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

Jonathan Wenstrup; Jakob Drachmann Havtorn; Lasse Borgholt; Stig Nikolaj Blomberg; Lars Maaloe; Michael R. Sayre; Hanne Christensen; Christina Kruuse; Helle Collatz Christensen

doi:10.1038/s41746-023-00980-y

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

Jonathan Wenstrup, Jakob Drachmann Havtorn, Lasse Borgholt, Stig Nikolaj Blomberg, Lars Maaloe, Michael R. Sayre, Hanne Christensen, Christina Kruuse, Helle Collatz Christensen

Research output: Contribution to journal › Journal article › Research › peer-review

39 Downloads (Pure)

Abstract

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

Original language	English
Article number	235
Journal	npj Digital Medicine
Volume	6
Number of pages	8
DOIs	https://doi.org/10.1038/s41746-023-00980-y
Publication status	Published - 2023

Access to Document

10.1038/s41746-023-00980-y

FulltextFinal published version, 1.44 MB

OpenUrl availability

Full text

Cite this

@article{5bb503c8501a41a2980494ac80c6eb02,

title = "A retrospective study on machine learning-assisted stroke recognition for medical helpline calls",

abstract = "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 ",

author = "Jonathan Wenstrup and Havtorn, {Jakob Drachmann} and Lasse Borgholt and Blomberg, {Stig Nikolaj} and Lars Maaloe and Sayre, {Michael R.} and Hanne Christensen and Christina Kruuse and Christensen, {Helle Collatz}",

year = "2023",

doi = "10.1038/s41746-023-00980-y",

language = "English",

volume = "6",

journal = "npj Digital Medicine",

issn = "2398-6352",

publisher = "Nature Research",

}

TY - JOUR

T1 - A retrospective study on machine learning-assisted stroke recognition for medical helpline calls

AU - Wenstrup, Jonathan

AU - Havtorn, Jakob Drachmann

AU - Borgholt, Lasse

AU - Blomberg, Stig Nikolaj

AU - Maaloe, Lars

AU - Sayre, Michael R.

AU - Christensen, Hanne

AU - Kruuse, Christina

AU - Christensen, Helle Collatz

PY - 2023

Y1 - 2023

N2 - 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

AB - 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

U2 - 10.1038/s41746-023-00980-y

DO - 10.1038/s41746-023-00980-y

M3 - Journal article

C2 - 38114611

SN - 2398-6352

VL - 6

JO - npj Digital Medicine

JF - npj Digital Medicine

M1 - 235

ER -

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

Abstract

Access to Document

OpenUrl availability

Fingerprint

Cite this