Elsevier

Methods

Volume 188, April 2021, Pages 112-121
Methods

Radiomics in neuro-oncology: Basics, workflow, and applications

https://doi.org/10.1016/j.ymeth.2020.06.003Get rights and content

Highlights

  • Radiomics is increasingly used and evaluated in patients with brain tumors.

  • Radiomics extracts additional information from routinely acquired imaging data.

  • Generated models allow prediction of, e.g., treatment response or molecular markers.

  • Radiomics adds important diagnostic information to highly relevant clinical questions.

Abstract

Over the last years, the amount, variety, and complexity of neuroimaging data acquired in patients with brain tumors for routine clinical purposes and the resulting number of imaging parameters have substantially increased. Consequently, a timely and cost-effective evaluation of imaging data is hardly feasible without the support of methods from the field of artificial intelligence (AI). AI can facilitate and shorten various time-consuming steps in the image processing workflow, e.g., tumor segmentation, thereby optimizing productivity. Besides, the automated and computer-based analysis of imaging data may help to increase data comparability as it is independent of the experience level of the evaluating clinician. Importantly, AI offers the potential to extract new features from the routinely acquired neuroimages of brain tumor patients. In combination with patient data such as survival, molecular markers, or genomics, mathematical models can be generated that allow, for example, the prediction of treatment response or prognosis, as well as the noninvasive assessment of molecular markers. The subdiscipline of AI dealing with the computation, identification, and extraction of image features, as well as the generation of prognostic or predictive mathematical models, is termed radiomics. This review article summarizes the basics, the current workflow, and methods used in radiomics with a focus on feature-based radiomics in neuro-oncology and provides selected examples of its clinical application.

Introduction

The diagnosis of brain cancer is predominantly based on neuroimaging findings, and, ultimately, the histomolecular evaluation of tissue samples obtained from tumor resection or biopsy. For decades, mostly anatomical neuroimaging techniques such as contrast-enhanced computed tomography (CT) or magnetic resonance imaging (MRI) have been used for brain tumor diagnostics, treatment planning, and follow-up. More recently, the increasing number of additional imaging parameters primarily derived from advanced MRI and amino acid PET, as well as technical developments such as the event of hybrid PET/CT and PET/MRI scanners, generate a large amount of complex neuroimaging data in patients with brain tumors.

A timely evaluation of this amount of diagnostic information that potentially can be implemented in clinical routine is costly and hardly feasible without considerable computer support. Here, methods from the emerging field of artificial intelligence (AI) offer new options to support clinicians further. In particular, AI provides the possibilities to partially or fully automate various steps within the diagnostic routine, so that especially time-consuming processes such as the manual detection and segmentation of lesions are performed by a computer and require only a final validation by a clinician. Furthermore, the speed of image processing and analysis can be enhanced using AI-based methods, thereby increasing productivity. Besides, the automated AI-based analysis of imaging data may help to increase the comparability of the obtained results as it is independent of the experience level of the evaluating clinician.

Moreover, AI offers the potential to extract yet undiscovered features from routinely acquired images. Specifically, quantitative and semi-quantitative image features can be extracted from routinely acquired neuroimaging data, which are usually beyond human perception. Finally, subsets of these image features, combined with patient information such as survival data, molecular markers, or genomics, can be used to develop mathematical models that characterize the underlying brain tumor biology. Subsequently, these models can be used for essential clinical questions, e.g., the assessment of prognosis or treatment response, as well as the noninvasive diagnosis of molecular markers. The computation, identification, and extraction of image features, as well as the generation of prognostic or predictive mathematical models, is summarized under the term radiomics [1], [2], [3], [4], [5], a specialized application within the broad field of AI.

This review article summarizes the basics, the current workflow, and methods used in radiomics with a focus on feature-based radiomics in neuro-oncology and provides selected examples of its clinical application.

Section snippets

Radiomics

“Images are more than pictures, they are data”. This intuitive and precise definition by Robert Gillies and colleagues nicely illustrates the basic idea of radiomics [3]. Radiomics can be subdivided into feature-based and deep learning-based radiomics and is usually applied to routinely acquired imaging data, thereby allowing additional data analysis at a low cost. Since radiomics features are either mathematically predefined (feature-based radiomics) or generated from the data by training

Applications of radiomics in Neuro-Oncology

Radiomics in patients with brain tumors is mainly based on the analysis of conventional MRI. Several studies have investigated the usefulness of radiomics for the differentiation of treatment-related changes from tumor progression in patients with gliomas and brain metastases, which is a clinical question of considerable importance. Furthermore, several studies have also evaluated radiomics for the classification as well as the molecular characterization of brain tumors, which is of high

Conclusions

In summary, feature-based and deep learning-based radiomics is increasingly evaluated in the field of neuro-oncology. Radiomics should be considered as an additional source of diagnostic information that, especially in combination with clinical, histopathological, molecular, and conventional imaging parameters, has a great potential to significantly improve the diagnostics and management of patients with brain tumors.

