Skip to main content
Log in

Feasibility of multi-parametric PET and MRI for prediction of tumour recurrence in patients with glioblastoma

  • Original Article
  • Published:
European Journal of Nuclear Medicine and Molecular Imaging Aims and scope Submit manuscript

Abstract

Background

Recurrence in glioblastoma patients often occur close to the original tumour and indicates that the current treatment is inadequate for local tumour control. In this study, we explored the feasibility of using multi-modality imaging at the time of radiotherapy planning. Specifically, we aimed to identify parameters from pre-treatment PET and MRI with potential to predict tumour recurrence.

Materials and methods

Sixteen patients were prospectively recruited and treated according to established guidelines. Multi-parametric imaging with 18F-FET PET/CT and 18F-FDG PET/MR including diffusion and dynamic contrast enhanced perfusion MRI were performed before radiotherapy. Correlations between imaging parameters were calculated. Imaging was related to the voxel-wise outcome at the time of tumour recurrence. Within the radiotherapy target, median differences of imaging parameters in recurring and non-recurring voxels were calculated for contrast-enhancing lesion (CEL), non-enhancing lesion (NEL), and normal appearing grey and white matter. Logistic regression models were created to predict the patient-specific probability of recurrence. The most important parameters were identified using standardized model coefficients.

Results

Significant median differences between recurring and non-recurring voxels were observed for FDG, FET, fractional anisotropy, mean diffusivity, mean transit time, extra-vascular, extra-cellular blood volume and permeability derived from scans prior to chemo-radiotherapy. Tissue-specific patterns of voxel-wise correlations were observed. The most pronounced correlations were observed for 18F-FDG- and 18F-FET-uptake in CEL and NEL. Voxel-wise modelling of recurrence probability resulted in area under the receiver operating characteristic curve of 0.77 from scans prior to therapy. Overall, FET proved to be the most important parameter for recurrence prediction.

Conclusion

Multi-parametric imaging before radiotherapy is feasible and significant differences in imaging parameters between recurring and non-recurring voxels were observed. Combining parameters in a logistic regression model enabled patient-specific maps of recurrence probability, where 18F-FET proved to be most important. This strategy could enable risk-adapted radiotherapy planning.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Similar content being viewed by others

References

  1. Stummer W, Pichlmeier U, Meinel T, Wiestler OD, Zanella F, Reulen H-J, et al. Fluorescence-guided surgery with 5-aminolevulinic acid for resection of malignant glioma: a randomised controlled multicentre phase III trial. Lancet Oncol. 2006;7:392–401.

    Article  CAS  PubMed  Google Scholar 

  2. Munck Af Rosenschöld P, Engelholm SSA, Ohlhues L, Law I, Vogelius I, Engelholm SSA. Photon and proton therapy planning comparison for malignant glioma based on CT, FDG-PET, DTI-MRI and fiber tracking. Acta Oncol. 2011;50:777–83.

    Article  PubMed  Google Scholar 

  3. Clarke JL, Molinaro AM, Phillips JJ, Butowski NA, Chang SM, Perry A, et al. A single-institution phase II trial of radiation, temozolomide, erlotinib, and bevacizumab for initial treatment of glioblastoma. Neuro-Oncology. 2014;16:984–90.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Stupp R, Hegi ME, Mason WP, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. Lancet Oncol. 2009;10:459–66.

    Article  CAS  PubMed  Google Scholar 

  5. Fogh S, Wahl M, Anwar M, Haas-Kogan D, Clarke JL, Sneed PK. Standardization and quality assurance of radiation therapy volumes for adults with high-grade gliomas. Semin Radiat Oncol. 2014;24:259–64.

    Article  PubMed  Google Scholar 

  6. Niyazi M, Brada M, Chalmers AJ, et al. ESTRO-ACROP guideline “target delineation of glioblastomas”. Radiother Oncol. 2016;118:35–42.

    Article  PubMed  Google Scholar 

  7. Minniti G, Amelio D, Amichetti M, Salvati M, Muni R, Bozzao A, et al. Patterns of failure and comparison of different target volume delineations in patients with glioblastoma treated with conformal radiotherapy plus concomitant and adjuvant temozolomide. Radiother Oncol. 2010;97:377–81.

    Article  CAS  PubMed  Google Scholar 

  8. Lawrence YR, Li XA, el Naqa I, Hahn CA, Marks LB, Merchant TE, Dicker AP (2010) Radiation dose-volume effects in the brain. Int J Radiat Oncol Biol Phys 76:S20--S27.

  9. Song SK, Sun SW, Ju WK, Lin SJ, Cross AH, Neufeld AH. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage. 2003;20:1714–22.

    Article  PubMed  Google Scholar 

  10. Kinoshita M, Hashimoto N, Goto T, Kagawa N, Kishima H, Izumoto S, et al. Fractional anisotropy and tumor cell density of the tumor core show positive correlation in diffusion tensor magnetic resonance imaging of malignant brain tumors. Neuroimage. 2008;43:29–35.

