A low-cost texture-based pipeline for predicting myocardial tissue remodeling and fibrosis using cardiac ultrasound.
Clustering
Echocardiography
Machine learning
Radiomics
Tissue characterization
Journal
EBioMedicine
ISSN: 2352-3964
Titre abrégé: EBioMedicine
Pays: Netherlands
ID NLM: 101647039
Informations de publication
Date de publication:
Apr 2020
Apr 2020
Historique:
received:
30
12
2019
revised:
04
03
2020
accepted:
04
03
2020
pubmed:
9
4
2020
medline:
26
1
2021
entrez:
9
4
2020
Statut:
ppublish
Résumé
Maturation of ultrasound myocardial tissue characterization may have far-reaching implications as a widely available alternative to cardiac magnetic resonance (CMR) for risk stratification in left ventricular (LV) remodeling. We extracted 328 texture-based features of myocardium from still ultrasound images. After we explored the phenotypes of myocardial textures using unsupervised similarity networks, global LV remodeling parameters were predicted using supervised machine learning models. Separately, we also developed supervised models for predicting the presence of myocardial fibrosis using another cohort who underwent cardiac magnetic resonance (CMR). For the prediction, patients were divided into a training and test set (80:20). Texture-based tissue feature extraction was feasible in 97% of total 534 patients. Interpatient similarity analysis delineated two patient groups based on the texture features: one group had more advanced LV remodeling parameters compared to the other group. Furthermore, this group was associated with a higher incidence of cardiac deaths (p = 0.001) and major adverse cardiac events (p < 0.001). The supervised models predicted reduced LV ejection fraction (<50%) and global longitudinal strain (<16%) with area under the receiver-operator-characteristics curves (ROC AUC) of 0.83 and 0.87 in the hold-out test set, respectively. Furthermore, the presence of myocardial fibrosis was predicted from only ultrasound myocardial texture with an ROC AUC of 0.84 (sensitivity 86.4% and specificity 83.3%) in the test set. Ultrasound texture-based myocardial tissue characterization identified phenotypic features of LV remodeling from still ultrasound images. Further clinical validation may address critical barriers in the adoption of ultrasound techniques for myocardial tissue characterization. None.
Sections du résumé
BACKGROUND
BACKGROUND
Maturation of ultrasound myocardial tissue characterization may have far-reaching implications as a widely available alternative to cardiac magnetic resonance (CMR) for risk stratification in left ventricular (LV) remodeling.
METHODS
METHODS
We extracted 328 texture-based features of myocardium from still ultrasound images. After we explored the phenotypes of myocardial textures using unsupervised similarity networks, global LV remodeling parameters were predicted using supervised machine learning models. Separately, we also developed supervised models for predicting the presence of myocardial fibrosis using another cohort who underwent cardiac magnetic resonance (CMR). For the prediction, patients were divided into a training and test set (80:20).
FINDINGS
RESULTS
Texture-based tissue feature extraction was feasible in 97% of total 534 patients. Interpatient similarity analysis delineated two patient groups based on the texture features: one group had more advanced LV remodeling parameters compared to the other group. Furthermore, this group was associated with a higher incidence of cardiac deaths (p = 0.001) and major adverse cardiac events (p < 0.001). The supervised models predicted reduced LV ejection fraction (<50%) and global longitudinal strain (<16%) with area under the receiver-operator-characteristics curves (ROC AUC) of 0.83 and 0.87 in the hold-out test set, respectively. Furthermore, the presence of myocardial fibrosis was predicted from only ultrasound myocardial texture with an ROC AUC of 0.84 (sensitivity 86.4% and specificity 83.3%) in the test set.
INTERPRETATION
CONCLUSIONS
Ultrasound texture-based myocardial tissue characterization identified phenotypic features of LV remodeling from still ultrasound images. Further clinical validation may address critical barriers in the adoption of ultrasound techniques for myocardial tissue characterization.
FUNDING
BACKGROUND
None.
Identifiants
pubmed: 32268274
pii: S2352-3964(20)30101-8
doi: 10.1016/j.ebiom.2020.102726
pmc: PMC7139137
pii:
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
102726Informations de copyright
Copyright © 2020 The Author(s). Published by Elsevier B.V. All rights reserved.
Déclaration de conflit d'intérêts
Declaration of Competing Interest Nobuyuki Kagiyama is supported by a research grant from Hitachi Healthcare; Partho P. Sengupta is a consultant to Heart Sciences, Ultromics, and Kencor Health. The other authors have nothing to disclose
Références
J Am Coll Cardiol. 2015 Nov 10;66(19):2059-2071
pubmed: 26541915
Bioinformatics. 2010 Sep 1;26(17):2136-44
pubmed: 20591905
Radiology. 2016 Feb;278(2):563-77
pubmed: 26579733
JACC Cardiovasc Imaging. 2011 Jan;4(1):98-108
pubmed: 21232712
J Cardiovasc Magn Reson. 2017 Jan 11;19(1):9
pubmed: 28122618
Circ Cardiovasc Imaging. 2011 Mar;4(2):179-90
pubmed: 21406664
J Am Soc Echocardiogr. 2011 Mar;24(3):277-313
pubmed: 21338865
Q J Nucl Med Mol Imaging. 2019 Dec;63(4):323-338
pubmed: 31527580
Eur Heart J Cardiovasc Imaging. 2015 Feb;16(2):210-6
pubmed: 25246502
Lancet. 2015 Feb 26;385 Suppl 1:S42
pubmed: 26312864
Eur Heart J. 2019 Nov 14;40(43):3529-3543
pubmed: 31504423
J Am Coll Cardiol. 1989 Mar 15;13(4):832-40
pubmed: 2926037
JACC Cardiovasc Imaging. 2019 Feb;12(2):236-248
pubmed: 30732719
Sci Rep. 2013;3:1236
pubmed: 23393618
J Magn Reson Imaging. 2020 Sep;52(3):906-919
pubmed: 31971296
Clin Radiol. 2004 Dec;59(12):1061-9
pubmed: 15556588
Circulation. 1988 Feb;77(2):328-36
pubmed: 3338129
Acad Radiol. 2020 Nov;27(11):e254-e262
pubmed: 31982342
JACC Cardiovasc Imaging. 2018 Dec;11(12):1837-1853
pubmed: 30522686
Nature. 2007 Feb 8;445(7128):627-30
pubmed: 17287805
Cardiovasc Diagn Ther. 2018 Feb;8(1):101-117
pubmed: 29541615
Cancer Res. 2018 Aug 15;78(16):4786-4789
pubmed: 29959149
Nat Commun. 2014 Jun 03;5:4006
pubmed: 24892406
Expert Rev Precis Med Drug Dev. 2019;4(2):59-72
pubmed: 31080889
J Am Coll Cardiol. 2013 Oct 15;62(16):e147-239
pubmed: 23747642
Circulation. 2018 Mar 20;137(12):e67-e492
pubmed: 29386200
J Am Coll Cardiol. 1985 Oct;6(4):737-43
pubmed: 4031287
Int J Med Inform. 2004 Feb;73(1):1-23
pubmed: 15036075
Circulation. 1989 Sep;80(3):491-503
pubmed: 2766504
J Am Soc Echocardiogr. 2015 Jan;28(1):1-39.e14
pubmed: 25559473
Magn Reson Med. 2002 Feb;47(2):372-83
pubmed: 11810682
Circulation. 2003 Sep 16;108(11):1395-403
pubmed: 12975244