A mobile-optimized artificial intelligence system for gestational age and fetal malpresentation assessment.
Health care
Medical research
Journal
Communications medicine
ISSN: 2730-664X
Titre abrégé: Commun Med (Lond)
Pays: England
ID NLM: 9918250414506676
Informations de publication
Date de publication:
2022
2022
Historique:
received:
10
02
2022
accepted:
28
09
2022
entrez:
17
10
2022
pubmed:
18
10
2022
medline:
18
10
2022
Statut:
epublish
Résumé
Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings. Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial intelligence (AI) models that used blind sweeps to predict gestational age (GA) and fetal malpresentation. AI GA estimates and standard fetal biometry estimates were compared to a previously established ground truth, and evaluated for difference in absolute error. Fetal malpresentation (non-cephalic vs cephalic) was compared to sonographer assessment. On-device AI model run-times were benchmarked on Android mobile phones. Here we show that GA estimation accuracy of the AI model is non-inferior to standard fetal biometry estimates (error difference -1.4 ± 4.5 days, 95% CI -1.8, -0.9, The gestational age model is non-inferior to the clinical standard and the fetal malpresentation model has high AUC-ROCs across operators and devices. Our AI models are able to run on-device, without internet connectivity, and provide feedback scores to assist in upleveling the capabilities of lightly trained ultrasound operators in low resource settings.
Sections du résumé
Background
UNASSIGNED
Fetal ultrasound is an important component of antenatal care, but shortage of adequately trained healthcare workers has limited its adoption in low-to-middle-income countries. This study investigated the use of artificial intelligence for fetal ultrasound in under-resourced settings.
Methods
UNASSIGNED
Blind sweep ultrasounds, consisting of six freehand ultrasound sweeps, were collected by sonographers in the USA and Zambia, and novice operators in Zambia. We developed artificial intelligence (AI) models that used blind sweeps to predict gestational age (GA) and fetal malpresentation. AI GA estimates and standard fetal biometry estimates were compared to a previously established ground truth, and evaluated for difference in absolute error. Fetal malpresentation (non-cephalic vs cephalic) was compared to sonographer assessment. On-device AI model run-times were benchmarked on Android mobile phones.
Results
UNASSIGNED
Here we show that GA estimation accuracy of the AI model is non-inferior to standard fetal biometry estimates (error difference -1.4 ± 4.5 days, 95% CI -1.8, -0.9,
Conclusions
UNASSIGNED
The gestational age model is non-inferior to the clinical standard and the fetal malpresentation model has high AUC-ROCs across operators and devices. Our AI models are able to run on-device, without internet connectivity, and provide feedback scores to assist in upleveling the capabilities of lightly trained ultrasound operators in low resource settings.
Identifiants
pubmed: 36249461
doi: 10.1038/s43856-022-00194-5
pii: 194
pmc: PMC9553916
doi:
Types de publication
Journal Article
Langues
eng
Pagination
128Subventions
Organisme : Bill & Melinda Gates Foundation
ID : INV-003266
Pays : United States
Informations de copyright
© The Author(s) 2022.
Déclaration de conflit d'intérêts
Competing interestsThe authors declare the following competing interests: this study was partially funded by Google Inc. R.G.G., C. Lee, A.W., M.S., J.A.T., S.M.M., C.C., S.S., D.T., A.U., K.C., J.G., G.E.D., T. Sp., T. Sa., K.L., T.T., G.C., L.P., J.W., and R.P. are employees of Google Inc. and own stock as part of the standard employee compensation package. The remaining authors declare no competing interests.
Références
Curr Probl Diagn Radiol. 2017 May - Jun;46(3):210-215
pubmed: 28057388
Reprod Health. 2018 Jul 20;15(1):129
pubmed: 30029609
Am J Obstet Gynecol. 1992 Mar;166(3):851-2
pubmed: 1550152
Commun Med (Lond). 2022 Oct 11;2:128
pubmed: 36249461
Radiology. 1984 Aug;152(2):497-501
pubmed: 6739822
Acta Obstet Gynecol Scand. 2019 Mar;98(3):300-308
pubmed: 30414270
J Perinatol. 2014 Jul;34(7):508-12
pubmed: 24699218
BJOG. 2017 May;124(6):860-862
pubmed: 28190290
BMC Public Health. 2021 May 20;21(1):952
pubmed: 34016085
Ultrasound Obstet Gynecol. 2016 Dec;48(6):719-726
pubmed: 26924421
Ultrasound Med Biol. 2018 Nov;44(11):2250-2260
pubmed: 30093339
NEJM Evid. 2022 May;1(5):
pubmed: 36875289
Obstet Gynecol. 2017 May;129(5):e150-e154
pubmed: 28426621
J Med Imaging (Bellingham). 2020 Jan;7(1):014501
pubmed: 31956665
Acta Obstet Gynecol Scand. 2006;85(7):797-804
pubmed: 16817076
Glob Health Sci Pract. 2016 Dec 28;4(4):675-683
pubmed: 28031304
Am J Obstet Gynecol. 2012 Oct;207(4):285.e1-9
pubmed: 23021690
Acta Obstet Gynecol Scand. 2007;86(2):240-3
pubmed: 17364290
Ultrasound Med Biol. 2019 Mar;45(3):773-785
pubmed: 30573305
Reprod Health. 2020 Dec 17;17(Suppl 3):179
pubmed: 33334337
Am J Perinatol. 2016 Jan;33(1):90-8
pubmed: 26238329
Lancet Glob Health. 2020 Apr;8(4):e545-e554
pubmed: 32199122
Ultrasound Q. 2014 Dec;30(4):262-6
pubmed: 25415862
Am J Obstet Gynecol. 2002 Dec;187(6):1694-8
pubmed: 12501085