Machine Learning–Based Radiomics for Molecular Subtyping of Gliomas

Clinical Cancer Research - Tập 24 Số 18 - Trang 4429-4436 - 2018
Chia‐Feng Lu1,2,3, Fei‐Ting Hsu2,4,5, Kevin Li‐Chun Hsieh2,4,5, Yu‐Chieh Jill Kao2,5, Cheng‐Yu Chen4, Justin Bo‐Kai Hsu6, Ping‐Huei Tsai2,7,8, Ray‐Jade Chen9,10, Chao‐Ching Huang11,12,13, Yun Yen14
11Department of Biomedical Imaging and Radiological Sciences, National Yang-Ming University, Taipei, Taiwan.
22Research Center of Translational Imaging, College of Medicine, Taipei Medical University, Taipei, Taiwan.
33Department of Anatomy and Cell Biology, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
44Department of Medical Imaging, Taipei Medical University Hospital, Taipei, Taiwan.
55Department of Radiology, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
66Department of Medical Research, Taipei Medical University Hospital, Taipei, Taiwan.
77Department of Medical Imaging and Radiological Sciences, Chung Shan Medical University, Taichung, Taiwan.
88Department of Medical Imaging, Chung Shan Medical University Hospital, Taichung, Taiwan.
910Division of General Surgery, Department of Surgery, Taipei Medical University Hospital, Taipei, Taiwan.
109Department of Surgery, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
1111Department of Pediatrics, School of Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
1212Department of Pediatrics, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
1313Department of Pediatrics, Wan-Fang Hospital, Taipei Medical University, Taipei, Taiwan.
1414Ph.D. Program for Cancer Biology and Drug Discovery, College of Medical Science and Technology, Taipei Medical University, Taipei, Taiwan.

Tóm tắt

Abstract Purpose: The new classification announced by the World Health Organization in 2016 recognized five molecular subtypes of diffuse gliomas based on isocitrate dehydrogenase (IDH) and 1p/19q genotypes in addition to histologic phenotypes. We aim to determine whether clinical MRI can stratify these molecular subtypes to benefit the diagnosis and monitoring of gliomas. Experimental Design: The data from 456 subjects with gliomas were obtained from The Cancer Imaging Archive. Overall, 214 subjects, including 106 cases of glioblastomas and 108 cases of lower grade gliomas with preoperative MRI, survival data, histology, IDH, and 1p/19q status were included. We proposed a three-level machine-learning model based on multimodal MR radiomics to classify glioma subtypes. An independent dataset with 70 glioma subjects was further collected to verify the model performance. Results: The IDH and 1p/19q status of gliomas can be classified by radiomics and machine-learning approaches, with areas under ROC curves between 0.922 and 0.975 and accuracies between 87.7% and 96.1% estimated on the training dataset. The test on the validation dataset showed a comparable model performance with that on the training dataset, suggesting the efficacy of the trained classifiers. The classification of 5 molecular subtypes solely based on the MR phenotypes achieved an 81.8% accuracy, and a higher accuracy of 89.2% could be achieved if the histology diagnosis is available. Conclusions: The MR radiomics-based method provides a reliable alternative to determine the histology and molecular subtypes of gliomas. Clin Cancer Res; 24(18); 4429–36. ©2018 AACR.

Từ khóa


Tài liệu tham khảo

Eckel-Passow, 2015, Glioma groups based on 1p/19q, IDH, and TERT promoter mutations in tumors, N Engl J Med, 372, 2499, 10.1056/NEJMoa1407279

Brat, 2015, Comprehensive, integrative genomic analysis of diffuse lower-grade gliomas, N Engl J Med, 372, 2481, 10.1056/NEJMoa1402121

Ceccarelli, 2016, Molecular profiling reveals biologically discrete subsets and pathways of progression in diffuse glioma, Cell, 164, 550, 10.1016/j.cell.2015.12.028

Louis, 2016, World Health Organization Histological Classification of Tumours of the Central Nervous System.

Louis, 2016, The 2016 World Health Organization classification of tumors of the central nervous system: a summary, Acta Neuropathol, 131, 803, 10.1007/s00401-016-1545-1

Diehn, 2008, Identification of noninvasive imaging surrogates for brain tumor gene-expression modules, Proc Natl Acad Sci U S A, 105, 5213, 10.1073/pnas.0801279105

Lambin, 2012, Radiomics: extracting more information from medical images using advanced feature analysis, Eur J Cancer, 48, 441, 10.1016/j.ejca.2011.11.036

Aerts, 2014, Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach, Nat Commun, 5, 4006, 10.1038/ncomms5006

Hsieh, 2017, Quantitative glioma grading using transformed gray-scale invariant textures of MRI, Computers Biol Med, 83, 102, 10.1016/j.compbiomed.2017.02.012

