Diffusion & Perfusion Imaging Protocols for Gliomas
Kathleen M Schmainda1

1Radiology, Medical College of Wisconsin, Milwaukee, WI, United States

Synopsis

We have much to gain by greater incorporation of advanced physiologic MRI methods such as diffusion MRI (DWI) and dynamic susceptibility contrast perfusion MRI (DSC-pMRI) methods into the treatment management protocols for patients with glioma. To motivate greater use this course will describe how these methods can be used at several critical junctures in the management of patients with glioma. Current questions and limitations, both scientific and technical, will also be discussed.

Highlights:

  • Use of dynamic susceptibility contrast perfusion MRI (DSC-pMRI) methods and DWI (diffusion MRI) methods to guide tumor tissue sampling may be necessary to achieve the goal of the 2016 CNS WHO classification for greater diagnostic accuracy.
  • The ability of DSC-pMRI methods to distinguish low-grade from high-grade glioma depends on the way the DSC-MRI data is collected and processed. Consideration of the contrast agent leakage effects, which occurs with a disrupted blood brain barrier, is an important factor that is driving the consensus protocol recommendations.
  • After undergoing radiotherapy (RT) with concurrent and adjuvant temozolomide (TMZ), it can be difficult to distinguish tumor from treatment effect (TE) using standard anatomic MRI alone. Relative cerebral blood volume (rCBV) derived from DSC-pMRI has demonstrated the ability to make this distinction. ·
  • The number of studies demonstrating the utility of rCBV to predict response to bevacizumab are increasing, suggesting that bevacizumab does prolong overall survival in a subpopulation of patients. This demonstrates how advanced imaging methods can significantly impact treatment management paradigms.
  • Diffusion MRI has demonstrated the ability to predict response to chemotherapy and detect invading tumor cells, which are invisible with other imaging modalities and methods. Yet changes in edema, or the presence of coagulative necrosis, can confound interpretation and require further study and technical innovation.
  • Response to immunotherapies are confounded by inflammation that can result in new FLAIR hyperintensity or lesion enhancement on T1w images, without true progressing tumor. These confounding factors motivate a new role for pMRI and DWI, for immunotherapy and other new therapies such as tumor treating-fields.

Target Audience:

Persons interested in the care and management of patients with gliomas including imaging students and scientists, neuro-oncologists, neurosurgeons and neuroradiologists.

Objective:

The management of patients with gliomas largely relies on MRI for detection, surgical and radiation treatment (RT) planning and evaluation of treatment response, progression or recurrence. For the most part anatomic and not advanced physiologic MRI methods have been used. The anatomic MRI scans typically include pre- and post-contrast T1 weighted images to identify contrast-agent enhancing lesions and FLAIR (fluid attenuated inversion recovery) images to delineate tumor-associated edema. While these anatomic methods form the basis of the response assessment criteria for high-grade gliomas it has it becoming increasingly apparent that there are significant limitations to using anatomic MRI methods alone 1. Therefore, we have much to gain by greater incorporation of advanced physiologic MRI methods such as diffusion MRI (DWI) and perfusion MRI (pMRI) methods into the treatment management protocols for patients with glioma. In this context, the objective of this course is to describe how DWI and dynamic susceptibility contrast pMRI (DSC-pMRI) methods can be used at several critical junctures in the management of glioma patients. In addition, current questions and limitations, both scientific and technical, will be discussed to further motivate research in these areas.

Background:

The role of pMRI and DWI at the time of diagnosis will first be presented. This will be followed by the present and future roles that each method can play during treatment follow-up, first in the context of chemo-radiation therapy, next for predicting response to bevacizumab treatment for recurrent glioblastoma and finally in the context of newer treatments that include immunotherapies and tumor-treating fields.

i. DSC-pMRI and DWI at diagnosis For the first time in over a century an entirely new classification scheme for the diagnosis of both adult and pediatric CNS (central nervous system) tumors has been released (2,3). This new 2016 CNS WHO (World Health Organization) classification adds molecular parameters to the histopathological characteristics that have comprised the WHO standard for decades. It is hoped that this new classification will lead to greater diagnostic accuracy, as well as improved patient management and more accurate determination of prognosis and treatment response. Yet, this promise can only be fully realized with appropriate tissue sampling. In this context pMRI and DWI markers may be of help, an idea supported by several recent studies where both the apparent diffusion coefficient (ADC) derived from DWI, and relative cerebral blood volume (rCBV) derived from DSC-pMRI, predicted differences in the expression of several molecular markers and/or directed tissue sampling to areas of most concern (4,5).

