Deep learning for detection of radiographic sacroiliitis: achieving expert-level performance.
Artificial intelligence
Axial spondyloarthritis
Deep learning
Machine learning
Sacroiliitis
Journal
Arthritis research & therapy
ISSN: 1478-6362
Titre abrégé: Arthritis Res Ther
Pays: England
ID NLM: 101154438
Informations de publication
Date de publication:
08 04 2021
08 04 2021
Historique:
received:
11
02
2021
accepted:
22
03
2021
entrez:
9
4
2021
pubmed:
10
4
2021
medline:
22
6
2021
Statut:
epublish
Résumé
Radiographs of the sacroiliac joints are commonly used for the diagnosis and classification of axial spondyloarthritis. The aim of this study was to develop and validate an artificial neural network for the detection of definite radiographic sacroiliitis as a manifestation of axial spondyloarthritis (axSpA). Conventional radiographs of the sacroiliac joints obtained in two independent studies of patients with axSpA were used. The first cohort comprised 1553 radiographs and was split into training (n = 1324) and validation (n = 229) sets. The second cohort comprised 458 radiographs and was used as an independent test dataset. All radiographs were assessed in a central reading session, and the final decision on the presence or absence of definite radiographic sacroiliitis was used as a reference. The performance of the neural network was evaluated by calculating areas under the receiver operating characteristic curves (AUCs) as well as sensitivity and specificity. Cohen's kappa and the absolute agreement were used to assess the agreement between the neural network and the human readers. The neural network achieved an excellent performance in the detection of definite radiographic sacroiliitis with an AUC of 0.97 and 0.94 for the validation and test datasets, respectively. Sensitivity and specificity for the cut-off weighting both measurements equally were 88% and 95% for the validation and 92% and 81% for the test set. The Cohen's kappa between the neural network and the reference judgements were 0.79 and 0.72 for the validation and test sets with an absolute agreement of 90% and 88%, respectively. Deep artificial neural networks enable the accurate detection of definite radiographic sacroiliitis relevant for the diagnosis and classification of axSpA.
Sections du résumé
BACKGROUND
Radiographs of the sacroiliac joints are commonly used for the diagnosis and classification of axial spondyloarthritis. The aim of this study was to develop and validate an artificial neural network for the detection of definite radiographic sacroiliitis as a manifestation of axial spondyloarthritis (axSpA).
METHODS
Conventional radiographs of the sacroiliac joints obtained in two independent studies of patients with axSpA were used. The first cohort comprised 1553 radiographs and was split into training (n = 1324) and validation (n = 229) sets. The second cohort comprised 458 radiographs and was used as an independent test dataset. All radiographs were assessed in a central reading session, and the final decision on the presence or absence of definite radiographic sacroiliitis was used as a reference. The performance of the neural network was evaluated by calculating areas under the receiver operating characteristic curves (AUCs) as well as sensitivity and specificity. Cohen's kappa and the absolute agreement were used to assess the agreement between the neural network and the human readers.
RESULTS
The neural network achieved an excellent performance in the detection of definite radiographic sacroiliitis with an AUC of 0.97 and 0.94 for the validation and test datasets, respectively. Sensitivity and specificity for the cut-off weighting both measurements equally were 88% and 95% for the validation and 92% and 81% for the test set. The Cohen's kappa between the neural network and the reference judgements were 0.79 and 0.72 for the validation and test sets with an absolute agreement of 90% and 88%, respectively.
CONCLUSION
Deep artificial neural networks enable the accurate detection of definite radiographic sacroiliitis relevant for the diagnosis and classification of axSpA.
Identifiants
pubmed: 33832519
doi: 10.1186/s13075-021-02484-0
pii: 10.1186/s13075-021-02484-0
pmc: PMC8028815
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
106Références
Ann Rheum Dis. 2012 Dec;71(12):1998-2001
pubmed: 22915622
Ann Rheum Dis. 1987 Feb;46(2):139-45
pubmed: 3827336
Nature. 2017 Feb 2;542(7639):115-118
pubmed: 28117445
Lancet Digit Health. 2019 Oct;1(6):e271-e297
pubmed: 33323251
Ann Rheum Dis. 2011 Aug;70(8):1369-74
pubmed: 21622969
Nature. 2020 Jan;577(7788):89-94
pubmed: 31894144
Ann Rheum Dis. 2019 Nov;78(11):1545-1549
pubmed: 31362994
Arthritis Rheumatol. 2014 Sep;66(9):2403-11
pubmed: 24909765
IEEE Trans Med Imaging. 2016 May;35(5):1285-98
pubmed: 26886976
J Rheumatol. 2017 Jan;44(1):70-77
pubmed: 27744397
Korean J Radiol. 2019 Mar;20(3):405-410
pubmed: 30799571
Ann Rheum Dis. 2009 Jun;68(6):777-83
pubmed: 19297344
J Rheumatol. 2004 Jan;31(1):125-32
pubmed: 14705231
Arthritis Rheum. 2009 Mar;60(3):717-27
pubmed: 19248087
Ann Rheum Dis. 2015 Jul;74(7):1327-39
pubmed: 25837448
Bioinformatics. 2005 Oct 15;21(20):3940-1
pubmed: 16096348
Arthritis Rheum. 1984 Apr;27(4):361-8
pubmed: 6231933