Artificial intelligence for nailfold capillaroscopy analyses - a proof of concept application in juvenile dermatomyositis.
Journal
Pediatric research
ISSN: 1530-0447
Titre abrégé: Pediatr Res
Pays: United States
ID NLM: 0100714
Informations de publication
Date de publication:
22 Nov 2023
22 Nov 2023
Historique:
received:
25
05
2023
accepted:
25
10
2023
revised:
10
10
2023
medline:
23
11
2023
pubmed:
23
11
2023
entrez:
22
11
2023
Statut:
aheadofprint
Résumé
Biomarkers for idiopathic inflammatory myopathies are difficult to identify and may involve expensive laboratory tests. We assess the potential for artificial intelligence (AI) to differentiate children with juvenile dermatomyositis (JDM) from healthy controls using nailfold capillaroscopy (NFC) images. We also assessed the potential of NFC images to reflect the range of disease activity with JDM. A total of 1,120 NFC images from 111 children with active JDM, diagnosed between 1990 and 2020, and 321 NFC images from 31 healthy controls were retrieved from the CureJM JDM Registry. We built a lightweight and explainable deep neural network model called NFC-Net. Images were downscaled by interpolation techniques to reduce the computational cost. NFC-Net achieved high performance in differentiating patients with JDM from controls, with an area under the ROC curve (AUROC) of 0.93 (0.84, 0.99) and accuracy of 0.91 (0.82, 0.92). With sensitivity (0.85) and specificity (0.90) resulted in model precision of 0.95. The AUROC and accuracy for predicting clinical disease activity from inactivity were 0.75 (0.61, 0.81) and 0.74 (0.65, 0.79). The good performance of the NFC-Net demonstrates that NFC images are sufficient for detecting often unrecognized JDM disease activity, providing a reliable indicator of disease status. Proposed NFC-Net can accurately predict children with JDM from healthy controls using nailfold capillaroscopy (NFC) images. Additionally, it predicts the scores to JDM disease activity versus no activity. Equipped with gradients, NFC-Net is explainable and gives visual information beside the reported accuracies. NFC-Net is computationally efficient since it is applied to substantially downscaled NFC images. Furthermore, the model can be wrapped within an edge-based device like a mobile application that is accessible to both clinicians and patients.
Sections du résumé
BACKGROUND
BACKGROUND
Biomarkers for idiopathic inflammatory myopathies are difficult to identify and may involve expensive laboratory tests. We assess the potential for artificial intelligence (AI) to differentiate children with juvenile dermatomyositis (JDM) from healthy controls using nailfold capillaroscopy (NFC) images. We also assessed the potential of NFC images to reflect the range of disease activity with JDM.
METHODS
METHODS
A total of 1,120 NFC images from 111 children with active JDM, diagnosed between 1990 and 2020, and 321 NFC images from 31 healthy controls were retrieved from the CureJM JDM Registry. We built a lightweight and explainable deep neural network model called NFC-Net. Images were downscaled by interpolation techniques to reduce the computational cost.
RESULTS
RESULTS
NFC-Net achieved high performance in differentiating patients with JDM from controls, with an area under the ROC curve (AUROC) of 0.93 (0.84, 0.99) and accuracy of 0.91 (0.82, 0.92). With sensitivity (0.85) and specificity (0.90) resulted in model precision of 0.95. The AUROC and accuracy for predicting clinical disease activity from inactivity were 0.75 (0.61, 0.81) and 0.74 (0.65, 0.79).
CONCLUSION
CONCLUSIONS
The good performance of the NFC-Net demonstrates that NFC images are sufficient for detecting often unrecognized JDM disease activity, providing a reliable indicator of disease status.
IMPACT
CONCLUSIONS
Proposed NFC-Net can accurately predict children with JDM from healthy controls using nailfold capillaroscopy (NFC) images. Additionally, it predicts the scores to JDM disease activity versus no activity. Equipped with gradients, NFC-Net is explainable and gives visual information beside the reported accuracies. NFC-Net is computationally efficient since it is applied to substantially downscaled NFC images. Furthermore, the model can be wrapped within an edge-based device like a mobile application that is accessible to both clinicians and patients.
Identifiants
pubmed: 37993641
doi: 10.1038/s41390-023-02894-7
pii: 10.1038/s41390-023-02894-7
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Informations de copyright
© 2023. The Author(s), under exclusive licence to the International Pediatric Research Foundation, Inc.