At diagnosis it can also be difficult to distinguish low-grade from high-grade glioma using standard MRI. Though histopathology remains the gold standard for this determination, imaging also can serve a role in guiding this diagnosis. In response, several groups have demonstrated that brain tumor rCBV, which is derived from DSC-pMRI, correlates with tumor grade. Yet there is much evidence to indicate that this correlation is observed only when the DSC-MRI data is collected and processed in a way that appropriately accounts for contrast agent leakage effects. Specifically, gadolinium (Gd)-based MRI contrast agents “leak” out of the brain vasculature if the blood brain barrier (BBB) is disrupted, which is often the case with brain tumors. These leakage effects can lead to inaccuracies in the determination of rCBV and a loss of rCBV correlation with tumor grade. Both preclinical and clinical studies have validated the reliability of a leakage-corrected approach. In fact, a study addressing multi-site concordance of DSC-MRI analysis for brain tumors, presented at this meeting (6), suggests that when a preload of contrast agent is used before collecting DSC-MRI data there is excellent agreement across sites and platforms in their ability to distinguish low-grade from high-grade brain tumor. Taking these findings into consideration, the National Brain Tumor Society (NBTS) dynamic susceptibility contrast (DSC) MRI working group is in the process of finalizing the details of a recommended consensus DSC-MRI protocol. These recommendations will be briefly outlined.

ii. DSC-pMRI and DWI for treatment monitoring Treatment Follow-Up consists of MRI scans every 6-8 weeks or 3-6 months, depending on diagnosis of high-grade or low-grade glioma as well as symptoms and adjuvant therapy. Standard therapy for high-grade brain tumor includes maximal safe tumor resection followed by radiotherapy (RT) with concurrent and adjuvant temozolomide, a chemotherapeutic (1). In 20-30% of patients, their first post-RT MRI shows increased contrast agent enhancement that eventually subsides without any change in therapy. This phenomenon known as pseudoprogression is thought to result from transient increases in tumor vascular permeability resulting from RT that may be further enhanced by TMZ (1). This treatment-related effect has significant implications for patient management. It can result in premature discontinuation of therapy and/or improper enrollment or exclusion from clinical trials. Clearly, a biomarker that can distinguish tumor from treatment effect would be of substantial benefit.

Recent studies have demonstrated that DSC-pMRI methods have this potential to contribute information necessary to make these distinctions. Specifically rCBV image maps have demonstrated the ability to distinguish treatment effect (TE) from recurrent tumor (7-9) and pseudoprogression from true progression (10). While studies report different thresholds, to distingusih tumor from treatment effects, more recent studies, which have used image-guided neuronavigation to spatially correlate rCBV values with specimen histopathology are predicting a normalized rCBV (nRCBV) threshold close to 1 to distinguish tumor from TE (8,9). This distinction has led to another promising biomarker, called fractional tumor burden (FTB)(11), which identifies the portion of enhancing lesion that is true tumor. This metric has demonstrated promise to predict outcomes in the setting of both chemo-radiation therapy (12) and bevacizumab treatment (13). Finally, several groups (14-17) have demonstrated the potential of rCBV to predict response to anti-angiogenic therapy more reliably than standard MRI in both single center (18-22) and multi-center trials (23,24).

DWI methods are also routinely collected for the assessment of brain tumors and their response to therapies. DWI is sensitive to the molecular motion of water, and thus sensitive to diffusion restrictions such as cell membranes. Because of this, the apparent diffusion coefficient (ADC) calculated from DWI, has been shown to correlate with cell density exhibiting a marked decrease in regions with an increase in cellularity (i.e. hypercellularity) (25,26), and are therefore routinely used in the assessment of tumor progression. Yet, recent studies indicate that in patients treated with bevacizumab, a new category of ultra-low ADC, confirmed to be coagulative necrosis, may not indicate progression but rather treatment response (27,28).