Notwithstanding, most studies lack further validation of the generated models

References (71)

  • L. Peng et al.

    Distinguishing true progression from radionecrosis after stereotactic radiation therapy for brain metastases with machine learning and radiomics

    Int. J. Radiat. Oncol. Biol. Phys.

    (2018)
  • P. Lohmann et al.

    Combined FET PET/MRI radiomics differentiates radiation injury from recurrent brain metastasis

    Neuroimage Clin.

    (2018)
  • P.M. Szczypinski et al.

    MaZda–a software package for image texture analysis

    Comput. Methods Programs Biomed.

    (2009)
  • P. Lambin et al.

    Radiomics: the bridge between medical imaging and personalized medicine

    Nat. Rev. Clin. Oncol.

    (2017)
  • R.J. Gillies et al.

    Radiomics: images are more than pictures they are data

    Radiology

    (2016)
  • H.J. Aerts et al.

    Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach

    Nat. Commun.

    (2014)
  • S. Rizzo et al.

    Radiomics: the facts and the challenges of image analysis

    Eur. Radiol. Exp.

    (2018)
  • S.S. Yip et al.

    Applications and limitations of radiomics

    Phys. Med. Biol.

    (2016)
  • W. Deng et al.

    Brain tumor segmentation based on improved convolutional neural network in combination with non-quantifiable local texture feature

    J. Med. Syst.

    (2019)
  • P. Prasanna, A. Karnawat, M. Ismail, A. Madabhushi, P. Tiwari, Radiomics-based convolutional neural network for brain...
  • E. Grovik et al.

    Deep learning enables automatic detection and segmentation of brain metastases on multisequence MRI

    J. Magn. Reson. Imaging

    (2020)
  • R.M. Haralick et al.

    Textural features for image classification

    IEEE Trans. Syst. Man. Cyb Smc

    (1973)
  • M. Amadasun et al.

    Textural features corresponding to textural properties

    IEEE Trans. Syst. Man Cyb

    (1989)
  • D.-H. Xu, A.S. Kurani, J.D. Furst, D.S. Raicu, Run-length encoding for volumetric texture, Conference Proeedings from...
  • Y. Zhang et al.

    Radiomics-based prognosis analysis for non-small cell lung cancer

    Sci. Rep.

    (2017)
  • V. Parekh et al.

    Radiomics: a new application from established techniques

    Expert Rev. Precis Med. Drug Dev.

    (2016)
  • M. Kuhn et al.

    Applied Predictive Modeling

    (2013)
  • A. Krizhevsky et al.

    Imagenet classification with deep convolutional neural networks, Advances in Neural

    Inform. Process. Syst.

    (2012)
  • Y.J. Cha et al.

    Prediction of response to stereotactic radiosurgery for brain metastases using convolutional neural networks

    Anticancer Res.

    (2018)
  • C. Tan, F. Sun, T. Kong, W. Zhang, C. Yang, C. Liu, A Survey on Deep Transfer Learning, arXiv e-prints,...
  • D.N. Louis et al.

    The 2016 world Health organization classification of tumors of the central nervous system: a summary

    Acta Neuropathol.

    (2016)
  • C. Hartmann et al.

    Patients with IDH1 wild type anaplastic astrocytomas exhibit worse prognosis than IDH1-mutated glioblastomas, and IDH1 mutation status accounts for the unfavorable prognostic effect of higher age: implications for classification of gliomas

    Acta Neuropathol.

    (2010)
  • M.C. Chamberlain et al.

    Prognostic significance of relative 1p/19q codeletion in oligodendroglial tumors

    J. Neurooncol.

    (2015)
  • M.E. Hegi et al.

    Correlation of O6-methylguanine methyltransferase (MGMT) promoter methylation with clinical outcomes in glioblastoma and clinical strategies to modulate MGMT activity

    J. Clin. Oncol.

    (2008)
  • H.H. Cho et al.

    Classification of the glioma grading using radiomics analysis

    PeerJ

    (2018)
  • Cited by (78)

    View all citing articles on Scopus
    View full text