Références
Feldman, B. M., Rider, L. G., Reed, A. M. & Pachman, L. M. Juvenile dermatomyositis and other idiopathic inflammatory myopathies of childhood. Lancet 371, 2201–2212 (2008).
doi: 10.1016/S0140-6736(08)60955-1
pubmed: 18586175
Pachman, L. M. Juvenile dermatomyositis (JDMS): new clues to diagnosis and pathogenesis. Clin. Exp. Rheumatol. 12, S69–S73 (1994).
pubmed: 7955631
Pachman, L. M., Nolan, B. E., DeRanieri, D. & Khojah, A. M. Juvenile Dermatomyositis: New Clues to Diagnosis and Therapy. Curr. Treatm. Opt. Rheumatol. 7, 1–24 (2021).
Pachman, L. M. et al. Juvenile dermatomyositis at diagnosis: clinical characteristics of 79 children. J. Rheumatol. 25, 1198–1204 (1998).
pubmed: 9632086
Rider, L. G. et al. Validation of manual muscle testing and a subset of eight muscles for adult and juvenile idiopathic inflammatory myopathies. Arthritis Care Res. 62, 465–472 (2010).
doi: 10.1002/acr.20035
McCann, L. J. et al. Development of a consensus core dataset in juvenile dermatomyositis for clinical use to inform research. Ann. Rheum. Dis. 77, 241–250 (2018).
doi: 10.1136/annrheumdis-2017-212141
pubmed: 29084729
Butbul Aviel, Y. et al. Sleep and fatigue and the relationship to pain, disease activity and quality of life in juvenile idiopathic arthritis and juvenile dermatomyositis. Rheumatology 50, 2051–2060 (2011).
doi: 10.1093/rheumatology/ker256
pubmed: 21873265
Ruperto, N. et al. Preliminary core sets of measures for disease activity and damage assessment in juvenile systemic lupus erythematosus and juvenile dermatomyositis. Rheumatology 42, 1452–1459 (2003).
doi: 10.1093/rheumatology/keg403
pubmed: 12832713
Bode, R. K., Klein-Gitelman, M. S., Miller, M. L., Lechman, T. S. & Pachman, L. M. Disease activity score for children with juvenile dermatomyositis: reliability and validity evidence. Arthritis Care Res.: Off. J. Am. Coll. Rheumatol. 49, 7–15 (2003).
doi: 10.1002/art.10924
Szabo, N. et al. Functional and morphological evaluation of hand microcirculation with nailfold capillaroscopy and laser Doppler imaging in Raynaud’s and Sjögren’s syndrome and poly/dermatomyositis. Scand. J. Rheumatol. 37, 23–29 (2008).
doi: 10.1080/03009740701640209
pubmed: 18189191
Cutolo, M., Sulli, A., Pizzorni, C. & Accardo, S. Nailfold videocapillaroscopy assessment of microvascular damage in systemic sclerosis. J. Rheumatol. 27, 155–160 (2000).
pubmed: 10648032
Ingegnoli, F. & Herrick, A. L. Nailfold capillaroscopy in pediatrics. Arthritis Care Res. 65, 1393–1400 (2013).
doi: 10.1002/acr.22026
Maricq, H. R. & LeRoy, C. E. Patterns of finger capillary abnormalities in connective tissue disease by “wide-field” microscopy. Arthritis Rheum. 16, 619–628 (1973).
doi: 10.1002/art.1780160506
pubmed: 4742842
Kurowski, J. A. et al. Nailfold Capillaroscopy as a Biomarker in the Evaluation of Pediatric Inflammatory Bowel Disease. Crohns Colitis 360, 3, 1–5 (2021).
Smith, R. L., Sundberg, J., Shamiyah, E., Dyer, A. & Pachman, L. M. Skin involvement in juvenile dermatomyositis is associated with loss of end row nailfold capillary loops. J. Rheumatol. 31, 1644–1649 (2004).
pubmed: 15290747
Ostrowski, R. A., Sullivan, C. L., Seshadri, R., Morgan, G. A. & Pachman, L. M. Association of normal nailfold end row loop numbers with a shorter duration of untreated disease in children with juvenile dermatomyositis. Arthritis Rheum. 62, 1533–1538 (2010).
doi: 10.1002/art.27379
pubmed: 20213809
pmcid: 3304541
Tawalbeh, S. M. et al. Serum protein biomarkers for juvenile dermatomyositis: a pilot study. BMC Rheumatol. 4, 1–15 (2020).
doi: 10.1186/s41927-020-00150-7
Wienke, J. et al. Galectin-9 and CXCL10 as biomarkers for disease activity in juvenile dermatomyositis: a longitudinal cohort study and multicohort validation. Arthritis Rheumatol. 71, 1377–1390 (2019).
doi: 10.1002/art.40881
pubmed: 30861625
pmcid: 6973145
Martin, N. et al. A national registry for juvenile dermatomyositis and other paediatric idiopathic inflammatory myopathies: 10 years’ experience; the Juvenile Dermatomyositis National (UK and Ireland) Cohort Biomarker Study and Repository for Idiopathic Inflammatory Myopathies. Rheumatology 50, 137–145 (2011).