Functional diffusion maps (fDMs), also referred to as parametric response maps (PRMs), were developed to examine voxel-by-voxel changes in ADC (29-31), thus providing spatial localization of changing cell density. This measure has demonstrated the ability to predict response to chemotherapy when evaluating the primary tumor burden (26,32,33) as well as predict response to both chemo-RT, as well as bevacizumab when assessing infiltrating tumor outside of the primary enhancing lesion (34-37). The technical challenges that may be limiting a more widespread adoption of this approach include inconsistencies in diffusion values across vendors and scanner platforms, and the ability to sufficiently register diffusion images across time and with tumor changes.


Conclusions:

No matter how many new drugs and new pathways are discovered or targeted for the treatment of brain tumors, changes in tumor cell density and vascularization will always be key factors indicating response or failure. Given that perfusion and diffusion MRI are unique biomarkers that provide this information it would seem that these methods should be used more routinely in the diagnosis and management of brain tumors. Therefore, it is hoped that the information and recommendations provided in this course will lead to a greater understanding and translation of these methods into regular clinical use.

Acknowledgements

Funding support provided by NIH/NCI R01 CA082500, NIH/NCI U01 CA176110 and the many investigators cited and uncited who have made significant contributions to this field of study.

References

1. Wen PY, et. al. J Clin Oncol 2010;28(11):1963-1972.

2. Louis DN, et. al.. Acta neuropathologica 2016;131(6):803-820.

3. Rodriguez FJ, et. al. J Mol Diagn 2016;18(5):620-634.

4. Pedeutour-Braccini Z, et. al.. Virchows Arch 2015;466(4):433-444.

5. Maeda M, I et. al. Radiology 1993;189:233-238.

6. Schmainda KM, et. al.. ISMRM 2017; Honolulu, Hawaii.

7. Sugahara T, et. al. AJNR Am J Neuroradiol 2000;21(5):901-909.

8. Hu L, et. al. Am J Neuroradiol 2009;30(3):552-558.

9. Prah MA, et. al.. ISMRM 2015; Toronto, Ontario, Candada. p 70.

10. Kong DS, et. al.. Am J Neuroradiol 2011;32:382-387.

11. Hu LS, et al. Neuro Oncol 2012;14(7):919-930.

12. Prah MA, et. al. ISMRM. 2017; Honolulu, Hawaii. p 707.

13. Prah MA, et. al. ISMRM 2017; Honolulu, Hawaii. p 708.

14. Pechman KR, et. al. J Neurooncol 2011;Epub Ahead of print.

15. Darpolor MM, et. al. PLOS One 2011;6(1):e16621.

16. Badruddoja MA, et. al.. Neuro-Oncology 2003;5(4):235-243.

17. Quarles CC, Schmainda KM. Magnetic Resonance in Medicine 2007;57(4):680-687.

18. Schmainda KM, et. al. Neuro Oncol 2014;16(6):880-888.

19. Kickingereder P, et. al. Neuro Oncol 2015;17(8):1139-1147.

20. Harris RJ, et. al. J Neurooncol 2015;122(3):497-505.

21. Liu TT, et. al. Neuro Oncol 2016.

22. Bennett IE, et. al. J Neurooncol 2017;131(2):321-329.

23. Schmainda KM, et. al. Neuro Oncol 2015;17(8):1148-1156.

24. Gerstner ER, et. al. Clin Cancer Res 2016;22(20):5079-5086.

25. Sugahara T, et. al. Topics in Magnetic Resonance Imaging 1999;10(2):114-124.

26. Chenevert TL, et. al. J Natl Cancer Institute 2000;92:2029-2036.

27. Nguyen, HS, et. al. Am J Neurorad 2016 37(12); 2201-2208.

28. Rieger J, et. al. J Neurooncol 2010;99(1):49-56.

29. Hamstra DA, et. al. Proc Natl Acad Sci USA 2005;102(46):16759-16764.

30. Hamstra DA, et. al. J Clin Oncol 2008;26(20):3387-3394.

31. Moffat BA, et. al. Neoplasia 2006;8(4):259-267.

32. Tsien C, et. al. J Clin Oncol 2010;28(13):2293-2299.

33. Galban S, et. al. PLOS One 2012;7(4).

34. Ellingson BM, et. al. Neuro Oncol 2011;13(10):1151-1161.

35. Ellingson BM, et. al. J Neurooncol 2010;97(3):419-423.

36. Cohen AD, et. al.. J Magn Reson Imaging 2013;38(4):868-875.

37. Ellingson BM, et. al. J Neurooncol 2011;102(1):95-103.


Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)