Multicenter Validation of a Deep Learning Detection Algorithm for Focal Cortical Dysplasia.
Adolescent
Adult
Child
Child, Preschool
Deep Learning
Female
Humans
Image Interpretation, Computer-Assisted
/ methods
Imaging, Three-Dimensional
/ methods
Magnetic Resonance Imaging
/ methods
Male
Malformations of Cortical Development
/ diagnostic imaging
Middle Aged
Neuroimaging
/ methods
Young Adult
Journal
Neurology
ISSN: 1526-632X
Titre abrégé: Neurology
Pays: United States
ID NLM: 0401060
Informations de publication
Date de publication:
19 10 2021
19 10 2021
Historique:
received:
29
08
2020
accepted:
26
07
2021
pubmed:
16
9
2021
medline:
30
10
2021
entrez:
15
9
2021
Statut:
ppublish
Résumé
To test the hypothesis that a multicenter-validated computer deep learning algorithm detects MRI-negative focal cortical dysplasia (FCD). We used clinically acquired 3-dimensional (3D) T1-weighted and 3D fluid-attenuated inversion recovery MRI of 148 patients (median age 23 years [range 2-55 years]; 47% female) with histologically verified FCD at 9 centers to train a deep convolutional neural network (CNN) classifier. Images were initially deemed MRI-negative in 51% of patients, in whom intracranial EEG determined the focus. For risk stratification, the CNN incorporated bayesian uncertainty estimation as a measure of confidence. To evaluate performance, detection maps were compared to expert FCD manual labels. Sensitivity was tested in an independent cohort of 23 cases with FCD (13 ± 10 years). Applying the algorithm to 42 healthy controls and 89 controls with temporal lobe epilepsy disease tested specificity. Overall sensitivity was 93% (137 of 148 FCD detected) using a leave-one-site-out cross-validation, with an average of 6 false positives per patient. Sensitivity in MRI-negative FCD was 85%. In 73% of patients, the FCD was among the clusters with the highest confidence; in half, it ranked the highest. Sensitivity in the independent cohort was 83% (19 of 23; average of 5 false positives per patient). Specificity was 89% in healthy and disease controls. This first multicenter-validated deep learning detection algorithm yields the highest sensitivity to date in MRI-negative FCD. By pairing predictions with risk stratification, this classifier may assist clinicians in adjusting hypotheses relative to other tests, increasing diagnostic confidence. Moreover, generalizability across age and MRI hardware makes this approach ideal for presurgical evaluation of MRI-negative epilepsy. This study provides Class III evidence that deep learning on multimodal MRI accurately identifies FCD in patients with epilepsy initially diagnosed as MRI negative.
Sections du résumé
BACKGROUND AND OBJECTIVE
To test the hypothesis that a multicenter-validated computer deep learning algorithm detects MRI-negative focal cortical dysplasia (FCD).
METHODS
We used clinically acquired 3-dimensional (3D) T1-weighted and 3D fluid-attenuated inversion recovery MRI of 148 patients (median age 23 years [range 2-55 years]; 47% female) with histologically verified FCD at 9 centers to train a deep convolutional neural network (CNN) classifier. Images were initially deemed MRI-negative in 51% of patients, in whom intracranial EEG determined the focus. For risk stratification, the CNN incorporated bayesian uncertainty estimation as a measure of confidence. To evaluate performance, detection maps were compared to expert FCD manual labels. Sensitivity was tested in an independent cohort of 23 cases with FCD (13 ± 10 years). Applying the algorithm to 42 healthy controls and 89 controls with temporal lobe epilepsy disease tested specificity.
RESULTS
Overall sensitivity was 93% (137 of 148 FCD detected) using a leave-one-site-out cross-validation, with an average of 6 false positives per patient. Sensitivity in MRI-negative FCD was 85%. In 73% of patients, the FCD was among the clusters with the highest confidence; in half, it ranked the highest. Sensitivity in the independent cohort was 83% (19 of 23; average of 5 false positives per patient). Specificity was 89% in healthy and disease controls.
