Combining automatic speech recognition with semantic natural language processing in schizophrenia.

Natural language processing (NLP) tools are increasingly used to quantify semantic anomalies in schizophrenia. Automatic speech recognition (ASR) technology, if robust enough, could significantly speed up the NLP research process. In this study, we assessed the performance of a state-of-the-art ASR tool and its impact on diagnostic classification accuracy based on a NLP model. We compared ASR to human transcripts quantitatively (Word Error Rate (WER)) and qualitatively by analyzing error type and position. Subsequently, we evaluated the impact of ASR on classification accuracy using semantic similarity measures. Two random forest classifiers were trained with similarity measures derived from automatic and manual transcriptions, and their performance was compared. The ASR tool had a mean WER of 30.4%. Pronouns and words in sentence-final position had the highest WERs. The classification accuracy was 76.7% (sensitivity 70%; specificity 86%) using automated transcriptions and 79.8% (sensitivity 75%; specificity 86%) for manual transcriptions. The difference in performance between the models was not significant. These findings demonstrate that using ASR for semantic analysis is associated with only a small decrease in accuracy in classifying schizophrenia, compared to manual transcripts. Thus, combining ASR technology with semantic NLP models qualifies as a robust and efficient method for diagnosing schizophrenia.

[Automated speech analysis can distinguish frontotemporal dementia from depression: a pilot study]. / Geautomatiseerde spraakanalyse bij onderscheiden van frontotemporale dementie en depressie.

BACKGROUND: Differentiating the behavioural variant of frontotemporal dementia from a depression is challenging. Recent development of automated speech analyses might add to diagnostic. AIM: To investigate the value of automated speech analyses in differentiating bvFTD from a depressive disorder. METHOD: A semistructured interview was recorded in 15 patients with bvFTD, 15 patients with a depressive disorder and 15 healthy controls, which was transcribed and analysed. Acoustic and semantic values were extracted and classified using machine learning. RESULTS: Acoustic values showed an 80% accuracy for differentiating bvFTD from depressive disorder and semantic values showed an 70.8% accuracy. CONCLUSION: Acoustic as well as semantic values show significant differences between bvFTD and depressive disorder. In automated speech analyses researches should consider privacy matters as well as possible confounders like age, sex and ethnicity. This study should be repeated in a larger population.

[Recognizing depression through computational language analysis]. / Herkennen van depressie met computationele taalanalyse.

BACKGROUND: Currently, clinical practice lacks a usable biomarker for the detection and differentiation of depression. Such a biomarker may be found in speech, from which important information can be distilled using automated speech analysis. AIM: To provide an overview of the fast-developing field of automated speech analysis for depression. METHOD: We summarize the current literature on speech features in depression. RESULTS: Current computational models can detect depression with high accuracy, rendering them applicable for diagnostic tools based on automatic speech analysis. Such tools are developing at a fast rate. CONCLUSION: Some challenges are still in the way of clinical implementation. For example, results differ largely between studies due to much variation in methodology. Furthermore, privacy and ethical issues need to be addressed before tools can be used.

Assessing coherence through linguistic connectives: Analysis of speech in patients with schizophrenia-spectrum disorders.

