Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models

Edward Campbell; Judith Dineley; Pauline Conde; Faith Matcham; Katie White; Carolin Oetzmann; Sara Simblett; Stuart Bruce; Amos Folarin; Til Wykes; Srinivasan  Vairavan; Richard Dobson; Laura  Docio-Fernandez; Carmen  Garcia-Mateo; Vaibhav A Narayan; Matthew Hotopf; Nicholas Cummins; on behalf of the RADAR CNS Consortium

doi:10.21437/Interspeech.2023-1721

Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models

Edward Campbell, Judith Dineley, Pauline Conde, Faith Matcham, Katie White, Carolin Oetzmann, Sara Simblett, Stuart Bruce, Amos Folarin, Til Wykes, Srinivasan Vairavan, Richard Dobson, Laura Docio-Fernandez, Carmen Garcia-Mateo, Vaibhav A Narayan, Matthew Hotopf, Nicholas Cummins, on behalf of the RADAR CNS Consortium

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

7 Citations (Scopus)

Abstract

There is an urgent need for new methods that improve the management and treatment of Major Depressive Disorder (MDD). Speech has long been regarded as a promising digital marker in this regard, with many works highlighting that speech changes associated with MDD can be captured through machine learning models. Typically, findings are based on cross-sectional data, with little work exploring the advantages of personalization in building more robust and reliable models. This work assesses the strengths of different combinations of speech representations and machine learning models, in personalized and generalized settings in a two-class depression severity classification paradigm. Key results on a longitudinal dataset highlight the benefits of personalization. Our strongest performing model set-up utilized self-supervised learning features and convolutional neural network (CNN) and long short-term memory (LSTM) back-end.

Original language	English
Title of host publication	Proc. INTERSPEECH 2023
Publisher	ISCA-INST SPEECH COMMUNICATION ASSOC
Pages	1738-1742
Number of pages	5
Volume	2023-August
DOIs	https://doi.org/10.21437/Interspeech.2023-1721
Publication status	Published - 20 Aug 2023

Publication series

Name	Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
ISSN (Print)	2308-457X

Keywords

Major depressive disorder
personalization
self- supervised learning
remote monitoring technologies

Access to Document

10.21437/Interspeech.2023-1721

Cite this

Campbell, E., Dineley, J., Conde, P., Matcham, F., White, K., Oetzmann, C., Simblett, S., Bruce, S., Folarin, A., Wykes, T., Vairavan, S., Dobson, R., Docio-Fernandez, L., Garcia-Mateo, C., Narayan, V. A., Hotopf, M., Cummins, N., & Consortium, O. B. O. T. RADAR. CNS. (2023). Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models. In Proc. INTERSPEECH 2023 (Vol. 2023-August, pp. 1738-1742). (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH). ISCA-INST SPEECH COMMUNICATION ASSOC. https://doi.org/10.21437/Interspeech.2023-1721

Campbell, Edward ; Dineley, Judith ; Conde, Pauline et al. / Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models. Proc. INTERSPEECH 2023. Vol. 2023-August ISCA-INST SPEECH COMMUNICATION ASSOC, 2023. pp. 1738-1742 (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH).

@inbook{ce58383d2f514f3d90e109266cda009e,

title = "Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models",

abstract = "There is an urgent need for new methods that improve the management and treatment of Major Depressive Disorder (MDD). Speech has long been regarded as a promising digital marker in this regard, with many works highlighting that speech changes associated with MDD can be captured through machine learning models. Typically, findings are based on cross-sectional data, with little work exploring the advantages of personalization in building more robust and reliable models. This work assesses the strengths of different combinations of speech representations and machine learning models, in personalized and generalized settings in a two-class depression severity classification paradigm. Key results on a longitudinal dataset highlight the benefits of personalization. Our strongest performing model set-up utilized self-supervised learning features and convolutional neural network (CNN) and long short-term memory (LSTM) back-end.",

keywords = "Major depressive disorder, personalization, self- supervised learning, remote monitoring technologies",

author = "Edward Campbell and Judith Dineley and Pauline Conde and Faith Matcham and Katie White and Carolin Oetzmann and Sara Simblett and Stuart Bruce and Amos Folarin and Til Wykes and Srinivasan Vairavan and Richard Dobson and Laura Docio-Fernandez and Carmen Garcia-Mateo and Narayan, {Vaibhav A} and Matthew Hotopf and Nicholas Cummins and Consortium, {on behalf of the RADAR CNS}",

