摘要
HomeRadiologyVol. 290, No. 2 PreviousNext Reviews and CommentaryFree AccessEditorialMRI of Recurrent Glioblastoma: Reliability and RealityRonald L. Wolf Ronald L. Wolf Author AffiliationsFrom the Department of Radiology, Neuroradiology Section, Perelman School of Medicine, University of Pennsylvania Health System, Dulles 219, 3400 Spruce St, Philadelphia, PA 19104.Address correspondence to the author (e-mail: [email protected]).Ronald L. Wolf Published Online:Nov 27 2018https://doi.org/10.1148/radiol.2018182530MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In See also the article by Barboriak et al in this issue.IntroductionNeuroimaging plays a central role in diagnosis, treatment planning, and posttreatment evaluation in the setting of glioblastoma, and complexity in imaging approach and interpretation is rapidly increasing. On posttreatment MR images, findings concerning for progression may actually represent primarily treatment-related effects (eg, pseudoprogression or more delayed treatment-related necrosis). In practice, variable fractions of tumor and treatment effect are usually present. An apparent “positive” response on imaging studies can also be seen in spite of neoplastic progression (pseudoresponse). Conventional MRI relying primarily on fluid-attenuated inversion recovery (FLAIR) and pre- and postcontrast T1-weighted acquisitions is the cornerstone of imaging surveillance. Along with clinical assessment, MRI forms the basis for clinical follow-up and response criteria in clinical trials such as the Response Assessment in Neuro-oncology (RANO) criteria, the successor to the original Macdonald criteria (1). As understanding of the disease and treatment strategies evolve, including antiangiogenic and immunotherapies, modifications to response criteria have become necessary (eg, modified, or mRANO, and immunotherapy, or iRANO) (2,3). Standardized imaging protocols have been proposed in support of these efforts (4).Conventional MRI is limited in tumor assessment before and after therapy. Physiologic MRI techniques, including dynamic susceptibility contrast perfusion (DSC) MRI and dynamic contrast-enhanced permeability (DCE) MRI, have been incorporated into clinical and trial protocols to augment evaluation and improve accuracy (5). The need for standardization in physiologic imaging has also become evident and guidelines are becoming available (6). Working to standardize MRI acquisition protocols across many institutions with different scanner configurations, field strength, and software capabilities is a substantial challenge. Even after standardization of image acquisition, remaining obstacles include variability in postprocessing approaches as well as interobserver and interinstitutional variability.DCE MRI and DSC MRI are the two primary contrast-enhanced advanced MRI strategies. Experience with DSC MRI is greater, but experience with DCE MRI is growing. In comparison, DCE MRI has some advantages, including higher spatial resolution and a relative insensitivity to susceptibility artifacts due to surgical hardware, blood products, and/or air interfaces near skull base. Disadvantages of DCE MRI include longer scan time and greater complexity in pharmacokinetic modeling for calculating permeability metrics (7). Expert recommendations for standardization of DCE MRI have been published (8), but challenges remain for ensuring reliable use of DCE MRI. These include standardization of quantitative metrics across different vendors and imaging systems, different institutions, and multiple users. The complexity of glioblastoma and associated treatment-related changes in response to standard and experimental therapies creates additional challenges. The RSNA Quantitative Imaging Biomarkers Alliance (QIBA) proposed quantitative imaging for solid neoplasms including glioma, and profiles in development include DCE MRI. A detailed approach for standardization can be found on the QIBA site (https://www.rsna.org/research/quantitative-imaging-biomarkers-alliance or http://qibawiki.rsna.org/images/1/12/DCE-MRI_Quantification_Profile_v1.0.pdf).In this issue of Radiology, Barboriak and colleagues present an analysis of DCE MRI data obtained as part of ACRIN 6677, which studied recurrent glioblastoma treated with bevacizumab with or without irinotecan (9). All DCE MRI data were processed centrally and image analysis subsequently performed by two teams, each with a technologist and a neuroradiologist. The technologist presegmented enhancing tissue and this was adjusted by the neuroradiologist. From this volume, multiple permeability metrics (Ktrans, IAUGC, and ve) were compared between these reading groups. They found relatively high variability in interreader variation (IRV) for permeability metrics on average exceeding 16%. Posttreatment variability was greater than for baseline assessments. Reader segmentation of enhancing tumor volume was a major source of variability.Variability in tumor segmentation is likely related to bevacizumab administration and alterations in the enhancement pattern that occur with antiangiogenic agents. As the authors note, treatment with bevacizumab leads to uncertainty in the definition of enhancing margins. This has been my experience as well.Another limitation is the segmentation strategy. Recent modifications to RANO incorporate volumetric measures. Enhanced three-dimensional (3D) T1-weighted series were manually segmented by the ACRIN investigators. However, standardized imaging protocols also include matched precontrast 3D T1-weighted series, and subtractions for segmentation could have been used to decrease variability (2). Automated segmentations may have improved results and are ultimately preferable (and probably inevitable). Although a standardized postprocessing strategy was used for the DCE MRI data, variation in data acquisition was still present, including different