Comparison of the structure of speech and phonetic properties with clinical features in patients with schizophrenia and their healthy siblings

Introduction: Schizophrenia; is a chronic psychiatric disorder in which early brain development is affected by heterogeneous genetic and neurobiological mechanisms in the form of psychotic symptoms characterized with positive symptoms such as hallucinations, delusions and disorganization, negative symptoms such as motivational and cognitive decline. Schizophrenia has always been one of the most interesting diagnostic groups in language research in psychiatry. Even in the early years when the disease was firstly described, the deterioration in the language attracted attention, and the speech samples of the patients were included in the first texts where the disease was described. Speech in schizophrenia is usually irregular and chaotic. In schizophrenia, different deteriorations have been detected in the form of prolongation in the number and duration of pauses, decrease in speech rate, vocal pitch and intensity variability, and these have been associated with especially negative symptoms. In neuroscience research, reliable, reproducible, clinically relevant biomarkers that can be used in the diagnosis, treatment and follow-up of schizophrenia have not yet been identified in more than a century. Recent advances in computational linguistics may offer the opportunity to transform language outputs into a new biomarker that is cheap, fast, and non-invasive. Our aim in our study is to evaluate the effects of schizophrenia on speech in the Turkish language by using computational linguistic and phonetic acoustic analysis methods, to observe the relationship with clinical findings, to compare these effects with healthy siblings and to reach more detailed findings about language acquisition and the effects of the disease itself. Methods: We included 26 patients with schizophrenia monitored at Istanbul University Istanbul Faculty of Medicine, Department of Psychiatry, 26 healthy siblings of the patients and 24 healthy controls matched with patients in terms of age, gender and educational status characteristics in our study. Recordings were taken from all participants for free conversational speech, personal narrative, picture descriptions, dream report and reading aloud tasks. The recordings were deciphered into text files by another blind researcher, speech graphs for each speech task, audio recordings were digitally analyzed and acoustic phonetic analysis of data related to the duration, frequency and pitch changes of speech were created. The clinical characteristics of patients with schizophrenia were measured by the Brief Psychiatric Rating Scale (BPRS), the Brief Negative Symptoms Scale (BNSS), the Scale for the Assessment of Positive Symnptoms (SAPS) and the Calgary Depression Scale for Schizophrenia (CDSS). The structural and acoustic phonetic features of speech were compared between schizophrenia, healthy sibling and the control group, and its relationship with clinical scale scores for the schizophrenia group was examined. Results: Schizophrenia group differed significantly from the control group in terms of the measures associated with quantity and connectivity in the speech they produced in all speech tasks. Patients create simpler sentence structures with fewer words for the same speech tasks. The density of speech was also reduced compared to the control group. The healthy siblings of the patients produced speeches that were similar to the patients in terms of the quantity, connectivity and intensity of speech and significantly different from the control group. In regard to acoustic phonetics, patients spend a larger percentage of speech with pauses and waits, and in many speech tasks, they produce a more static speech. However, the differences seen in the literature in terms of speech rate, fundamental frequency values and variability were not detected in our study. Conclusions: In patients with schizophrenia, there are significant disturbances in the structure of speech and acoustic phonetic features. The severity of these disturbances correlates with the severity of clinical symptoms. In particular, as the scale subscores where negative symptoms are measured increase, verbosity is decreasing and speech becomes simpler and poorer. There is also a relationship between clinical symptoms and acoustic phonetic analysis results. As the clinical symptoms worsen, the voice appears to become deeper and blunter with less variability throughout speech. At the same time, similar disturbances are detected in some speech attributes in the healthy siblings of the patients. Our study also shows that computational linguistic analyzes can also be used in Turkish and can be a clinical biomarker candidate.