BACKGROUND: Incoherent speech is a core diagnostic symptom of schizophrenia-spectrum disorders (SSD) that can be studied using semantic space models. Since linguistic connectives signal relations between words, they and their surrounding words might represent linguistic loci to detect unusual coherence in speech. Therefore, we investigated whether connectives' measures are useful to assess incoherent speech in SSD. METHODS: Connectives and their surrounding words were extracted from transcripts of spontaneous speech of 50 SSD-patients and 50 control participants. Using word2vec, two different cosine similarities were calculated: those of connectives and their surrounding words (connectives-related similarity), and those of free-of-connectives words-chunks (non-connectives similarity). Differences between groups in proportion of five types of connectives were assessed using generalized logistic models, and connectives-related similarity was analyzed through non-parametric multivariate analysis of variance. These features were evaluated in classification tasks to differentiate between groups. RESULTS: SSD-patients used less contingency (e.g., because) (p = .008) and multiclass connectives (e.g., as) (p < .001) than control participants. SSD-patients had higher minimum similarity of multiclass (adj-p = .04) and temporality connectives (e.g., after) (adj-p < .001), narrower similarity-range of expansion (e.g., and) (adj-p = .002) and multiclass connectives (adj-p = .04), and lower maximum similarity of expansion connectives (adj-p = .005). Using connectives' features alone, SSD-patients and controls could be distinguished with 85 % accuracy. DISCUSSION: Our results show that SSD-speech can be distinguished from speech of control participants with high accuracy, based solely on connectives' features. We conclude that including connectives could strengthen computational models to categorize SSD.

Acoustic speech markers for schizophrenia-spectrum disorders: a diagnostic and symptom-recognition tool.

BACKGROUND: Clinicians routinely use impressions of speech as an element of mental status examination. In schizophrenia-spectrum disorders, descriptions of speech are used to assess the severity of psychotic symptoms. In the current study, we assessed the diagnostic value of acoustic speech parameters in schizophrenia-spectrum disorders, as well as its value in recognizing positive and negative symptoms. METHODS: Speech was obtained from 142 patients with a schizophrenia-spectrum disorder and 142 matched controls during a semi-structured interview on neutral topics. Patients were categorized as having predominantly positive or negative symptoms using the Positive and Negative Syndrome Scale (PANSS). Acoustic parameters were extracted with OpenSMILE, employing the extended Geneva Acoustic Minimalistic Parameter Set, which includes standardized analyses of pitch (F0), speech quality and pauses. Speech parameters were fed into a random forest algorithm with leave-ten-out cross-validation to assess their value for a schizophrenia-spectrum diagnosis, and PANSS subtype recognition. RESULTS: The machine-learning speech classifier attained an accuracy of 86.2% in classifying patients with a schizophrenia-spectrum disorder and controls on speech parameters alone. Patients with predominantly positive v. negative symptoms could be classified with an accuracy of 74.2%. CONCLUSIONS: Our results show that automatically extracted speech parameters can be used to accurately classify patients with a schizophrenia-spectrum disorder and healthy controls, as well as differentiate between patients with predominantly positive v. negatives symptoms. Thus, the field of speech technology has provided a standardized, powerful tool that has high potential for clinical applications in diagnosis and differentiation, given its ease of comparison and replication across samples.

Implementation of automatic speech analysis for early detection of psychiatric symptoms: What do patients want?

Psychiatry is in dire need of a method to aid early detection of symptoms. Recent developments in automatic speech analysis prove promising in this regard, and open avenues for implementation of speech-based applications to detect psychiatric symptoms. The current survey was conducted to assess positions with regard to speech recordings among a group (n = 675) of individuals who experience psychiatric symptoms. Overall, respondents are open to the idea of speech recordings in light of their mental welfare. Importantly, concerns with regard to privacy were raised. Given that speech recordings are privacy sensitive, this requires special attention upon implementation of automatic speech analysis techniques. Furthermore, respondents indicated a preference for speech recordings in the presence of a clinician, as opposed to a recording made at home without the clinician present. In developing a speech marker for psychiatry, close collaboration with the intended users is essential to arrive at a truly valid and implementable method.

Quantified language connectedness in schizophrenia-spectrum disorders.