vendors, field strengths, and scan parameters. Also, heterogeneity in the tissue itself can contribute to variability. The enhancing tissue in a posttreatment glioblastoma setting is usually a mixture of viable neoplasm, treated nonproliferating neoplasm, and nonneoplastic treatment effect, which may also contribute to the effect of segmentation volume on variability. The ACRIN investigators focused on median and mean descriptive statistics from the full enhancing volume, but other statistical descriptors to address tissue heterogeneity may be informative. Finally, the number of patients studied (n = 21) is relatively small.As cancer therapy evolves and more sophisticated targeted therapies are developed, response assessment must also evolve. Conventional postcontrast T1-weighted MRI is still necessary, but reliability in assessment still needs to be improved. Advanced imaging modalities like DCE MRI will need to be thoughtfully incorporated into response assessment protocols and must also be reliable (9,10). Limitations aside, the implications of the study by Barboriak et al are important not only for clinical trials but also for our routine clinical interpretations. Given the high level of variability of DCE MRI, we still have work to do to effectively use such metrics as biomarkers. Continued efforts to standardize imaging protocols and MRI analysis should be supplemented by automated image analysis methods to improve our ability to predict glioblastoma response to therapy.Disclosures of Conflicts of Interest: R.L.W. disclosed no relevant relationships.References1. Ellingson BM, Chung C, Pope WB, Boxerman JL, Kaufmann TJ. Pseudoprogression, radionecrosis, inflammation or true tumor progression? challenges associated with glioblastoma response assessment in an evolving therapeutic landscape. J Neurooncol 2017;134(3):495–504. Crossref, Medline, Google Scholar2. Ellingson BM, Wen PY, Cloughesy TF. Modified Criteria for Radiographic Response Assessment in Glioblastoma Clinical Trials. Neurotherapeutics 2017;14(2):307–320. Crossref, Medline, Google Scholar3. Wen PY, Chang SM, Van den Bent MJ, Vogelbaum MA, Macdonald DR, Lee EQ. Response Assessment in Neuro-Oncology Clinical Trials. J Clin Oncol 2017;35(21):2439–2449. Crossref, Medline, Google Scholar4. Ellingson BM, Bendszus M, Boxerman J, et al. Consensus recommendations for a standardized Brain Tumor Imaging Protocol in clinical trials. Neuro Oncol 2015;17(9):1188–1198. Medline, Google Scholar5. Shiroishi MS, Boxerman JL, Pope WB. Physiologic MRI for assessment of response to therapy and prognosis in glioblastoma. Neuro Oncol 2016;18(4):467–478. Crossref, Medline, Google Scholar6. Welker K, Boxerman J, Kalnin A, et al. ASFNR recommendations for clinical performance of MR dynamic susceptibility contrast perfusion imaging of the brain. AJNR Am J Neuroradiol 2015;36(6):E41–E51. Crossref, Medline, Google Scholar7. Huang RY, Neagu MR, Reardon DA, Wen PY. Pitfalls in the neuroimaging of glioblastoma in the era of antiangiogenic and immuno/targeted therapy - detecting illusive disease, defining response. Front Neurol 2015;6:33. Crossref, Medline, Google Scholar8. Leach MO, Brindle KM, Evelhoch JL, et al. Assessment of antiangiogenic and antivascular therapeutics using MRI: recommendations for appropriate methodology for clinical trials. Br J Radiol 2003;76(Spec No 1):S87–S91. Crossref, Medline, Google Scholar9. Barboriak DP, Zhang Z, Desai P, et al. Interreader variability of dynamic contrast-enhanced MRI of recurrent glioblastoma: The Multicenter ACRIN 6677/RTOG 0625 Study. Radiology 2019;290:467–476. Link, Google Scholar10. Yankeelov TE, Mankoff DA, Schwartz LH, et al. Quantitative Imaging in Cancer Clinical Trials. Clin Cancer Res 2016;22(2):284–290. Crossref, Medline, Google ScholarArticle HistoryReceived: Nov 1 2018Revision requested: Nov 5 2018Revision received: Nov 7 2018Accepted: Nov 8 2018Published online: Nov 27 2018Published in print: Feb 2019 FiguresReferencesRelatedDetailsCited ByLearning-based analysis of amide proton transfer-weighted MRI to identify true progression in glioma patientsPengfeiGuo, MathiasUnberath, Hye-YoungHeo, Charles G.Eberhart, MichaelLim, Jaishri O.Blakeley, ShanshanJiang2022 | NeuroImage: Clinical, Vol. 35Accompanying This ArticleInterreader Variability of Dynamic Contrast-enhanced MRI of Recurrent Glioblastoma: The Multicenter ACRIN 6677/RTOG 0625 StudyNov 27 2018RadiologyRecommended Articles Longitudinal Assessment of Posttreatment Diffuse Glioma Tissue Volumes with Three-dimensional Convolutional Neural NetworksRadiology: Artificial Intelligence2022Volume: 4Issue: 5World Health Organization Grade II/III Glioma Molecular Status: Prediction by MRI Morphologic Features and Apparent Diffusion CoefficientRadiology2020Volume: 296Issue: 1pp. 111-121The University of California San Francisco Preoperative Diffuse Glioma MRI DatasetRadiology: Artificial Intelligence2022Volume: 4Issue: 6Emerging Applications of Artificial Intelligence in Neuro-OncologyRadiology2019Volume: 290Issue: 3pp. 607-618Robust Deep Learning–based Segmentation of Glioblastoma on Routine Clinical MRI Scans Using Sparsified TrainingRadiology: Artificial Intelligence2020Volume: 2Issue: 5See More RSNA Education Exhibits Technical Optimization of Contrast Enhanced Sequences in Central Nervous System: Making the Most for Every Single Gadolinium DropDigital Posters2018Response Assessment in Neuro-Oncology (RANO) Criteria for Gliomas: Practical Topics Using Conventional and Advanced Imaging TechniquesDigital Posters2018âPercentage Signal Intensity Recoveryâ: An Underutilised MR Perfusion Parameter in the Differentiation of Common Brain Tumors!Digital Posters2018 RSNA Case Collection Trigeminal schwannomaRSNA Case Collection2020Creutzfeldt-Jakob DiseaseRSNA Case Collection2021Primary intraosseous meningiomaRSNA Case Collection2020 Vol. 290, No. 2 Metrics Altmetric Score PDF download