Assessment of right ventricular size and function from cardiovascular magnetic resonance images using artificial intelligence.
Artificial intelligence
Deep learning
Right ventricular ejection fraction
Right ventricular function
Journal
Journal of cardiovascular magnetic resonance : official journal of the Society for Cardiovascular Magnetic Resonance
ISSN: 1532-429X
Titre abrégé: J Cardiovasc Magn Reson
Pays: England
ID NLM: 9815616
Informations de publication
Date de publication:
11 04 2022
11 04 2022
Historique:
received:
15
12
2021
accepted:
29
03
2022
entrez:
12
4
2022
pubmed:
13
4
2022
medline:
14
4
2022
Statut:
epublish
Résumé
Theoretically, artificial intelligence can provide an accurate automatic solution to measure right ventricular (RV) ejection fraction (RVEF) from cardiovascular magnetic resonance (CMR) images, despite the complex RV geometry. However, in our recent study, commercially available deep learning (DL) algorithms for RVEF quantification performed poorly in some patients. The current study was designed to test the hypothesis that quantification of RV function could be improved in these patients by using more diverse CMR datasets in addition to domain-specific quantitative performance evaluation metrics during the cross-validation phase of DL algorithm development. We identified 100 patients from our prior study who had the largest differences between manually measured and automated RVEF values. Automated RVEF measurements were performed using the original version of the algorithm (DL1), an updated version (DL2) developed from a dataset that included a wider range of RV pathology and validated using multiple domain-specific quantitative performance evaluation metrics, and conventional methodology performed by a core laboratory (CORE). Each of the DL-RVEF approaches was compared against CORE-RVEF reference values using linear regression and Bland-Altman analyses. Additionally, RVEF values were classified into 3 categories: ≤ 35%, 35-50%, and ≥ 50%. Agreement between RVEF classifications made by the DL approaches and the CORE measurements was tested. CORE-RVEF and DL-RVEFs were obtained in all patients (feasibility of 100%). DL2-RVEF correlated with CORE-RVEF better than DL1-RVEF (r = 0.87 vs. r = 0.42), with narrower limits of agreement. As a result, DL2 algorithm also showed increasing accuracy from 0.53 to 0.80 for categorizing RV function. The use of a new DL algorithm cross-validated on a dataset with a wide range of RV pathology using multiple domain-specific metrics resulted in a considerable improvement in the accuracy of automated RVEF measurements. This improvement was demonstrated in patients whose images were the most challenging and resulted in the largest RVEF errors. These findings underscore the critical importance of this strategy in the development of DL approaches for automated CMR measurements.
Sections du résumé
BACKGROUND
Theoretically, artificial intelligence can provide an accurate automatic solution to measure right ventricular (RV) ejection fraction (RVEF) from cardiovascular magnetic resonance (CMR) images, despite the complex RV geometry. However, in our recent study, commercially available deep learning (DL) algorithms for RVEF quantification performed poorly in some patients. The current study was designed to test the hypothesis that quantification of RV function could be improved in these patients by using more diverse CMR datasets in addition to domain-specific quantitative performance evaluation metrics during the cross-validation phase of DL algorithm development.
METHODS
We identified 100 patients from our prior study who had the largest differences between manually measured and automated RVEF values. Automated RVEF measurements were performed using the original version of the algorithm (DL1), an updated version (DL2) developed from a dataset that included a wider range of RV pathology and validated using multiple domain-specific quantitative performance evaluation metrics, and conventional methodology performed by a core laboratory (CORE). Each of the DL-RVEF approaches was compared against CORE-RVEF reference values using linear regression and Bland-Altman analyses. Additionally, RVEF values were classified into 3 categories: ≤ 35%, 35-50%, and ≥ 50%. Agreement between RVEF classifications made by the DL approaches and the CORE measurements was tested.
RESULTS
CORE-RVEF and DL-RVEFs were obtained in all patients (feasibility of 100%). DL2-RVEF correlated with CORE-RVEF better than DL1-RVEF (r = 0.87 vs. r = 0.42), with narrower limits of agreement. As a result, DL2 algorithm also showed increasing accuracy from 0.53 to 0.80 for categorizing RV function.
