Domain Mapping and Deep Learning from Multiple MRI Clinical Datasets for Prediction of Molecular Subtypes in Low Grade Gliomas

Brain tumors, such as low grade gliomas (LGG), are molecularly classified which require the surgical collection of tissue samples. The pre-surgical or non-operative identification of LGG molecular type could improve patient counseling and treatment decisions. However, radiographic approaches to LGG molecular classification are currently lacking, as clinicians are unable to reliably predict LGG molecular type using magnetic resonance imaging (MRI) studies. Machine learning approaches may improve the prediction of LGG molecular classification through MRI, however, the development of these techniques requires large annotated data sets. Merging clinical data from different hospitals to increase case numbers is needed, but the use of different scanners and settings can affect the results and simply combining them into a large dataset often have a significant negative impact on performance. This calls for efficient domain adaption methods. Despite some previous studies on domain adaptations, mapping MR images from different datasets to a common domain without affecting subtitle molecular-biomarker information has not been reported yet. In this paper, we propose an effective domain adaptation method based on Cycle Generative Adversarial Network (CycleGAN). The dataset is further enlarged by augmenting more MRIs using another GAN approach. Further, to tackle the issue of brain tumor segmentation that requires time and anatomical expertise to put exact boundary around the tumor, we have used a tight bounding box as a strategy. Finally, an efficient deep feature learning method, multi-stream convolutional autoencoder (CAE) and feature fusion, is proposed for the prediction of molecular subtypes (1p/19q-codeletion and IDH mutation). The experiments were conducted on a total of 161 patients consisting of FLAIR and T1 weighted with contrast enhanced (T1ce) MRIs from two different institutions in the USA and France. The proposed scheme is shown to achieve the test accuracy of 74.81% on 1p/19q codeletion and 81.19% on IDH mutation, with marked improvement over the results obtained without domain mapping. This approach is also shown to have comparable performance to several state-of-the-art methods.

[1]  Maximilian Kohlbrenner,et al.  Pre-Training CNNs Using Convolutional Autoencoders , 2017 .

[2]  Chuan Li,et al.  Precomputed Real-Time Texture Synthesis with Markovian Generative Adversarial Networks , 2016, ECCV.

[3]  Tao Xu,et al.  SegAN: Adversarial Network with Multi-scale L1 Loss for Medical Image Segmentation , 2017, Neuroinformatics.

[4]  S. Klein,et al.  Predicting the 1p/19q Codeletion Status of Presumed Low-Grade Glioma with an Externally Validated Machine Learning Algorithm , 2019, Clinical Cancer Research.

[5]  D. Dong,et al.  Non-invasive genotype prediction of chromosome 1p/19q co-deletion by development and validation of an MRI-based radiomics signature in lower-grade gliomas , 2018, Journal of Neuro-Oncology.

[6]  Muazzam Maqsood,et al.  A Deep Siamese Convolution Neural Network for Multi-Class Classification of Alzheimer Disease , 2020, Brain sciences.

[7]  Jinhua Yu,et al.  A Universal Intensity Standardization Method Based on a Many-to-One Weak-Paired Cycle Generative Adversarial Network for Magnetic Resonance Images , 2019, IEEE Transactions on Medical Imaging.

[8]  Diego Castillo-Barnes,et al.  Studying the Manifold Structure of Alzheimer's Disease: A Deep Learning Approach Using Convolutional Autoencoders , 2020, IEEE Journal of Biomedical and Health Informatics.

[9]  Mark W. Woolrich,et al.  FSL , 2012, NeuroImage.

[10]  Ninon Burgos,et al.  New advances in the Clinica software platform for clinical neuroimaging studies , 2019 .

[11]  Seung Hong Choi,et al.  Gliomas: Histogram analysis of apparent diffusion coefficient maps with standard- or high-b-value diffusion-weighted MR imaging--correlation with tumor grade. , 2011, Radiology.

[12]  Jayaram K. Udupa,et al.  New variants of a method of MRI scale standardization , 2000, IEEE Transactions on Medical Imaging.

[13]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[14]  Mihaela van der Schaar,et al.  RadialGAN: Leveraging multiple datasets to improve target-specific predictive models using Generative Adversarial Networks , 2018, ICML.

