Natural Language Processing markers in first episode psychosis and people at clinical high-risk

Sarah E Morgan; Kelly Diederen; Petra E Vértes; Samantha H Y Ip; Bo Wang; Bethany Thompson; Arsime Demjaha; Andrea De Micheli; Dominic Oliver; Maria Liakata; Paolo Fusar-Poli; Tom J Spencer; Philip McGuire; Sarah Morgan

doi:10.1038/s41398-021-01722-y

Natural Language Processing markers in first episode psychosis and people at clinical high-risk

Sarah E Morgan, Kelly Diederen, Petra E Vértes, Samantha H Y Ip, Bo Wang, Bethany Thompson, Arsime Demjaha, Andrea De Micheli, Dominic Oliver, Maria Liakata, Paolo Fusar-Poli, Tom J Spencer, Philip McGuire, Sarah Morgan

Research output: Contribution to journal › Article › peer-review

35 Citations (Scopus)

Abstract

Recent work has suggested that disorganised speech might be a powerful predictor of later psychotic illness in clinical high risk subjects. To that end, several automated measures to quantify disorganisation of transcribed speech have been proposed. However, it remains unclear which measures are most strongly associated with psychosis, how different measures are related to each other and what the best strategies are to collect speech data from participants. Here, we assessed whether twelve automated Natural Language Processing markers could differentiate transcribed speech excerpts from subjects at clinical high risk for psychosis, first episode psychosis patients and healthy control subjects (total N = 54). In-line with previous work, several measures showed significant differences between groups, including semantic coherence, speech graph connectivity and a measure of whether speech was on-topic, the latter of which outperformed the related measure of tangentiality. Most NLP measures examined were only weakly related to each other, suggesting they provide complementary information. Finally, we compared the ability of transcribed speech generated using different tasks to differentiate the groups. Speech generated from picture descriptions of the Thematic Apperception Test and a story re-telling task outperformed free speech, suggesting that choice of speech generation method may be an important consideration. Overall, quantitative speech markers represent a promising direction for future clinical applications.

Original language	English
Article number	630
Journal	Translational psychiatry
Volume	11
Issue number	1
DOIs	https://doi.org/10.1038/s41398-021-01722-y
Publication status	Published - 13 Dec 2021

Access to Document

10.1038/s41398-021-01722-y

Cite this

Morgan, S. E., Diederen, K., Vértes, P. E., Ip, S. H. Y., Wang, B., Thompson, B., Demjaha, A., De Micheli, A., Oliver, D., Liakata, M., Fusar-Poli, P., Spencer, T. J., McGuire, P., & Morgan, S. (2021). Natural Language Processing markers in first episode psychosis and people at clinical high-risk. Translational psychiatry, 11(1), Article 630. https://doi.org/10.1038/s41398-021-01722-y

@article{75983907238540dd879d0ce7db0fe8cf,

title = "Natural Language Processing markers in first episode psychosis and people at clinical high-risk",

abstract = "Recent work has suggested that disorganised speech might be a powerful predictor of later psychotic illness in clinical high risk subjects. To that end, several automated measures to quantify disorganisation of transcribed speech have been proposed. However, it remains unclear which measures are most strongly associated with psychosis, how different measures are related to each other and what the best strategies are to collect speech data from participants. Here, we assessed whether twelve automated Natural Language Processing markers could differentiate transcribed speech excerpts from subjects at clinical high risk for psychosis, first episode psychosis patients and healthy control subjects (total N = 54). In-line with previous work, several measures showed significant differences between groups, including semantic coherence, speech graph connectivity and a measure of whether speech was on-topic, the latter of which outperformed the related measure of tangentiality. Most NLP measures examined were only weakly related to each other, suggesting they provide complementary information. Finally, we compared the ability of transcribed speech generated using different tasks to differentiate the groups. Speech generated from picture descriptions of the Thematic Apperception Test and a story re-telling task outperformed free speech, suggesting that choice of speech generation method may be an important consideration. Overall, quantitative speech markers represent a promising direction for future clinical applications.",

author = "Morgan, {Sarah E} and Kelly Diederen and V{\'e}rtes, {Petra E} and Ip, {Samantha H Y} and Bo Wang and Bethany Thompson and Arsime Demjaha and {De Micheli}, Andrea and Dominic Oliver and Maria Liakata and Paolo Fusar-Poli and Spencer, {Tom J} and Philip McGuire and Sarah Morgan",

note = "Funding Information: We thank the services users and volunteers who took part in this study, and the members of the Outreach and Support in South London (OASIS) team who were involved in the recruitment, management and clinical follow-up of the participants reported in this manuscript. We are also grateful to the experts by experience from the Cambridge and Peterborough Foundation Trust Service User and Carers Research Group, who gave constructive feedback on the manuscript. SEM was supported by the Accelerate Programme for Scientific Discovery, funded by Schmidt Futures, a Fellowship from The Alan Turing Institute, London, and a Henslow Fellowship at Lucy Cavendish College, University of Cambridge, funded by the Cambridge Philosophical Society. PEV is supported by a fellowship from MQ: Transforming Mental Health (MQF17_24). This work was supported by The Alan Turing Institute under the EPSRC grant EP/N510129/1, the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014), the UK Medical Research Council (MRC) and the National Institute for Health Research (NIHR) Mental Health Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King{\textquoteright}s College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, MRC or the Department of Health and Social Care. The funder had no influence on the design of the study or interpretation of the results. Publisher Copyright: {\textcopyright} 2021, The Author(s).",