CONCLUSIONS
The use of a new DL algorithm cross-validated on a dataset with a wide range of RV pathology using multiple domain-specific metrics resulted in a considerable improvement in the accuracy of automated RVEF measurements. This improvement was demonstrated in patients whose images were the most challenging and resulted in the largest RVEF errors. These findings underscore the critical importance of this strategy in the development of DL approaches for automated CMR measurements.
Identifiants
pubmed: 35410226
doi: 10.1186/s12968-022-00861-5
pii: 10.1186/s12968-022-00861-5
pmc: PMC8996592
doi:
Types de publication
Journal Article
Research Support, N.I.H., Extramural
Langues
eng
Sous-ensembles de citation
IM
Pagination
27Informations de copyright
© 2022. The Author(s).
Références
J Cardiovasc Magn Reson. 2019 Apr 25;21(1):24
pubmed: 31023305
Curr Treat Options Cardiovasc Med. 2017 Sep 25;19(11):82
pubmed: 28948501
Circ Arrhythm Electrophysiol. 2017 Jan;10(1):
pubmed: 28087564
BMC Med Imaging. 2015 Aug 12;15:29
pubmed: 26263899
J Am Coll Cardiol. 2018 Oct 2;72(14):e91-e220
pubmed: 29097296
ESC Heart Fail. 2020 Feb;7(1):253-263
pubmed: 31903694
Circ Cardiovasc Imaging. 2014 Jan;7(1):190-7
pubmed: 24449548
IEEE Trans Med Imaging. 2018 Nov;37(11):2514-2525
pubmed: 29994302
Circ Heart Fail. 2021 Apr;14(4):e007840
pubmed: 33866826
JACC Cardiovasc Imaging. 2017 Oct;10(10 Pt B):1211-1221
pubmed: 28412423
Circ Cardiovasc Imaging. 2021 Mar;14(3):e011337
pubmed: 33722059
JACC Cardiovasc Imaging. 2022 Mar;15(3):413-427
pubmed: 34656471
J Am Coll Cardiol. 2007 Feb 27;49(8):855-62
pubmed: 17320743
Eur Heart J. 2015 Dec 7;36(46):3227-37
pubmed: 26216920
Sci Rep. 2021 Jun 2;11(1):11648
pubmed: 34078942
JACC Cardiovasc Imaging. 2021 May;14(5):931-942
pubmed: 33008758
J Cardiovasc Magn Reson. 2012 Feb 01;14:11
pubmed: 22296860
Eur Heart J. 2016 Jul 14;37(27):2129-2200
pubmed: 27206819
Circ Cardiovasc Imaging. 2014 Jan;7(1):115-24
pubmed: 24192452
JACC Cardiovasc Imaging. 2019 Mar;12(3):516-531
pubmed: 30846125
Circulation. 2013 Oct 8;128(15):1623-33
pubmed: 23965488
JACC Cardiovasc Imaging. 2012 Jan;5(1):28-37
pubmed: 22239890
Radiology. 2017 May;283(2):381-390
pubmed: 28092203
Circ Arrhythm Electrophysiol. 2019 Mar;12(3):e007022
pubmed: 30866666
JACC Cardiovasc Imaging. 2017 Oct;10(10 Pt B):1225-1236
pubmed: 29025576
Magn Reson Med. 2017 Dec;78(6):2439-2448
pubmed: 28205298
J Am Soc Echocardiogr. 2019 Aug;32(8):969-977
pubmed: 31174940
Med Image Anal. 2019 Aug;56:26-42
pubmed: 31154149
Heart. 2015 Nov;101(21):1724-30
pubmed: 26276804
J Am Heart Assoc. 2020 Sep 15;9(18):e016612
pubmed: 32873121
Front Cardiovasc Med. 2020 Jun 30;7:105
pubmed: 32714943
Proc IEEE Int Conf Acoust Speech Signal Process. 2018 Apr;2018:6677-6681
pubmed: 31892860
IEEE Trans Med Imaging. 2021 Dec;40(12):3543-3554
pubmed: 34138702
Circulation. 2010 Jan 19;121(2):252-8
pubmed: 20048206
J Cardiovasc Magn Reson. 2021 Apr 26;23(1):47
pubmed: 33896419