Automated analysis of free speech predicts psychosis onset in high-risk youths.

Bedi G, Carrillo F, Cecchi GA, Slezak DF, Sigman M, Mota NB, Ribeiro S, Javitt DC, Copelli M, Corcoran CM. · 2015

Read Automated analysis of free speech predicts psychosis onset in high-risk youths. by Bedi G, Carrillo F, Cecchi GA, Slezak DF, Sigman M, Mota NB, Ribeiro S, Javitt DC, Copelli M, Corcoran CM. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4849456.txt.

Introduction

Gillinder Bedi and co-authors tested in 2015 whether automated speech analysis could predict psychosis in high-risk youths. Psychiatry lacks the lab tests common in other branches of medicine. The group recorded interviews with 34 young people at high clinical risk, then tracked them for up to two and a half years. Five went on to develop psychosis.

Software scored meaning flow and sentence build in the interview transcripts. Those scores, fed to a learning method, flagged who later fell ill, and they also tracked with early symptom ratings. Speech is cheap to collect and score by machine. With larger tests, such tools could give clinics an early, neutral risk signal.

Catalog record

CREATORBedi G, Carrillo F, Cecchi GA, Slezak DF, Sigman M, Mota NB, Ribeiro S, Javitt DC, Copelli M, Corcoran CM.
YEAR2015
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC4849456

