Semantic and syntactic analysis of speech in patients at ultra-high-risk for psychosis: A proof of concept study

Poster B78, Tuesday, October%209, 11:30%20am%20-%201:00%20pm, Essex%20Ballroom

Pierre-françois Bazziconi1, Christophe Lemey1, DH Kim-Duffor, Laure Bleton1, Michel Walter1; 1CHRU BREST, Psychiatry

In the early psychosis detection field, the challenge is to identify a predictive marker of transition to schizophrenia. Language disorders,could be one of these markers. Computerized speech analysis techniques such as Latent Semantic Analysis (LSA) have already proven their reliability in schizophrenia (Elvevag, 2007, Schizophrenia research). One study showed that a combination of semantic and syntactic analysis would accurately predict the psychotic transition (Bedi G,2015, NPJ Schizophrenia). The aim of our study is to validate this model in french language as well as identifying specific linguistic markers of the psychotic transition. Patients will be recruited during two years from the early psychosis detection centers in Brest, Paris and in Lausanne. The initial report, including the Comprehensive Assessment of At-Risk Mental States will be completed with an audio recording from the initial medical interview. The recording will be transcribed and analyzed by computer following the method of LSA. An analysis of the grammatical function (number of words, rate of the various grammatical functions) will also be performed. With this first analysis, linguistic markers correlated to transition to psychosis, can be identified. Then, we wiil use these markers and the global coherence index to construct a predictive model for transition to schizophrenia. Finally this model will be tested through machine learning, where each new inclusion will enrich the algorithmic model. Our study leads to demonstrate that language abnormalities may be predictive markers of the psychotic transition in high-risk psychotic patients, using a model based on automatize semantic and syntactic analysis

Topic Area: Ultra High Risk / Prodromal Research

Back to Poster Schedule