Interdisciplinary approach to identify language markers for post-traumatic stress disorder using machine learning and...

Area of Science:

• Psychiatry and Computational Linguistics
• Natural Language Processing (NLP) applications in mental health

Background:

• Post-traumatic stress disorder (PTSD) lacks established clinical biomarkers.
• Investigating language as a potential diagnostic tool for PTSD is crucial for early detection and treatment.

Purpose of the Study:

• To explore the relationship between language patterns and PTSD in survivors of the 2015 Paris terrorist attacks.
• To evaluate the efficacy of interdisciplinary methods integrating psychiatry, linguistics, and NLP for PTSD diagnosis.

Main Methods:

• Analysis of interview transcripts from 148 Paris attack survivors.
• A three-step methodology: clinical psychiatrist diagnosis, psycholinguistic feature-based machine learning, and a hypothesis-free deep learning approach.
• Controlled for confounding factors and analyzed associations with DSM-5 subsymptoms.

Main Results:

• Clinical psychiatrist diagnosis achieved an Area Under Curve (AUC) of 0.72.
• Machine learning models yielded an AUC of 0.69, and deep learning achieved an AUC of 0.64.
• Established associations between language and PTSD subsymptoms, demonstrating the utility of linguistic markers.

Conclusions:

• Language analysis, particularly through NLP, shows promise as a diagnostic biomarker for PTSD.
• The findings support the development of advanced diagnostic tools and methods for assessing treatment effectiveness in PTSD.
• This research provides a robust foundation for utilizing linguistic data in clinical psychiatric practice.