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Automated analysis of free speech predicts psychosis onset in high-risk youths By Gillinder Bedi, Facundo Carrillo, Guillermo A Cecchi, Diego Fernández Slezak, Mariano Sigman, Natália B Mota, Sidarta Ribeiro, Daniel C Javitt, Mauro Copelli, Cheryl M Corcoran Published in: NPJ Schizophrenia DOI: 10.1038/npjschz.2015.30 ABSTRACT Abstract Background/Objectives: Psychiatry lacks the objective clinical tests routinely used in other specializations. Novel computerized methods to characterize complex behaviors such as speech could be used to identify and predict psychiatric illness in individuals. AIMS: In this proof-of-principle study, our aim was to test automated speech analyses combined with Machine Learning to predict later psychosis onset in youths at clinical high-risk (CHR) for psychosis. Methods: Thirty-four CHR youths (11 females) had baseline interviews and were assessed quarterly for up to 2.5 years; five transitioned to psychosis. Using automated analysis, transcripts of interviews were evaluated for semantic and syntactic features predicting later psychosis onset. Speech features were fed into a convex hull classification algorithm with leave-one-subject-out cross-validation to assess their predictive value for psychosis outcome. The canonical correlation between the speech features and prodromal symptom ratings was computed. Results: Derived speech features included a Latent Semantic Analysis measure of semantic coherence and two syntactic markers of speech complexity: maximum phrase length and use of determiners (e.g., which). These speech features predicted later psychosis development with 100% accuracy, outperforming classification from clinical interviews. Speech features were significantly correlated with prodromal symptoms. Conclusions: Findings support the utility of automated speech analysis to measure subtle, clinically relevant mental state changes in emergent psychosis. Recent developments in computer science, including natural language processing, could provide the foundation for future development of objective clinical tests for psychiatry. INTRODUCTION The capacity of psychiatry to diagnose and treat serious mental illness has been hampered by the absence of objective clinical tests of the type routinely used in other fields of medicine. Although recent years have seen substantial advances in understanding of the neurobiology of mental illness,1 these developments have yet to yield markers that reliably differentiate psychiatric health from illness at the level of the individual patient. Whereas clinical neuroscience has focused on the brain in mental illness, computer science has, in parallel, developed increasingly sophisticated automated approaches to characterize and predict human behavior. Such advances are now commonly utilized in industry (the private business sector): models combining demographic data and purchasing behavior are used to personalize advertising content2 and automated language assessment is employed to screen job candidates and score essays.3 The degree to which such technologies might also aid diagnosis and prognosis in psychiatry is only now beginning to be explored (e.g., see ref. 4). Developments in automated natural language processing5 present one promising avenue for psychiatry. Although speech may present a unique ‘window into the mind’ in a variety of altered states,6 it is particularly relevant to psychosis. Thought disorder, a cardinal symptom of schizophrenia in which thought processes lose coherence, is typically diagnosed on the basis of clinical observation of disorganized speech.7 As a complement to clinical observation, automated analysis methods have previously been used to assess speech correlates of thought disorder in schizophrenia.8 For example, Latent Semantic Analysis (LSA), an automated high-dimensional associative analysis of semantic structure in speech, has been used to identify a reduction in semantic coherence in schizophrenia that correlates with clinical ratings and has comparable diagnostic accuracy.3 LSA combined with structural speech analysis was also able to accurately differentiate between first-degree relatives of schizophrenia patients and unrelated healthy individuals, suggesting that subtle differences indicative of underlying genetic vulnerabilities to schizophrenia can be distinguished with computerized speech analysis.9 As yet, however, these methods have not been applied to the prediction of psychosis onset, even though clinically diagnosed subtle disorganization in speech has consistently been identified as predictive of psychosis (i.e., with classification accuracy of ~60%) among young people identified as at clinical high risk (CHR) for psychosis (reviewed in ref. 10), as well as those at genetic high risk for psychosis.11 There are several reasons to test automated prediction approaches in this population. Schizophrenia, although relatively rare (lifetime prevalence ~1%), is among the most catastrophic mental illnesses both personally and societally. Schizophrenia and related psychotic illnesses typically emerge in young adults at the point of maximal societal and parental investment when individuals are poised to begin to contribute socially and economically.12 Although those at CHR for developing schizophrenia by virtue of subthreshold or attenuated psychotic symptoms can be identified,13 to date reliable prediction of psychosis onset among high-risk youths has proven elusive. Improving the capacity to predict psychosis among high-risk populations would have important ramifications for early identification and preventive intervention, potentially critically altering the long-term life trajectory of people with emergent psychotic disorders. Here, we present a proof-of-principle test of automated speech analysis to predict, at the level of the individual, the later onset of psychosis. Specifically, we employed analysis of free speech at baseline to predict psychosis onset over a subsequent period of up to 2.5 years in teens and young adults identified as at CHR for psychosis.13 On the basis of earlier findings in schizophrenia,3,9,14 in which automated text analyses yielded parameters that accurately discriminated between patients and controls, we hypothesized that automated semantic and syntactic analysis of baseline interview transcripts would yield speech features capable of predicting subsequent psychosis outcome among CHR individuals. MATERIALS AND METHODS Participants Participants were 34 help-seeking youths aged 14 to 27 years who were fluent in English (three were immigrants who learned English as children). They were referred from schools and clinicians, or self-referred through the Center of Prevention and Evaluation website. Exclusion criteria included history of threshold psychosis or Axis I psychotic disorder, risk of harm to self or others incommensurate with outpatient care, any major medical or neurological disorder, and Intelligence Quotient<70 (assessed with the Wechsler Abbreviated Scale of Intelligence). The attenuated psychotic symptoms characteristic of the CHR participants could not have occurred solely in the context of substance use or withdrawal. Adults provided written informed consent; participants under 18 provided written assent, with consent provided by a parent. All experiments were performed in accordance with the relevant guidelines and regulations, and all procedures were approved by the Institutional Review Board at the New York State Psychiatric Institute at Columbia University. Five participants transitioned to psychosis within 2.5 years of follow-up (CHR+), whereas 29 did not (CHR−). Demographics for CHR individuals, stratified by psychosis outcome, are presented in Table 1. [Table: Demographics] Procedures Ascertainment and prospective characterization The Structured Interview for Prodromal Syndromes/Scale of Prodromal Symptoms (SIPS/SOPS)13 was used for ascertainment of CHR status, for baseline and quarterly symptom ratings,10 and to determine psychosis outcome. The SIPS/SOPS evaluates positive (subthreshold psychotic), negative, disorganized, and general symptoms. Participants had to meet baseline criteria for one of three prodromal syndromes, assessed with the SIPS/SOPS: (i) attenuated positive symptom syndrome (⩾1 SOPS-positive item in the prodromal range with symptoms beginning or worsening in the past year, and symptoms occurring ⩾once/week in the prior month); (ii) genetic risk and deterioration syndrome (psychosis in a first-degree relative or schizotypal disorder accompanied by a 30% drop in global assessment of function over the past year); or (iii) brief intermittent psychotic symptom syndrome (⩾1 SOPS-positive items in the psychotic range with symptoms

Read the complete text