TY - JOUR
T1 - Multi-class glioma segmentation on real-world data with missing MRI sequences
T2 - comparison of three deep learning algorithms
AU - Pemberton, Hugh G.
AU - Wu, Jiaming
AU - Kommers, Ivar
AU - Müller, Domenique M. J.
AU - Hu, Yipeng
AU - Goodkin, Olivia
AU - Vos, Sjoerd B.
AU - Bisdas, Sotirios
AU - Robe, Pierre A.
AU - Ardon, Hilko
AU - Bello, Lorenzo
AU - Rossi, Marco
AU - Sciortino, Tommaso
AU - Nibali, Marco Conti
AU - Berger, Mitchel S.
AU - Hervey-Jumper, Shawn L.
AU - Bouwknegt, Wim
AU - van den Brink, Wimar A.
AU - Furtner, Julia
AU - Han, Seunggu J.
AU - Idema, Albert J. S.
AU - Kiesel, Barbara
AU - Widhalm, Georg
AU - Kloet, Alfred
AU - Wagemakers, Michiel
AU - Zwinderman, Aeilko H.
AU - Krieg, Sandro M.
AU - Mandonnet, Emmanuel
AU - Prados, Ferran
AU - de Witt Hamer, Philip
AU - Barkhof, Frederik
AU - Eijgelaar, Roelant S.
N1 - Funding Information: The authors would like to thank all patients whose data was used in this study. HP is a full-time employee of Deloitte. FB, SB, FP and JW are supported by the National Institute for Health Research (NIHR) biomedical research centre at UCLH. FP received a Guarantors of Brain fellowship 2017–2020 and is also supported by the Biomedical Research Centre initiative at University College London Hospitals (UCLH). The PICTURE project is sponsored by an unrestricted grant of Stichting Hanarth fonds, “Machine learning for better neurosurgical decisions in patients with glioblastoma”; a grant for public-private partnerships (Amsterdam UMC PPP-grant) sponsored by the Dutch government (Ministry of Economic Affairs) through the Rijksdienst voor Ondernemend Nederland (RVO) and Topsector Life Sciences and Health (LSH), “Picturing predictions for patients with brain tumors”; a grant from the Innovative Medical Devices Initiative program, project number 10-10400-96-14003; The Netherlands Organisation for Scientific Research (NWO), 2020.027; a grant from the Dutch Cancer Society, VU2014-7113 and the Anita Veldman foundation, CCA2018-2-17. Publisher Copyright: © 2023, The Author(s).
PY - 2023/12/1
Y1 - 2023/12/1
N2 - This study tests the generalisability of three Brain Tumor Segmentation (BraTS) challenge models using a multi-center dataset of varying image quality and incomplete MRI datasets. In this retrospective study, DeepMedic, no-new-Unet (nn-Unet), and NVIDIA-net (nv-Net) were trained and tested using manual segmentations from preoperative MRI of glioblastoma (GBM) and low-grade gliomas (LGG) from the BraTS 2021 dataset (1251 in total), in addition to 275 GBM and 205 LGG acquired clinically across 12 hospitals worldwide. Data was split into 80% training, 5% validation, and 15% internal test data. An additional external test-set of 158 GBM and 69 LGG was used to assess generalisability to other hospitals’ data. All models’ median Dice similarity coefficient (DSC) for both test sets were within, or higher than, previously reported human inter-rater agreement (range of 0.74–0.85). For both test sets, nn-Unet achieved the highest DSC (internal = 0.86, external = 0.93) and the lowest Hausdorff distances (10.07, 13.87 mm, respectively) for all tumor classes (p < 0.001). By applying Sparsified training, missing MRI sequences did not statistically affect the performance. nn-Unet achieves accurate segmentations in clinical settings even in the presence of incomplete MRI datasets. This facilitates future clinical adoption of automated glioma segmentation, which could help inform treatment planning and glioma monitoring.
AB - This study tests the generalisability of three Brain Tumor Segmentation (BraTS) challenge models using a multi-center dataset of varying image quality and incomplete MRI datasets. In this retrospective study, DeepMedic, no-new-Unet (nn-Unet), and NVIDIA-net (nv-Net) were trained and tested using manual segmentations from preoperative MRI of glioblastoma (GBM) and low-grade gliomas (LGG) from the BraTS 2021 dataset (1251 in total), in addition to 275 GBM and 205 LGG acquired clinically across 12 hospitals worldwide. Data was split into 80% training, 5% validation, and 15% internal test data. An additional external test-set of 158 GBM and 69 LGG was used to assess generalisability to other hospitals’ data. All models’ median Dice similarity coefficient (DSC) for both test sets were within, or higher than, previously reported human inter-rater agreement (range of 0.74–0.85). For both test sets, nn-Unet achieved the highest DSC (internal = 0.86, external = 0.93) and the lowest Hausdorff distances (10.07, 13.87 mm, respectively) for all tumor classes (p < 0.001). By applying Sparsified training, missing MRI sequences did not statistically affect the performance. nn-Unet achieves accurate segmentations in clinical settings even in the presence of incomplete MRI datasets. This facilitates future clinical adoption of automated glioma segmentation, which could help inform treatment planning and glioma monitoring.
UR - http://www.scopus.com/inward/record.url?scp=85175837438&partnerID=8YFLogxK
U2 - https://doi.org/10.1038/s41598-023-44794-0
DO - https://doi.org/10.1038/s41598-023-44794-0
M3 - Article
C2 - 37919354
SN - 2045-2322
VL - 13
JO - Scientific reports
JF - Scientific reports
IS - 1
M1 - 18911
ER -