Investigation of linguistic parameters in affective and psychotic disorders and their brain structural correlates.

• Applicant: Frederike Stein, Ph.D.
• Subject Area: Biological Psychiatry ; Applied Linguistics, Computational Linguistics ; Human Cognitive and Systems Neuroscience
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 527712970

Mental disorders are diagnosed using standardized categorization systems. However, there is ample evidence that there is a great overlap between many categorical diagnoses at various levels. Therefore, alternative methods are needed to gain a better understanding of underlying pathomechanisms. Hereof, multivariate, transdiagnostic dimensional approaches are a novel framework, with the idea that brain structural and physiological changes across diagnoses cluster along latent variables. Ideally, these are easy to measure and include large brain networks. Spontaneous speech, as the most complex human cognitive-motor task, fulfills these criteria because it involves coordinated motor, auditory, cognitive, and emotional processes. The proposed project aims to paradigmatically investigate latent linguistic parameters (LLP) of speech production in N=392 patients with Major Depressive Disorder (MDD), Bipolar Disorder (BD), and Schizophrenia (SZ). Using computational linguistic tools, algorithms for extracting LLP of phonology, syntax and semantics will be developed, tested and validated in work package 1 (WP). Subsequently, extracted LLP will be analyzed for transdiagnostic dimensions and subgroups using multivariate methods. Here, transdiagnostic linguistic dimensions and clusters are expected to be present across MDD, BD, and SZ. Finally, algorithms will be documented and made available to other researchers via a GitHub repository. WP 2 will integrate the identified dimensions and subclusters into a transdiagnostic model of spontaneous speech in affective and psychotic disorders by examining brain structural correlates (white and gray matter) using MRI at both the whole-brain level i.e., local associations and at the global macroscopic level i.e., structural networks (connectomics). With respect to brain structural correlates of linguistic parameters, associations are expected primarily in the language network. WP3, will conduct a validation study with n=60 by using further language production tasks in addition to the image description, which should provide information about reliability and generalizability. In summary, the proposed project aims to decode linguistic changes in "endogenous psychoses" in a systematic and automated way. The long-term goal is to develop an integrative neurobiological model of language in affective and psychotic disorders. In a next step, findings from this project could be used to train prediction algorithms investigating the course of affective and psychotic disorders.

• DFG Programme: Research Grants