A retrospective study on machine learning-assisted stroke recognition for medical helpline calls

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningfagfællebedømt

Dokumenter

  • Fulltext

    Forlagets udgivne version, 1,44 MB, PDF-dokument

Advanced stroke treatment is time-dependent and, therefore, relies on recognition by call-takers at prehospital telehealth services to ensure fast hospitalisation. This study aims to develop and assess the potential of machine learning in improving prehospital stroke recognition during medical helpline calls. We used calls from 1 January 2015 to 31 December 2020 in Copenhagen to develop a machine learning-based classification pipeline. Calls from 2021 are used for testing. Calls are first transcribed using an automatic speech recognition model and then categorised as stroke or non-stroke using a text classification model. Call-takers achieve a sensitivity of 52.7% (95% confidence interval 49.2–56.4%) with a positive predictive value (PPV) of 17.1% (15.5–18.6%). The machine learning framework performs significantly better (p < 0.0001) with a sensitivity of 63.0% (62.0–64.1%) and a PPV of 24.9% (24.3–25.5%). Thus, a machine learning framework for recognising stroke in prehospital medical helpline calls may become a supportive tool for call-takers, aiding in early and accurate stroke recognition.
OriginalsprogEngelsk
Artikelnummer235
Tidsskriftnpj Digital Medicine
Vol/bind6
Udgave nummer1
Antal sider8
DOI
StatusUdgivet - 2023

Bibliografisk note

Funding Information:
J.D.H. and L.B. received funding from the Innovation Fund Denmark. J.D.H. and L.B. used Corti and held stock warrants. L.M. is a co-founder, stockholder, and the Chief Technology Officer of Corti. J.W. received funding from Trygfonden. S.N.F.B. has no conflicts of interest to declare. H.C. has received funding from the Velux Foundation, Tværsfonden, Helsefonden, Hartmann Fonden, Lundbeck Foundation, and Novo Nordisk Foundation; royalties from Gyldendal; honoraria from Bayer and Bristol Meyers Squibb, and is chair of Action Plan for stroke in Europe Implementation, Co-chair of the Scientific Stroke Panel EAN and Senior Guest Editor of AHA Stroke. M.S. has no conflicts of interest to declare. H.C.C. has no conflicts of interest to declare. C.K. received funding from the Novo Nordisk Foundation and is the chair of the Danish Resuscitation Council and vice chair of the Danish Stroke Society. Both positions are unpaid.

Funding Information:
We thank the staff of the CEMS for their role in generating the data used in this study. We thank Emilie Grunddal Pedersen, Mette Bjerg Lindhøj, and Jens Morten Haugård for their help and cooperation in accessing the data sources. We also thank the Centre for IT and Medical Technology (CIMT) and Corti employees Akihiro Inui and Nathaniel Joselson for their assistance in setting up and using the cloud-computing environment for training and evaluating the machine learning framework of this study. Funding for the work was received from Innovation Fund Denmark, Trygfonden, Copenhagen University Hospital—Herlev, Gentofte, and the University of Copenhagen. The grant providers had no role in the study design, data collection, analysis, interpretation, manuscript writing, or publication decision. Corti provided additional funds and technical expertise to develop the models. Corti was not financially compensated for this, and the project was part of its research initiatives, which were conducted in cooperation with several universities and the Innovation Fund Denmark. Corti owns the rights to the models and source code.

Funding Information:
We thank the staff of the CEMS for their role in generating the data used in this study. We thank Emilie Grunddal Pedersen, Mette Bjerg Lindhøj, and Jens Morten Haugård for their help and cooperation in accessing the data sources. We also thank the Centre for IT and Medical Technology (CIMT) and Corti employees Akihiro Inui and Nathaniel Joselson for their assistance in setting up and using the cloud-computing environment for training and evaluating the machine learning framework of this study. Funding for the work was received from Innovation Fund Denmark, Trygfonden, Copenhagen University Hospital—Herlev, Gentofte, and the University of Copenhagen. The grant providers had no role in the study design, data collection, analysis, interpretation, manuscript writing, or publication decision. Corti provided additional funds and technical expertise to develop the models. Corti was not financially compensated for this, and the project was part of its research initiatives, which were conducted in cooperation with several universities and the Innovation Fund Denmark. Corti owns the rights to the models and source code.

Publisher Copyright:
© 2023, The Author(s).

ID: 377807167