Hsieh, 2017, Computer-aided grading of gliomas based on local and global MRI features, Computer Methods Prog Biomed, 139, 31, 10.1016/j.cmpb.2016.10.021

Zhang, 2017, Multimodal MRI features predict isocitrate dehydrogenase genotype in high-grade gliomas, Neuro-oncol, 19, 109, 10.1093/neuonc/now121

Hsieh, 2017, Radiomic model for predicting mutations in the isocitrate dehydrogenase gene in glioblastomas, Oncotarget, 8, 45888, 10.18632/oncotarget.17585

Kickingereder, 2016, Radiomic profiling of glioblastoma: identifying an imaging predictor of patient survival with improved performance over established clinical and radiologic risk models, Radiology, 280, 880, 10.1148/radiol.2016160845

Clark, 2013, The Cancer Imaging Archive (TCIA): maintaining and operating a public information repository, J Digit Imaging, 26, 1045, 10.1007/s10278-013-9622-7

Scarpace, 2016, Radiology data from The Cancer Genome Atlas Glioblastoma Multiforme [TCGA-GBM] collection

Pedano, 2016, Radiology Data from The Cancer Genome Atlas Low Grade Glioma [TCGA-LGG] collection

Scarpace, 2016, Data from REMBRANDT

Starck, 2007, The undecimated wavelet decomposition and its reconstruction, IEEE Trans Image Process, 16, 297, 10.1109/TIP.2006.887733

Ojala, 2000, Gray scale and rotation invariant texture classification with local binary patterns, 10.1007/3-540-45054-8_27

Rister, 2017, Volumetric Image Registration From Invariant Keypoints, IEEE Trans Image Process, 26, 4900, 10.1109/TIP.2017.2722689

Cheung, 2009, N-SIFT: N-Dimensional Scale Invariant Feature Transform, IEEE Trans Image Process, 18, 2012, 10.1109/TIP.2009.2024578

Schölkopf, 2002, Learning with kernels: support vector machines, regularization, optimization, and beyond

Breiman, 2001, Random Forests, Machine Learn, 45, 5, 10.1023/A:1010933404324

Rätsch, 2001, Soft margins for AdaBoost, Machine Learn, 42, 287, 10.1023/A:1007618119488

Seiffert, 2010, RUSBoost: A hybrid approach to alleviating class imbalance, IEEE Transactions on Systems, 40, 185

Guyon, 2003, An introduction to variable and feature selection, J Machine Learn Res, 3, 1157

Matthews, 1975, Comparison of the predicted and observed secondary structure of T4 phage lysozyme, Biochim Biophys Acta, 405, 442, 10.1016/0005-2795(75)90109-9

Powers, 2011, Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation, J Mach Learn Tech, 2, 37

Mukaka, 2012, A guide to appropriate use of correlation coefficient in medical research, Malawi Med J, 24, 69

Hinkle, 2003, Applied statistics for the behavioral sciences

Upadhyay, 2011, Conventional MRI evaluation of gliomas, Br J Radiol, 84, S107, 10.1259/bjr/65711810

Scott, 2002, How often are nonenhancing supratentorial gliomas malignant? A population study, Neurology, 59, 947, 10.1212/WNL.59.6.947

Wiestler, 2013, ATRX loss refines the classification of anaplastic gliomas and identifies a subgroup of IDH mutant astrocytic tumors with better prognosis, Acta Neuropathol, 126, 443, 10.1007/s00401-013-1156-z

Law, 2008, Gliomas: predicting time to progression or survival with cerebral blood volume measurements at dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging, Radiology, 247, 490, 10.1148/radiol.2472070898

Law, 2004, Comparison of cerebral blood volume and vascular permeability from dynamic susceptibility contrast-enhanced perfusion MR imaging with glioma grade, Am J Neuroradiol, 25, 746

Choi, 2012, 2-hydroxyglutarate detection by magnetic resonance spectroscopy in IDH-mutated patients with gliomas, Nat Med, 18, 624, 10.1038/nm.2682

Van Cauter, 2012, Gliomas: diffusion kurtosis MR imaging in grading, Radiology, 263, 492, 10.1148/radiol.12110927

Raab, 2010, Cerebral gliomas: diffusional kurtosis imaging analysis of microstructural differences 1, Radiology, 254, 876, 10.1148/radiol.09090819

Pereira, 2015, Deep convolutional neural networks for the segmentation of gliomas in multi-sequence MRI, Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries

Nie, 2016, 3D deep learning for multi-modal imaging-guided survival time prediction of brain tumor patients, Medical Image Computing and Computer-Assisted Intervention – MICCAI 2016. Lecture Notes in Computer Science, 10.1007/978-3-319-46723-8_25

Shin, 2016, Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning, IEEE Trans Med Imaging, 35, 1285, 10.1109/TMI.2016.2528162