Robust deep learning-based segmentation of glioblastoma on routine clinical mri scans using sparsified training
AuthorsEijgelaar, R. S.
Müller, D. M. J.
van Herk, Marcel
Conti Nibali, M.
Berger, M. S.
De Witt Hamer, P. C.
de Munck, J. C.
Witte, M. G.
AffiliationDepartment of Radiation Oncology, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, the Netherlands
MetadataShow full item record
AbstractPurpose: To improve the robustness of deep learning-based glioblastoma segmentation in a clinical setting with sparsified datasets. Materials and methods: In this retrospective study, preoperative T1-weighted, T2-weighted, T2-weighted fluid-attenuated inversion recovery, and postcontrast T1-weighted MRI from 117 patients (median age, 64 years; interquartile range [IQR], 55-73 years; 76 men) included within the Multimodal Brain Tumor Image Segmentation (BraTS) dataset plus a clinical dataset (2012-2013) with similar imaging modalities of 634 patients (median age, 59 years; IQR, 49-69 years; 382 men) with glioblastoma from six hospitals were used. Expert tumor delineations on the postcontrast images were available, but for various clinical datasets, one or more sequences were missing. The convolutional neural network, DeepMedic, was trained on combinations of complete and incomplete data with and without site-specific data. Sparsified training was introduced, which randomly simulated missing sequences during training. The effects of sparsified training and center-specific training were tested using Wilcoxon signed rank tests for paired measurements. Results: A model trained exclusively on BraTS data reached a median Dice score of 0.81 for segmentation on BraTS test data but only 0.49 on the clinical data. Sparsified training improved performance (adjusted P < .05), even when excluding test data with missing sequences, to median Dice score of 0.67. Inclusion of site-specific data during sparsified training led to higher model performance Dice scores greater than 0.8, on par with a model based on all complete and incomplete data. For the model using BraTS and clinical training data, inclusion of site-specific data or sparsified training was of no consequence. Conclusion: Accurate and automatic segmentation of glioblastoma on clinical scans is feasible using a model based on large, heterogeneous, and partially incomplete datasets. Sparsified training may boost the performance of a smaller model based on public and site-specific data.Supplemental material is available for this article.Published under a CC BY 4.0 license.
CitationEijgelaar RS, Visser M, Müller DMJ, Barkhof F, Vrenken H, van Herk M, et al. Robust Deep Learning–based Segmentation of Glioblastoma on Routine Clinical MRI Scans Using Sparsified Training. Radiology: Artificial Intelligence. 2020 Sep 1;2(5):e190103.
JournalRadiology: Artificial Intelligence
- Generative Adversarial Networks to Synthesize Missing T1 and FLAIR MRI Sequences for Use in a Multisequence Brain Tumor Segmentation Model.
- Authors: Conte GM, Weston AD, Vogelsang DC, Philbrick KA, Cai JC, Barbera M, Sanvito F, Lachance DH, Jenkins RB, Tobin WO, Eckel-Passow JE, Erickson BJ
- Issue date: 2021 May
- Clinical Evaluation of a Multiparametric Deep Learning Model for Glioblastoma Segmentation Using Heterogeneous Magnetic Resonance Imaging Data From Clinical Routine.
- Authors: Perkuhn M, Stavrinou P, Thiele F, Shakirin G, Mohan M, Garmpis D, Kabbasch C, Borggrefe J
- Issue date: 2018 Nov
- Brain tumor segmentation using holistically nested neural networks in MRI images.
- Authors: Zhuge Y, Krauze AV, Ning H, Cheng JY, Arora BC, Camphausen K, Miller RW
- Issue date: 2017 Oct
- Deep learning for glioblastoma segmentation using preoperative magnetic resonance imaging identifies volumetric features associated with survival.
- Authors: Wan Y, Rahmat R, Price SJ
- Issue date: 2020 Dec
- Handling missing MRI sequences in deep learning segmentation of brain metastases: a multicenter study.
- Authors: Grøvik E, Yi D, Iv M, Tong E, Nilsen LB, Latysheva A, Saxhaug C, Jacobsen KD, Helland Å, Emblem KE, Rubin DL, Zaharchuk G
- Issue date: 2021 Feb 22