyse de la parole peut-elle aider à prévenir les psychoses?
Les troubles psychotiques ont tendance à évoluer en dents de scie, les personnes vulnérables étant exposées au risque de rechute. Or, ces rechutes peuvent souvent être évitées lorsqu’un avertissement est reçu à temps et que les mesures appropriées sont prises. Le projet TRUSTING, financé par l’UE, part de l’hypothèse que la parole spontanée est la clé pour prédire l’imminence d’une psychose. L’approche utilisée consistera à demander aux personnes présentant un risque de rechute psychotique d’enregistrer de courts extraits de discours à intervalles réguliers. Ces échantillons seront analysés à l’aide de l’intelligence artificielle, dans le but de reconnaître les changements qui signalent des symptômes psychotiques subtils. L’objectif est de développer une application conviviale et fiable qui puisse être utilisée à domicile pour envoyer une alerte lorsque des écarts de langage prédictifs d’une rechute psychotique sont détectés.
Objectif
Schizophrenia affects a staggering 21 million people worldwide, with 80% of these citizens suffering from a relapsing disease, putting their health and safety at enormous risk. Timely detection of these psychotic relapses would require very frequent contact with clinicians, which is neither desirable nor feasible. An accurate online relapse predictor could alert clinicians of subtle deterioration, which enables timely intervention and allow safe discontinuation of long-term medication, which so many affected citizens desire. Our Consortium demonstrated that subtle alterations in speech carry a predictive signal for psychosis onset. This project will develop an AI monitoring system that leverages spoken language processing (SLP) and natural language processing (NLP) of speech recorded at home to calculate the relapse risk. The monitoring tool we develop will be validated retrospectively in a longitudinal cohort, cross-sectionally, across six languages, after which it will be tested prospectively in a multicenter randomized trial, with the end goal of improving functional and clinical outcomes of those affected by schizophrenia. Developing such a system for exceptionally vulnerable people requires ‘buy-in’ from clinicians and mental health care service users, namely trust. A lack of trust is the biggest obstacle to the real-world implementation of a speech-based monitoring system. TRUSTING will develop a framework that systematically ensures addressing all the criteria for trustworthy AI put forward by the EU. This will ensure an empirically based and validated tool that can reliably detect pending relapse. As the core philosophy of trustworthiness is part of every aspect of the project, it will be a system more likely to be welcomed and embraced by service users and their carers. TRUSTING generates the scientific and social foundation for disruptive technology to deliver the unmet promise of an equitable and just form of healthcare for people at risk of relapse.
Champ scientifique
Mots‑clés
Programme(s)
Régime de financement
HORIZON-RIA - HORIZON Research and Innovation ActionsCoordinateur
9713 GZ Groningen
Pays-Bas