    Article  CAS  PubMed  Google Scholar 

  11. Jain R (2013) Measurements of tumor vascular leakiness using DCE in brain tumors: clinical applications. NMR Biomed 26:1042–1049.

  12. Larsson HBW, Courivaud F, Rostrup E, Hansen AE. Measurement of brain perfusion, blood volume, and blood-brain barrier permeability, using dynamic contrast-enhanced T(1)-weighted MRI at 3 tesla. Magn Reson Med. 2009;62:1270–81.

    Article  PubMed  Google Scholar 

  13. Kracht LW, Miletic H, Busch S, Jacobs AH, Voges J, Hoevels M, et al. Delineation of brain tumor extent with [11C]L-methionine positron emission tomography: local comparison with stereotactic histopathology. Clin Cancer Res. 2004;10:7163–70.

    Article  CAS  PubMed  Google Scholar 

  14. Pauleit D, Floeth FW, Hamacher K, Riemenschneider MJ, Reifenberger G, Müller H-W, et al. O-(2-[18F]fluoroethyl)-L-tyrosine PET combined with MRI improves the diagnostic assessment of cerebral gliomas. Brain. 2005;128:678–87.

    Article  PubMed  Google Scholar 

  15. Poulsen SH, Urup T, Grunnet K, Christensen IJ, Larsen VA, Jensen ML, et al. The prognostic value of FET PET at radiotherapy planning in newly diagnosed glioblastoma. Eur J Nucl Med Mol Imaging. 2017;44:373–81.

    Article  CAS  PubMed  Google Scholar 

  16. Lee IH, Piert M, Gomez-Hassan D, Junck L, Rogers L, Hayman J, Ten Haken RK, Lawrence TS, Cao Y, Tsien CI (2009) Association of 11C-methionine PET uptake with site of failure after concurrent temozolomide and radiation for primary glioblastoma multiforme. Int J Radiat Oncol Biol Phys 73:479–485.

  17. Padma MV, Said S, Jacobs M, et al. Prediction of pathology and survival by FDG PET in gliomas. J Neuro-Oncol. 2003;64:227–37.

    Article  CAS  Google Scholar 

  18. Stupp R, Mason WP, van den Bent MJ, et al (2005) Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med 352:987–996.

  19. Wen PY, Macdonald DR, Reardon DA, et al. Updated response assessment criteria for high-grade gliomas: response assessment in neuro-oncology working group. J Clin Oncol. 2010;28:1963–72.

    Article  PubMed  Google Scholar 

  20. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31:1116–28.

    Article  PubMed  Google Scholar 

  21. Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging. 2001;20:45–57.

    Article  CAS  PubMed  Google Scholar 

  22. Avants BB, Tustison NJ, Song G, Cook PA, Klein A, Gee JC. A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54:2033–44.

    Article  PubMed  Google Scholar 

  23. Ou Y, Sotiras A, Paragios N, Davatzikos C. DRAMMS: deformable registration via attribute matching and mutual-saliency weighting. Med Image Anal. 2011;15:622–39.

    Article  PubMed  Google Scholar 

  24. Agresti A. Analysis of ordinal categorical data, 2nd ed. Wiley: Series in Probability and Statistics. 2010. 

  25. Louis DN, Perry A, Reifenberger G, von Deimling A, Figarella-Branger D, Cavenee WK, et al. The 2016 World Health Organization classification of tumors of the central nervous system: a summary. Acta Neuropathol. 2016;131:803–20.

    Article  PubMed  Google Scholar 

  26. Pirotte B, Goldman S, Massager N, David P, Wikler D, Vandesteene A, et al. Comparison of 18F-FDG and 11C-methionine for PET-guided stereotactic brain biopsy of gliomas. J Nucl Med. 2004;45:1293–8.

    CAS  PubMed  Google Scholar 

  27. Labak CM, Wang PY, Arora R, Guda MR, Asuthkar S, Tsung AJ, et al. Glucose transport: meeting the metabolic demands of cancer, and applications in glioblastoma treatment. Am J Cancer Res. 2016;6:1599–608.

    CAS  PubMed  PubMed Central  Google Scholar 

  28. Price SJ, Jena R, Burnet NG, Hutchinson PJ, Dean a F, Peña a, et al. Improved delineation of glioma margins and regions of infiltration with the use of diffusion tensor imaging: an image-guided biopsy study. Am J Neuroradiol. 2006;27:1969–74.

    CAS  PubMed  Google Scholar 

  29. Khalifa J, Tensaouti F, Lotterie JA, et al. Do perfusion and diffusion MRI predict glioblastoma relapse sites following chemoradiation? J Neuro-Oncol. 2016;130:1–12.