Language abnormalities are a core symptom of schizophrenia-spectrum disorders and could serve as a potential diagnostic marker. Natural language processing enables quantification of language connectedness, which may be lower in schizophrenia-spectrum disorders. Here, we investigated connectedness of spontaneous speech in schizophrenia-spectrum patients and controls and determine its accuracy in classification. Using a semi-structured interview, speech of 50 patients with a schizophrenia-spectrum disorder and 50 controls was recorded. Language connectedness in a semantic word2vec model was calculated using consecutive word similarity in moving windows of increasing sizes (2-20 words). Mean, minimal and variance of similarity were calculated per window size and used in a random forest classifier to distinguish patients and healthy controls. Classification based on connectedness reached 85% cross-validated accuracy, with 84% specificity and 86% sensitivity. Features that best discriminated patients from controls were variance of similarity at window sizes between 5 and 10. We show impaired connectedness in spontaneous speech of patients with schizophrenia-spectrum disorders even in patients with low ratings of positive symptoms. Effects were most prominent at the level of sentence connectedness. The high sensitivity, specificity and tolerability of this method show that language analysis is an accurate and feasible digital assistant in diagnosing schizophrenia-spectrum disorders.

Language disturbances in schizophrenia: the relation with antipsychotic medication.

Language disturbances are key aberrations in schizophrenia. Little is known about the influence of antipsychotic medication on these symptoms. Using computational language methods, this study evaluated the impact of high versus low dopamine D2 receptor (D2R) occupancy antipsychotics on language disturbances in 41 patients with schizophrenia, relative to 40 healthy controls. Patients with high versus low D2R occupancy antipsychotics differed by total number of words and type-token ratio, suggesting medication effects. Both patient groups differed from the healthy controls on percentage of time speaking and clauses per utterance, suggesting illness effects. Overall, more severe negative language disturbances (i.e. slower articulation rate, increased pausing, and shorter utterances) were seen in the patients that used high D2R occupancy antipsychotics, while less prominent disturbances were seen in low D2R occupancy patients. Language analyses successfully predicted drug type (sensitivity = 80.0%, specificity = 76.5%). Several language disturbances were more related to drug type and dose, than to other psychotic symptoms, suggesting that language disturbances may be aggravated by high D2R antipsychotics. This negative impact of high D2R occupancy drugs may have clinical implications, as impaired language production predicts functional outcome and degrades the quality of life.

Language in schizophrenia: relation with diagnosis, symptomatology and white matter tracts.

Language deviations are a core symptom of schizophrenia. With the advances in computational linguistics, language can be easily assessed in exact and reproducible measures. This study investigated how language characteristics relate to schizophrenia diagnosis, symptom, severity and integrity of the white matter language tracts in patients with schizophrenia and healthy controls. Spontaneous speech was recorded and diffusion tensor imaging was performed in 26 schizophrenia patients and 22 controls. We were able to classify both groups with a sensitivity of 89% and a specificity of 82%, based on mean length of utterance and clauses per utterance. Language disturbances were associated with negative symptom severity. Computational language measures predicted language tract integrity in patients (adjusted R2 = 0.467) and controls (adjusted R2 = 0.483). Quantitative language analyses have both clinical and biological validity, offer a simple, helpful marker of both severity and underlying pathology, and provide a promising tool for schizophrenia research and clinical practice.

Clinical use of semantic space models in psychiatry and neurology: A systematic review and meta-analysis.

Verbal communication disorders are a hallmark of many neurological and psychiatric illnesses. Recent developments in computational analysis provide objective characterizations of these language abnormalities. We conducted a meta-analysis assessing semantic space models as a diagnostic or prognostic tool in psychiatric or neurological disorders. Diagnostic test accuracy analyses revealed reasonable sensitivity and specificity and high overall efficacy in differentiating between patients and controls (n=1680: Hedges' g =.73, p=.001). Analyses of full sentences (Hedges' g =.95 p <.0001) revealed a higher efficacy than single words (Hedges' g = .51, p <.0001). Specifically, models examining psychotic patients (Hedges' g =.96, p=.003) and those with autism (Hedges' g = .84, p <.0001) were highly effective. Our results show semantic space models are effective as a diagnostic tool in a variety of psychiatric and neurological disorders. The field is still exploratory in nature; techniques differ and models are only used to distinguish patients from healthy controls so far. Future research should aim to distinguish between disorders and perhaps explore newer semantic space tools like word2vec.