doi: 10.1093/rheumatology/keq261
pubmed: 20823094
Huang, B. et al. Long-term follow-up of Janus-kinase inhibitor and novel active disease biomarker in juvenile dermatomyositis. Rheumatology https://doi.org/10.1093/rheumatology/keac399 . (2022)
Conklin, L. S. et al. Serum biomarkers of glucocorticoid response and safety in anti-neutrophil cytoplasmic antibody-associated vasculitis and juvenile dermatomyositis. Steroids 140, 159–166 (2018).
doi: 10.1016/j.steroids.2018.10.008
pubmed: 30352204
pmcid: 6640634
Deakin, C. T. et al. Association with HLA-DRβ1 position 37 distinguishes juvenile dermatomyositis from adult-onset myositis. Hum. Mol. Genet 31, 2471–2481 (2022).
doi: 10.1093/hmg/ddac019
pubmed: 35094092
pmcid: 9307311
Wienke, J. et al. Endothelial and inflammatory biomarker profile at diagnosis reflects clinical heterogeneity of juvenile dermatomyositis and is prognostic for response to treatment in two independent cohorts. Arthritis Rheumatol. 72, 1214 (2020).
doi: 10.1002/art.41236
pubmed: 32103637
pmcid: 7329617
Khojah, A., Morgan, G. & Pachman, L. M. Clues to disease activity in juvenile Dermatomyositis: Neopterin and other biomarkers. Diagnostics 12, 8 (2021).
doi: 10.3390/diagnostics12010008
pubmed: 35054173
pmcid: 8774344
Pascal Getreuer. Linear methods for image interpolation. Image Process. Line 1, 238–259 (2011).
doi: 10.5201/ipol.2011.g_lmii
Lian, X. & Liu, J. Revisit batch normalization: new understanding and refinement via composition optimization. Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, PMLR 89, 3254–3263 (2019).
Xiang, S. & Li, H. On the Effects of Batch and Weight Normalization in Generative Adversarial Networks. arXiv preprint: Machine Learning (2017).
He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, https://doi.org/10.1109/CVPR.2016.90 , 770–778 (2016).
Simonyan, K. & Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556 (2014).
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A. & Chen, L. C. MobileNetV2: Inverted Residuals and Linear Bottlenecks. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 4510–4520 https://doi.org/10.1109/CVPR.2018.00474 (2018).
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. & Wojna, Z. Rethinking the inception architecture for computer vision. in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 2818–2826 (2016).
Szegedy, C., Ioffe, S., Vanhoucke, V. & Alemi, A. A. Inception-v4, inception-resnet and the impact of residual connections on learning. in Thirty-first AAAI Conference on Artificial Intelligence (2017).
Chollet, F. Xception: Deep learning with depthwise separable convolutions. in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 1800–1807 (2017).
Tan, M. & Le, Q. Efficientnet: Rethinking model scaling for convolutional neural networks. in International Conference on Machine Learning, PMLR, Long Beach, CA, USA, 6105–6114 (2019).
Poulinakis, K., Drikakis, D., Kokkinakis, I. W. & Spottswood, S. M. Machine-learning methods on noisy and sparse. Data. Math. 11, 236 (2023).
Erion, G., Janizek, J. D., Sturmfels, P., Lundberg, S. M. & Lee, S.-I. Improving performance of deep learning models with axiomatic attribution priors and expected gradients. Nat. Mach. Intell. 3, 620–631 (2021).
doi: 10.1038/s42256-021-00343-w
Sundararajan, M., Taly, A. & Yan, Q. Axiomatic attribution for deep networks. in International Conference on Machine Learning, 70, 3319–3328 (2017).
Rozemberczki, B. et al. The Shapley Value in Machine Learning. in Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI-22 (ed. De Raedt, L.) 5572–5579 (International Joint Conferences on Artificial Intelligence Organization, 2022). https://doi.org/10.24963/ijcai.2022/778 .
Bach, S. et al. On pixel-wise explanations for non-linear classifier decisions by layer-wise relevance propagation. PLoS One 10, e0130140 (2015).
doi: 10.1371/journal.pone.0130140
pubmed: 26161953
pmcid: 4498753
Kassani, P. H., Lu, F., Le Guen, Y., Belloy, M. E. & He, Z. Deep neural networks with controlled variable selection for the identification of putative causal genetic variants. Nat. Mach. Intell. 4, 761–771 (2022).
doi: 10.1038/s42256-022-00525-0
pubmed: 37859729
pmcid: 10586424
Wang, A., Khojah, A., Morgan, G. & Pachman, L. M. Nailfold capillary dropout precedes the presentation of pneumatosis intestinalis and micro-perforation in juvenile dermatomyositis. Clin. Immunol. Commun. 3, 74–76 (2023).
doi: 10.1016/j.clicom.2023.06.001