Automatic deep learning multicontrast corpus callosum segmentation in multiple sclerosis.
atrophy
convolutional neural networks
corpus callosum
magnetic resonance imaging
multiple sclerosis
neurodegeneration
Journal
Journal of neuroimaging : official journal of the American Society of Neuroimaging
ISSN: 1552-6569
Titre abrégé: J Neuroimaging
Pays: United States
ID NLM: 9102705
Informations de publication
Date de publication:
05 2022
05 2022
Historique:
revised:
05
01
2022
received:
22
11
2021
accepted:
12
01
2022
pubmed:
28
1
2022
medline:
12
5
2022
entrez:
27
1
2022
Statut:
ppublish
Résumé
Corpus callosum (CC) atrophy is predictive of future disability in multiple sclerosis (MS). However, current segmentation methods are either labor- or computationally intensive. We therefore developed an automated deep learning-based CC segmentation tool and hypothesized that its output would correlate with disability. A cohort of 631 MS patients (449 females, baseline age 41 ± 11 years) with both 3-dimensional T1-weighted and T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI was used for the development. Data from 204 patients were manually segmented to train convolutional neural networks in extracting the midsagittal intracranial and CC areas. Remaining data were used to compare segmentations with FreeSurfer and benchmark the outputs with regard to clinical correlations. A 1.5 and 3 Tesla reproducibility cohort of 9 MS patients evaluated the segmentation robustness. The deep learning-based tool was accurate in selecting the appropriate slice for segmentation (98% accuracy within 3 mm of the manual ground truth) and segmenting the CC (Dice coefficient .88-.91) and intracranial areas (.97-.98). The accuracy was lower with higher atrophy. Reproducibility was excellent (intraclass correlation coefficient > .90) for T1-weighted scans and moderate-good for FLAIR (.74-.75). Segmentations were associated with baseline and future (average follow-up time 6-7 years) Expanded Disability Status Scale (ρ = -.13 to -.24) and Symbol Digit Modalities Test (r = .18-.29) scores. We present a fully automatic deep learning-based CC segmentation tool optimized to modern imaging in MS with clinical correlations on par with computationally expensive alternatives.
Sections du résumé
BACKGROUND AND PURPOSE
Corpus callosum (CC) atrophy is predictive of future disability in multiple sclerosis (MS). However, current segmentation methods are either labor- or computationally intensive. We therefore developed an automated deep learning-based CC segmentation tool and hypothesized that its output would correlate with disability.
METHODS
A cohort of 631 MS patients (449 females, baseline age 41 ± 11 years) with both 3-dimensional T1-weighted and T2-weighted fluid-attenuated inversion recovery (FLAIR) MRI was used for the development. Data from 204 patients were manually segmented to train convolutional neural networks in extracting the midsagittal intracranial and CC areas. Remaining data were used to compare segmentations with FreeSurfer and benchmark the outputs with regard to clinical correlations. A 1.5 and 3 Tesla reproducibility cohort of 9 MS patients evaluated the segmentation robustness.
RESULTS
The deep learning-based tool was accurate in selecting the appropriate slice for segmentation (98% accuracy within 3 mm of the manual ground truth) and segmenting the CC (Dice coefficient .88-.91) and intracranial areas (.97-.98). The accuracy was lower with higher atrophy. Reproducibility was excellent (intraclass correlation coefficient > .90) for T1-weighted scans and moderate-good for FLAIR (.74-.75). Segmentations were associated with baseline and future (average follow-up time 6-7 years) Expanded Disability Status Scale (ρ = -.13 to -.24) and Symbol Digit Modalities Test (r = .18-.29) scores.
CONCLUSIONS
We present a fully automatic deep learning-based CC segmentation tool optimized to modern imaging in MS with clinical correlations on par with computationally expensive alternatives.
Identifiants
pubmed: 35083815
doi: 10.1111/jon.12972
pmc: PMC9304261
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
459-470Informations de copyright
© 2022 The Authors. Journal of Neuroimaging published by Wiley Periodicals LLC on behalf of American Society of Neuroimaging.
Références
Ann Neurol. 2005 Dec;58(6):840-6
pubmed: 16283615
Stat Methods Med Res. 1999 Jun;8(2):135-60
pubmed: 10501650
Mult Scler. 2017 Aug;23(9):1233-1240
pubmed: 27754943
Lancet Neurol. 2021 Aug;20(8):653-670
pubmed: 34139157
Ann Neurol. 2001 Jul;50(1):121-7
pubmed: 11456302
Neuroimage Clin. 2020;27:102315
pubmed: 32593977
Neuroimage. 2010 Dec;53(4):1181-96
pubmed: 20637289
Front Neurol. 2017 Sep 04;8:433
pubmed: 28928705
J Med Syst. 2018 Oct 8;42(11):226
pubmed: 30298337
Nat Rev Neurol. 2014 Apr;10(4):225-38
pubmed: 24638138
Nat Rev Dis Primers. 2018 Nov 8;4(1):43
pubmed: 30410033
Neurosciences (Riyadh). 2020 Jul;25(3):193-199
pubmed: 32683399
Epidemiology. 2019 Mar;30(2):230-233
pubmed: 30721167
Neuroimage. 2012 Jul 16;61(4):1402-18
pubmed: 22430496
J Neuroimaging. 2020 Mar;30(2):198-204
pubmed: 31750599
Ann Neurol. 2011 Feb;69(2):292-302
pubmed: 21387374
Lancet Neurol. 2016 Mar;15(3):292-303
pubmed: 26822746
Mult Scler Relat Disord. 2018 Feb;20:154-158
pubmed: 29414290
J Neuroimaging. 2015 Jan-Feb;25(1):62-7
pubmed: 24816394
Front Neurosci. 2020 Jan 24;14:15
pubmed: 32226359
J Neuroimaging. 2015 Nov-Dec;25(6):996-1001
pubmed: 25786805
J Neuroimaging. 2022 May;32(3):459-470
pubmed: 35083815
Neurology. 2017 Mar 28;88(13):1265-1272
pubmed: 28235816
Mult Scler Relat Disord. 2018 Apr;21:110-116
pubmed: 29550717
Radiol Phys Technol. 2021 Dec;14(4):358-365
pubmed: 34338999
AJR Am J Roentgenol. 1993 May;160(5):949-55
pubmed: 8470609
J Neuroimaging. 2021 May;31(3):493-500
pubmed: 33587820
Neuroimage. 2006 Jul 1;31(3):1116-28
pubmed: 16545965
Arch Neurol. 1998 Feb;55(2):193-8
pubmed: 9482361