year = "2021",

month = dec,

day = "13",

doi = "10.1038/s41398-021-01722-y",

language = "English",

volume = "11",

journal = "Translational psychiatry",

issn = "2158-3188",

publisher = "Nature Publishing Group",

number = "1",

}

TY - JOUR

T1 - Natural Language Processing markers in first episode psychosis and people at clinical high-risk

AU - Morgan, Sarah E

AU - Diederen, Kelly

AU - Vértes, Petra E

AU - Ip, Samantha H Y

AU - Wang, Bo

AU - Thompson, Bethany

AU - Demjaha, Arsime

AU - De Micheli, Andrea

AU - Oliver, Dominic

AU - Liakata, Maria

AU - Fusar-Poli, Paolo

AU - Spencer, Tom J

AU - McGuire, Philip

AU - Morgan, Sarah

N1 - Funding Information: We thank the services users and volunteers who took part in this study, and the members of the Outreach and Support in South London (OASIS) team who were involved in the recruitment, management and clinical follow-up of the participants reported in this manuscript. We are also grateful to the experts by experience from the Cambridge and Peterborough Foundation Trust Service User and Carers Research Group, who gave constructive feedback on the manuscript. SEM was supported by the Accelerate Programme for Scientific Discovery, funded by Schmidt Futures, a Fellowship from The Alan Turing Institute, London, and a Henslow Fellowship at Lucy Cavendish College, University of Cambridge, funded by the Cambridge Philosophical Society. PEV is supported by a fellowship from MQ: Transforming Mental Health (MQF17_24). This work was supported by The Alan Turing Institute under the EPSRC grant EP/N510129/1, the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014), the UK Medical Research Council (MRC) and the National Institute for Health Research (NIHR) Mental Health Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, MRC or the Department of Health and Social Care. The funder had no influence on the design of the study or interpretation of the results. Publisher Copyright: © 2021, The Author(s).

PY - 2021/12/13

Y1 - 2021/12/13

N2 - Recent work has suggested that disorganised speech might be a powerful predictor of later psychotic illness in clinical high risk subjects. To that end, several automated measures to quantify disorganisation of transcribed speech have been proposed. However, it remains unclear which measures are most strongly associated with psychosis, how different measures are related to each other and what the best strategies are to collect speech data from participants. Here, we assessed whether twelve automated Natural Language Processing markers could differentiate transcribed speech excerpts from subjects at clinical high risk for psychosis, first episode psychosis patients and healthy control subjects (total N = 54). In-line with previous work, several measures showed significant differences between groups, including semantic coherence, speech graph connectivity and a measure of whether speech was on-topic, the latter of which outperformed the related measure of tangentiality. Most NLP measures examined were only weakly related to each other, suggesting they provide complementary information. Finally, we compared the ability of transcribed speech generated using different tasks to differentiate the groups. Speech generated from picture descriptions of the Thematic Apperception Test and a story re-telling task outperformed free speech, suggesting that choice of speech generation method may be an important consideration. Overall, quantitative speech markers represent a promising direction for future clinical applications.

AB - Recent work has suggested that disorganised speech might be a powerful predictor of later psychotic illness in clinical high risk subjects. To that end, several automated measures to quantify disorganisation of transcribed speech have been proposed. However, it remains unclear which measures are most strongly associated with psychosis, how different measures are related to each other and what the best strategies are to collect speech data from participants. Here, we assessed whether twelve automated Natural Language Processing markers could differentiate transcribed speech excerpts from subjects at clinical high risk for psychosis, first episode psychosis patients and healthy control subjects (total N = 54). In-line with previous work, several measures showed significant differences between groups, including semantic coherence, speech graph connectivity and a measure of whether speech was on-topic, the latter of which outperformed the related measure of tangentiality. Most NLP measures examined were only weakly related to each other, suggesting they provide complementary information. Finally, we compared the ability of transcribed speech generated using different tasks to differentiate the groups. Speech generated from picture descriptions of the Thematic Apperception Test and a story re-telling task outperformed free speech, suggesting that choice of speech generation method may be an important consideration. Overall, quantitative speech markers represent a promising direction for future clinical applications.

UR - http://www.scopus.com/inward/record.url?scp=85118456194&partnerID=8YFLogxK

U2 - 10.1038/s41398-021-01722-y

DO - 10.1038/s41398-021-01722-y

M3 - Article

C2 - 34903724

SN - 2158-3188

VL - 11

JO - Translational psychiatry

JF - Translational psychiatry

IS - 1

M1 - 630

ER -

Natural Language Processing markers in first episode psychosis and people at clinical high-risk

Abstract

Access to Document

Other files and links

Fingerprint

Cite this