For consortium members

TRUSTING: A TRUSTworthy speech-based AI monitoring system for the prediction of relapse in individuals with schizophrenia
Grant agreement: No. 101080251
Coordinated by: University Medical Center Groningen, Academisch Ziekenhuis Groningen – Universitair Medisch Centrum Groningen

Total cost: € 8.6 million



1 July 2023

31 Dec 2028

About the project

Psychotic disorders tend to have a waxing and waning course, with vulnerable persons being at risk for a relapse, especially when antipsychotic medication is no longer used. Such psychotic relapses can often be prevented, when a warning is received in time, so that the patient and their clinical team can take appropriate measures. 

These can consist of the reinstatement of medication, psychotherapy, social interventions or a combination of those. Accurate and timely warning can thus make a huge difference. We hypothesise that spontaneous speech specifically holds the key to predict whether psychosis is imminent. In this project, we will ask people at risk of psychotic relapse to record short speech samples at regular intervals through a private and safe (remote) electronic environment. 

Using artificial intelligence, we will analyse these speech fragments with the aim to recognise changes in speech. These changes may signal subtle tendencies of psychotic symptoms, such as hallucinations, delusions or thought disorder.

Our goal is to develop a user-friendly, trustworthy application that can be used from home to deliver a message when speech deviations predictive of psychotic relapse are detected. We hope that such an application will support people vulnerable for psychosis to live independently without fear of relapse with or without antipsychotic maintenance medication.

What We Want To Achieve

TRUSTING has three objectives

Relapse predictor

To create an accurate speech predictor for relapse in psychosis that is validated across languages, speech tasks and subgroups.

AI monitoring

To build a trustworthy speech-based AI monitoring tool for relapse prediction and test its efficacy in a randomised controlled trial.


To evaluate cost-effectiveness of the AI monitoring tool and define a protection and commercialisation roadmap.


Meet the enthusiastic experts leading the TRUSTING project