TOPSY: A picture description dataset to examine speech, language and communication in untreated first episode...

Recordings of human behaviour provide a rich source of data to detect features of conditions such as psychosis. Advanced technologies can now detect subtle speech aberrations that often go unnoticed by clinicians and family members. This is particularly important given that behavioral data, especially from untreated samples, are scarce. Speech is a crucial verbal behavioral data point for understanding psychosis, but the lack of comprehensive datasets has been a major limitation in research. Furthermore, long-term antipsychotic treatment can significantly affect speech, impacting both its tone and content. This makes it critical to collect longitudinal data to understand these effects. Tracking Outcomes of Psychosis (TOPSY) study aims to address these gaps by providing a rich dataset that includes not only speech data collected using a semi-structured picture description protocol but also a range of other valuable measures based on a validated Thought and Language Index. The TOPSY dataset offers naturalistic speech samples from individuals in an early-stage psychosis treatment program (<2 median days of antipsychotic treatment) and demographically matched healthy controls. This data is complemented by a comprehensive neurobiological and clinical battery, including: MRS (Magnetic Resonance Spectroscopy), Structural MRI, T1 structural data with myelin mapping, DTI (Diffusion Tensor Imaging), Resting fMRI (functional Magnetic Resonance Imaging). In addition to the neuroimaging data, we have collected detailed clinical and cognitive information, including TLI scores (Thought and Language Index) verbal fluency, modified DSST, and PANSS-8 item scores. The protocol and clinical assessments were repeated after a 12-month interval to assess the longitudinal stability of speech, symptom burden, and functional status. Transcripts were extracted from conversations lasting between 3 and 5 min, allowing for in-depth analysis of acoustic, semantic, syntactic, and pragmatic measures related to psychosis. We expect this dataset to be invaluable for future investigations into the clinical utility of speech measures for assessing thought disorder and other psychosis-related symptoms.