[15]  Günther Palm,et al.  Learning convolutional neural networks from few samples , 2013, The 2013 International Joint Conference on Neural Networks (IJCNN).

[16]  G. Reifenberger,et al.  The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary , 2016, Acta Neuropathologica.

[17]  Pedro Costa,et al.  Towards Adversarial Retinal Image Synthesis , 2017, ArXiv.

[18]  Hayit Greenspan,et al.  Virtual PET Images from CT Data Using Deep Convolutional Networks: Initial Results , 2017, SASHIMI@MICCAI.

[19]  F. Wacker,et al.  A New Method for MRI Intensity Standardization with Application to Lesion Detection in the Brain , 2006 .

[20]  Jirí Sedlár,et al.  Predicting Deletion of Chromosomal Arms 1p/19q in Low-Grade Gliomas from MR Images Using Machine Intelligence , 2017, Journal of Digital Imaging.

[21]  N. Kaabouch,et al.  A Deep Learning approach for Diagnosis of Mild Cognitive Impairment Based on MRI Images , 2019, Brain sciences.

[22]  G. Collewet,et al.  Influence of MRI acquisition protocols and image intensity normalization methods on texture classification. , 2004, Magnetic resonance imaging.

[23]  Soumith Chintala,et al.  Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks , 2015, ICLR.

[24]  Yan Wang,et al.  Multimodal 3D DenseNet for IDH Genotype Prediction in Gliomas , 2018, Genes.

[25]  Irene Y. H. Gu,et al.  Multi-stream Convolutional Autoencoder and 2D Generative Adversarial Network for Glioma Classification , 2019, CAIP.

[26]  Wenya Linda Bi,et al.  Machine learning reveals multimodal MRI patterns predictive of isocitrate dehydrogenase and 1p/19q status in diffuse low- and high-grade gliomas , 2019, Journal of Neuro-Oncology.

[27]  Anders Eklund,et al.  Generative Adversarial Networks for Image-to-Image Translation on Multi-Contrast MR Images - A Comparison of CycleGAN and UNIT , 2018, ArXiv.

[28]  Khalid Masood Khan,et al.  Brain Tumor Analysis Empowered with Deep Learning: A Review, Taxonomy, and Future Challenges , 2020, Brain sciences.

[29]  Arno Klein,et al.  A reproducible evaluation of ANTs similarity metric performance in brain image registration , 2011, NeuroImage.

[30]  Zeeshan Syed,et al.  Adapting Surgical Models to Individual Hospitals Using Transfer Learning , 2012, 2012 IEEE 12th International Conference on Data Mining Workshops.

[31]  Boudewijn P. F. Lelieveldt,et al.  Inter‐station intensity standardization for whole‐body MR data , 2016, Magnetic resonance in medicine.

[32]  K. Masamune,et al.  Prediction of lower-grade glioma molecular subtypes using deep learning , 2019, Journal of Neuro-Oncology.

[33]  Arie Perry,et al.  Molecular Diagnostics in Central Nervous System Tumors , 2005, Advances in anatomic pathology.

[34]  Baojuan Li,et al.  Radiomics Strategy for Molecular Subtype Stratification of Lower‐Grade Glioma: Detecting IDH and TP53 Mutations Based on Multimodal MRI , 2018, Journal of magnetic resonance imaging : JMRI.

[35]  Yuanyuan Wang,et al.  Noninvasive IDH1 mutation estimation based on a quantitative radiomics approach for grade II glioma , 2017, European Radiology.

[36]  I. Mader,et al.  Surgical management of lower-grade glioma in the spotlight of the 2016 WHO classification system , 2018, Journal of Neuro-Oncology.

[37]  S. Klein,et al.  Differences in spatial distribution between WHO 2016 low-grade glioma molecular subgroups , 2019, Neuro-oncology advances.

[38]  Jenna Wiens,et al.  A study in transfer learning: leveraging data from multiple hospitals to enhance hospital-specific predictions , 2014, J. Am. Medical Informatics Assoc..

[39]  Luc Van Gool,et al.  Deep Temporal Linear Encoding Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[40]  拓海 杉山,et al.  “Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告 , 2017 .