note = "Funding Information: Funding The RADAR-CNS project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 115902. This Joint Undertaking receives support from the European Union's Horizon 2020 research and innovation programme and EFPIA (www.imi.europa.eu). This communication reflects the views of the RADAR-CNS consortium and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. The funding body has not been involved in the design of the study, the collection or analysis of data, or the interpretation of data. We thank our colleagues both within the RADAR-CNS consortium and across all involved institutions for their contribution to the development of this protocol. We thank all the members of the RADAR-CNS patient advisory board for their contribution to the device selection procedures, and their invaluable advice throughout the study protocol design. This paper also represents independent research part funded by the National Institute for Health Research (NIHR) Maudsley 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 or the Department of Health and Social Care. C.O. is supported by the UK Medical Research Council (MR/N013700/1) and King's College London member of the MRC Doctoral Training Partnership in Biomedical Sciences This work has also received financial support from Axudas propias para a mobilidade de Persoal Investigador da Unversidade de Vigo 2021, the Xunta de Galicia (Centro singular de investigaci{\'o}n de Galicia accreditation 2019-2022), Conseller{\'i}a de Cultura (Educaci{\'o}n e Ordenaci{\'o}n Universitaria; axudas para a consolidaci{\'o}n e estruturaci{\'o}n de unidades de investigaci{\'o}n competitivas do Sistema Universitario de Galicia -ED431B 2021/24), and the European Union (European Regional Development Fund - ERDF). Funding Information: Funding The RADAR-CNS project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 115902. This Joint Undertaking receives support from the European Union{\textquoteright}s Horizon 2020 research and innovation programme and EFPIA (www.imi.europa.eu). This communication reflects the views of the RADAR-CNS consortium and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. The funding body has not been involved in the design of the study, the collection or analysis of data, or the interpretation of data. We thank our colleagues both within the RADAR-CNS consortium and across all involved institutions for their contribution to the development of this protocol. We thank all the members of the RADAR-CNS patient advisory board for their contribution to the device selection procedures, and their invaluable advice throughout the study protocol design. This paper also represents independent research part funded by the National Institute for Health Research (NIHR) Maudsley 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 or the Department of Health and Social Care. C.O. is supported by the UK Medical Research Council (MR/N013700/1) and King{\textquoteright}s College London member of the MRC Doctoral Training Partnership in Biomedical Sciences This work has also received financial support from Axudas propias para a mobilidade de Persoal Investigador da Unversidade de Vigo 2021, the Xunta de Galicia (Centro singular de investigaci{\'o}n de Galicia accreditation 2019-2022), Con-seller{\'i}a de Cultura (Educaci{\'o}n e Ordenaci{\'o}n Universitaria; ax-udas para a consolidaci{\'o}n e estruturaci{\'o}n de unidades de in-vestigaci{\'o}n competitivas do Sistema Universitario de Galicia -ED431B 2021/24), and the European Union (European Regional Development Fund - ERDF). Publisher Copyright: {\textcopyright} 2023 International Speech Communication Association. All rights reserved.",

year = "2023",

month = aug,

day = "20",

doi = "10.21437/Interspeech.2023-1721",

language = "English",

volume = "2023-August",

series = "Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",

publisher = "ISCA-INST SPEECH COMMUNICATION ASSOC",

pages = "1738--1742",

booktitle = "Proc. INTERSPEECH 2023",

}

Campbell, E, Dineley, J , Conde, P, Matcham, F, White, K , Oetzmann, C , Simblett, S, Bruce, S, Folarin, A , Wykes, T, Vairavan, S, Dobson, R, Docio-Fernandez, L, Garcia-Mateo, C, Narayan, VA, Hotopf, M , Cummins, N & Consortium, OBOTRADARCNS 2023, Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models. in Proc. INTERSPEECH 2023. vol. 2023-August, Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, ISCA-INST SPEECH COMMUNICATION ASSOC, pp. 1738-1742. https://doi.org/10.21437/Interspeech.2023-1721

Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models. / Campbell, Edward; Dineley, Judith ; Conde, Pauline et al.
Proc. INTERSPEECH 2023. Vol. 2023-August ISCA-INST SPEECH COMMUNICATION ASSOC, 2023. p. 1738-1742 (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH).