DISCUSSION
This first multicenter-validated deep learning detection algorithm yields the highest sensitivity to date in MRI-negative FCD. By pairing predictions with risk stratification, this classifier may assist clinicians in adjusting hypotheses relative to other tests, increasing diagnostic confidence. Moreover, generalizability across age and MRI hardware makes this approach ideal for presurgical evaluation of MRI-negative epilepsy.
CLASSIFICATION OF EVIDENCE
This study provides Class III evidence that deep learning on multimodal MRI accurately identifies FCD in patients with epilepsy initially diagnosed as MRI negative.
Identifiants
pubmed: 34521691
pii: WNL.0000000000012698
doi: 10.1212/WNL.0000000000012698
pmc: PMC8548962
doi:
Types de publication
Journal Article
Multicenter Study
Research Support, Non-U.S. Gov't
Validation Study
Langues
eng
Sous-ensembles de citation
IM
Pagination
e1571-e1582Subventions
Organisme : CIHR
ID : MOP-57840
Pays : Canada
Commentaires et corrections
Type : CommentIn
Type : ErratumIn
Type : CommentIn
Informations de copyright
© 2021 American Academy of Neurology.
Références
Epilepsy Behav. 2015 Jul;48:21-8
pubmed: 26037845
IEEE Trans Med Imaging. 1998 Feb;17(1):87-97
pubmed: 9617910
J Neurol Neurosurg Psychiatry. 2014 Jul;85(7):716-20
pubmed: 24293000
Epilepsia. 2013 May;54(5):840-7
pubmed: 23551133
Brain. 2009 Aug;132(Pt 8):2079-90
pubmed: 19506069
Epilepsia. 2015 Nov;56(11):1669-86
pubmed: 26434565
Neuroimage Clin. 2016 Dec 30;14:18-27
pubmed: 28123950
Med Image Anal. 2017 Dec;42:60-88
pubmed: 28778026
PLoS One. 2011 Feb 04;6(2):e16430
pubmed: 21326599
Nat Rev Neurol. 2011 Feb;7(2):99-108
pubmed: 21243016
Neurology. 2014 Jul 1;83(1):48-55
pubmed: 24898923
Prog Brain Res. 2012;195:123-63
pubmed: 22230626
JAMA Neurol. 2013 Aug;70(8):1003-8
pubmed: 23732844
Radiology. 2006 Mar;238(3):780-9
pubmed: 16505391
Neuroimage Clin. 2016 Feb 23;11:515-529
pubmed: 27114900
Neurology. 2017 Feb 21;88(8):734-742
pubmed: 28130467
Acta Neuropathol Commun. 2020 Mar 9;8(1):27
pubmed: 32151273
Neuroimage Clin. 2021;30:102565
pubmed: 33556791
Nature. 2015 May 28;521(7553):436-44
pubmed: 26017442
Nat Rev Neurosci. 2018 Feb 16;19(3):123-137
pubmed: 29449712
Curr Opin Neurol. 2014 Apr;27(2):206-12
pubmed: 24553461
Neurology. 2012 Jul 3;79(1):13-4
pubmed: 22722620
Epilepsia. 2019 Jun;60(6):1054-1068
pubmed: 31135062
Epilepsia. 2011 Jan;52(1):158-74
pubmed: 21219302
IEEE J Biomed Health Inform. 2021 May;25(5):1724-1734
pubmed: 32931436
Epilepsy Res. 2010 May;89(2-3):310-8
pubmed: 20227852
Comput Med Imaging Graph. 2020 Jan;79:101662
pubmed: 31812131
Brain. 2008 Dec;131(Pt 12):3246-55
pubmed: 18812443
Nat Med. 2019 Jan;25(1):44-56
pubmed: 30617339
Sci Rep. 2017 Dec 19;7(1):17816
pubmed: 29259224
Curr Biol. 2017 Jun 19;27(12):R580-R586
pubmed: 28633023