Combined molecular subtyping, grading, and segmentation of glioma using multi-task deep learning.
deep learning
glioma
multi-task
radiomics
segmentation
Journal
Neuro-oncology
ISSN: 1523-5866
Titre abrégé: Neuro Oncol
Pays: England
ID NLM: 100887420
Informations de publication
Date de publication:
14 02 2023
14 02 2023
Historique:
pubmed:
6
7
2022
medline:
16
2
2023
entrez:
5
7
2022
Statut:
ppublish
Résumé
Accurate characterization of glioma is crucial for clinical decision making. A delineation of the tumor is also desirable in the initial decision stages but is time-consuming. Previously, deep learning methods have been developed that can either non-invasively predict the genetic or histological features of glioma, or that can automatically delineate the tumor, but not both tasks at the same time. Here, we present our method that can predict the molecular subtype and grade, while simultaneously providing a delineation of the tumor. We developed a single multi-task convolutional neural network that uses the full 3D, structural, preoperative MRI scans to predict the IDH mutation status, the 1p/19q co-deletion status, and the grade of a tumor, while simultaneously segmenting the tumor. We trained our method using a patient cohort containing 1508 glioma patients from 16 institutes. We tested our method on an independent dataset of 240 patients from 13 different institutes. In the independent test set, we achieved an IDH-AUC of 0.90, an 1p/19q co-deletion AUC of 0.85, and a grade AUC of 0.81 (grade II/III/IV). For the tumor delineation, we achieved a mean whole tumor Dice score of 0.84. We developed a method that non-invasively predicts multiple, clinically relevant features of glioma. Evaluation in an independent dataset shows that the method achieves a high performance and that it generalizes well to the broader clinical population. This first-of-its-kind method opens the door to more generalizable, instead of hyper-specialized, AI methods.
Sections du résumé
BACKGROUND
Accurate characterization of glioma is crucial for clinical decision making. A delineation of the tumor is also desirable in the initial decision stages but is time-consuming. Previously, deep learning methods have been developed that can either non-invasively predict the genetic or histological features of glioma, or that can automatically delineate the tumor, but not both tasks at the same time. Here, we present our method that can predict the molecular subtype and grade, while simultaneously providing a delineation of the tumor.
METHODS
We developed a single multi-task convolutional neural network that uses the full 3D, structural, preoperative MRI scans to predict the IDH mutation status, the 1p/19q co-deletion status, and the grade of a tumor, while simultaneously segmenting the tumor. We trained our method using a patient cohort containing 1508 glioma patients from 16 institutes. We tested our method on an independent dataset of 240 patients from 13 different institutes.
RESULTS
In the independent test set, we achieved an IDH-AUC of 0.90, an 1p/19q co-deletion AUC of 0.85, and a grade AUC of 0.81 (grade II/III/IV). For the tumor delineation, we achieved a mean whole tumor Dice score of 0.84.
CONCLUSIONS
We developed a method that non-invasively predicts multiple, clinically relevant features of glioma. Evaluation in an independent dataset shows that the method achieves a high performance and that it generalizes well to the broader clinical population. This first-of-its-kind method opens the door to more generalizable, instead of hyper-specialized, AI methods.
Identifiants
pubmed: 35788352
pii: 6631283
doi: 10.1093/neuonc/noac166
pmc: PMC9925710
doi:
Substances chimiques
Isocitrate Dehydrogenase
EC 1.1.1.41
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
279-289Informations de copyright
© The Author(s) 2022. Published by Oxford University Press on behalf of the Society for Neuro-Oncology.
Références
Acad Radiol. 2021 Nov;28(11):1599-1621
pubmed: 32660755
Front Oncol. 2019 Aug 22;9:806
pubmed: 31508366
Radiology. 2016 Feb;278(2):563-77
pubmed: 26579733
Eur Radiol. 2018 Aug;28(8):3306-3317
pubmed: 29536240
Comput Med Imaging Graph. 2021 Mar;88:101831
pubmed: 33482430
AJNR Am J Neuroradiol. 2018 Feb;39(2):208-216
pubmed: 28982791
Med Image Comput Comput Assist Interv. 2019;11764:415-422
pubmed: 34085058
Neurooncol Pract. 2020 Jan;7(1):68-76
pubmed: 32025325
Neuroimage Clin. 2019;22:101727
pubmed: 30825711
Acta Neuropathol. 2016 Jun;131(6):803-20
pubmed: 27157931
Radiology. 2015 Dec;277(3):826-32
pubmed: 26509226
Front Neurosci. 2020 Apr 08;14:282
pubmed: 32322186
J Am Coll Radiol. 2018 Mar;15(3 Pt B):504-508
pubmed: 29402533
J Digit Imaging. 2013 Dec;26(6):1045-57
pubmed: 23884657
Acta Neuropathol. 2010 Sep;120(3):297-304
pubmed: 20644945
Science. 2018 May 11;360(6389):660-663
pubmed: 29748285
Nat Rev Cancer. 2018 Aug;18(8):500-510
pubmed: 29777175
Neuro Oncol. 2016 Mar;18(3):388-400
pubmed: 26354927
Sci Rep. 2020 Mar 6;10(1):4250
pubmed: 32144360
Sci Data. 2017 Sep 05;4:170117
pubmed: 28872634
IEEE Trans Med Imaging. 2015 Oct;34(10):1993-2024
pubmed: 25494501
Data Brief. 2021 Jun 02;37:107191
pubmed: 34159239
Cell. 2016 Jan 28;164(3):550-63
pubmed: 26824661
J Neuropathol Exp Neurol. 2020 Aug 1;79(8):855-862
pubmed: 32688383
Br J Radiol. 2016;89(1060):20150857
pubmed: 26849038
Neurology. 2010 Jun 8;74(23):1886-90
pubmed: 20427748
Clin Neurol Neurosurg. 2009 Dec;111(10):835-9
pubmed: 19765887
Neuro Oncol. 2021 Aug 2;23(8):1231-1251
pubmed: 34185076