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

TY - CHAP

T1 - Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models

AU - Campbell, Edward

AU - Dineley, Judith

AU - Conde, Pauline

AU - Matcham, Faith

AU - White, Katie

AU - Oetzmann, Carolin

AU - Simblett, Sara

AU - Bruce, Stuart

AU - Folarin, Amos

AU - Wykes, Til

AU - Vairavan, Srinivasan

AU - Dobson, Richard

AU - Docio-Fernandez, Laura

AU - Garcia-Mateo, Carmen

AU - Narayan, Vaibhav A

AU - Hotopf, Matthew

AU - Cummins, Nicholas

AU - Consortium, on behalf of the RADAR CNS

N1 - Funding Information: Funding The RADAR-CNS project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 115902. This Joint Undertaking receives support from the European Union's Horizon 2020 research and innovation programme and EFPIA (www.imi.europa.eu). This communication reflects the views of the RADAR-CNS consortium and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. The funding body has not been involved in the design of the study, the collection or analysis of data, or the interpretation of data. We thank our colleagues both within the RADAR-CNS consortium and across all involved institutions for their contribution to the development of this protocol. We thank all the members of the RADAR-CNS patient advisory board for their contribution to the device selection procedures, and their invaluable advice throughout the study protocol design. This paper also represents independent research part funded by the National Institute for Health Research (NIHR) Maudsley 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 or the Department of Health and Social Care. C.O. is supported by the UK Medical Research Council (MR/N013700/1) and King's College London member of the MRC Doctoral Training Partnership in Biomedical Sciences This work has also received financial support from Axudas propias para a mobilidade de Persoal Investigador da Unversidade de Vigo 2021, the Xunta de Galicia (Centro singular de investigación de Galicia accreditation 2019-2022), Consellería de Cultura (Educación e Ordenación Universitaria; axudas para a consolidación e estruturación de unidades de investigación competitivas do Sistema Universitario de Galicia -ED431B 2021/24), and the European Union (European Regional Development Fund - ERDF). Funding Information: Funding The RADAR-CNS project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 115902. This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA (www.imi.europa.eu). This communication reflects the views of the RADAR-CNS consortium and neither IMI nor the European Union and EFPIA are liable for any use that may be made of the information contained herein. The funding body has not been involved in the design of the study, the collection or analysis of data, or the interpretation of data. We thank our colleagues both within the RADAR-CNS consortium and across all involved institutions for their contribution to the development of this protocol. We thank all the members of the RADAR-CNS patient advisory board for their contribution to the device selection procedures, and their invaluable advice throughout the study protocol design. This paper also represents independent research part funded by the National Institute for Health Research (NIHR) Maudsley 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 or the Department of Health and Social Care. C.O. is supported by the UK Medical Research Council (MR/N013700/1) and King’s College London member of the MRC Doctoral Training Partnership in Biomedical Sciences This work has also received financial support from Axudas propias para a mobilidade de Persoal Investigador da Unversidade de Vigo 2021, the Xunta de Galicia (Centro singular de investigación de Galicia accreditation 2019-2022), Con-sellería de Cultura (Educación e Ordenación Universitaria; ax-udas para a consolidación e estruturación de unidades de in-vestigación competitivas do Sistema Universitario de Galicia -ED431B 2021/24), and the European Union (European Regional Development Fund - ERDF). Publisher Copyright: © 2023 International Speech Communication Association. All rights reserved.

PY - 2023/8/20

Y1 - 2023/8/20

N2 - There is an urgent need for new methods that improve the management and treatment of Major Depressive Disorder (MDD). Speech has long been regarded as a promising digital marker in this regard, with many works highlighting that speech changes associated with MDD can be captured through machine learning models. Typically, findings are based on cross-sectional data, with little work exploring the advantages of personalization in building more robust and reliable models. This work assesses the strengths of different combinations of speech representations and machine learning models, in personalized and generalized settings in a two-class depression severity classification paradigm. Key results on a longitudinal dataset highlight the benefits of personalization. Our strongest performing model set-up utilized self-supervised learning features and convolutional neural network (CNN) and long short-term memory (LSTM) back-end.

AB - There is an urgent need for new methods that improve the management and treatment of Major Depressive Disorder (MDD). Speech has long been regarded as a promising digital marker in this regard, with many works highlighting that speech changes associated with MDD can be captured through machine learning models. Typically, findings are based on cross-sectional data, with little work exploring the advantages of personalization in building more robust and reliable models. This work assesses the strengths of different combinations of speech representations and machine learning models, in personalized and generalized settings in a two-class depression severity classification paradigm. Key results on a longitudinal dataset highlight the benefits of personalization. Our strongest performing model set-up utilized self-supervised learning features and convolutional neural network (CNN) and long short-term memory (LSTM) back-end.

KW - Major depressive disorder

KW - personalization

KW - self- supervised learning

KW - remote monitoring technologies

UR - http://www.scopus.com/inward/record.url?scp=85171562208&partnerID=8YFLogxK

U2 - 10.21437/Interspeech.2023-1721

DO - 10.21437/Interspeech.2023-1721

M3 - Conference paper

VL - 2023-August

T3 - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

SP - 1738

EP - 1742

BT - Proc. INTERSPEECH 2023

PB - ISCA-INST SPEECH COMMUNICATION ASSOC

ER -

Campbell E, Dineley J , Conde P, Matcham F, White K , Oetzmann C et al. Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models. In Proc. INTERSPEECH 2023. Vol. 2023-August. ISCA-INST SPEECH COMMUNICATION ASSOC. 2023. p. 1738-1742. (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH). doi: 10.21437/Interspeech.2023-1721

Classifying depression symptom severity: Assessment of speech representations in personalized and generalized machine learning models

Abstract

Publication series

Keywords

Access to Document

Other files and links

Fingerprint

Cite this