    Article  CAS  Google Scholar 

  30. Anwar M, Molinaro AM, Morin O, Chang SM, Haas-Kogan DA, Nelson SJ, et al. Identifying voxels at risk for progression in glioblastoma based on dosimetry, physiologic and metabolic MRI. Radiat Res. 2017;188:303–13.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Piroth MD, Pinkawa M, Holy R, et al. Integrated boost IMRT with FET-PET-adapted local dose escalation in glioblastomas. results of a prospective phase II study. Strahlenther Onkol. 2012;188:334–9.

    Article  CAS  PubMed  Google Scholar 

  32. Lundemann M, Costa JC, Law I, Engelholm SA, Muhic A, Poulsen HS, et al. Patterns of failure for patients with glioblastoma following O-(2-[(18)F]fluoroethyl)-L-tyrosine PET- and MRI-guided radiotherapy. Radiother Oncol. 2017;122:380–6.

    Article  PubMed  Google Scholar 

  33. Henriksen OM, Larsen VA, Muhic A, Hansen AE, Larsson HBW, Poulsen HS, et al. Simultaneous evaluation of brain tumour metabolism, structure and blood volume using [18F]-fluoroethyltyrosine (FET) PET/MRI: feasibility, agreement and initial experience. Eur J Nucl Med Mol Imaging. 2016;43:103–12.

    Article  CAS  PubMed  Google Scholar 

  34. Verger A, Filss CP, Lohmann P, et al. Comparison of18F-FET PET and perfusion-weighted MRI for glioma grading: a hybrid PET/MR study. Eur J Nucl Med Mol Imaging. 2017;44:2257–65.

    Article  CAS  PubMed  Google Scholar 

  35. Andersen FL, Ladefoged CN, Beyer T, Keller SH, Hansen AE, Højgaard L, et al. Combined PET/MR imaging in neurology: MR-based attenuation correction implies a strong spatial bias when ignoring bone. Neuroimage. 2014;84:206–16.

    Article  PubMed  Google Scholar 

  36. Andersson JLR, Sotiropoulos SN. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. Neuroimage. 2016;125:1063–78.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Saremi F. Perfusion imaging in clinical practice, a multimodality approach to tissue perfusion analysis. Hagerstown: Wolters Kluwer Health; 2015.

    Google Scholar 

  38. Larsson HBW, Hansen AE, Berg HK, Rostrup E, Haraldseth O. Dynamic contrast-enhanced quantitative perfusion measurement of the brain using T1-weighted MRI at 3T. J Magn Reson Imaging. 2008;27:754–62.

    Article  PubMed  Google Scholar 

  39. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010. https://doi.org/10.18637/jss.v033.i01.

Download references

Acknowledgements

The authors would like to thank the John and Birthe Meyer Foundation for the donation of the Siemens mMR hybrid PET/MR system to Rigshospitalet. The authors would also like to thank Karin Stahr, Marianne Federspiel and Jakup Poulsen for help with data acquisition, Betina Rotbøll and Lotte S. Andersen for coordinating logistics and Kirsten Grunnet for clinical data management.

Funding

This study was funded by the Lundbeck Foundation, Department of Oncology (Rigshospitalet), Department of Clinical Physiology, Nuclearmedicine & PET (Rigshospitalet) and the Niels Bohr Institute, Copenhagen University, Copenhagen, Denmark.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Michael Lundemann.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethics statement

All procedures performed were in accordance with the 1964 Helsinki declaration and approved by the ethical committee for the Capital Region of Denmark (H-3-2013-162).

Informed consent

Informed consent was obtained from all individual participants included in the study.

Electronic supplementary material

ESM 1

(PDF 50 kb)

ESM 2

(PDF 1403 kb)

ESM 3

(PDF 150 kb)

Appendix

Appendix

Imaging parameters

The 18F-FET PET/CT was performed on a 64 slice Siemens Biograph mCT (Siemens, Erlangen, Germany) approximately 20 min after intravenous administration of 200 MBq FET. A spiral CT acquisition was immediately followed by 20 min static PET acquisition. CT images were reconstructed to a 512 × 512 matrix with a voxel size of 0.6 × 0.6 × 1 mm. PET images were reconstructed to a 400 × 400 matrix using 3D-OSEM with four iterations and 12 subsets and filtered with a 5 mm Gaussian filter to a nominal voxel size of 0.8 × 0.8 × 3 mm.

The 18F-FDG PET/MRI was done on a Siemens Biograph mMR, with a 3 T magnet and 8-channel headcoil. A 60-min dynamic list-mode PET acquisition was performed simultaneously with the MRI acquisition after injection of 200 MBq FDG. An average of the PET frames from 40 to 60 min were reconstructed to a 344 × 344 matrix using OP-OSEM with four iterations and 21 subsets and filtered with a 3 mm Gaussian filter to a voxel size of 0.8 × 0.8 × 2 mm. The CT from the FET-PET/CT was used for attenuation correction [35]. Pre- and post-contrast 3D isotropic T1-weighted magnetization prepared rapid gradient echo (MPRAGE) were acquired with identical parameters; flip angle (FA) 9 degrees, echo time (TE) 2.52 ms, inversion time (TI) 900 ms, repetition time (TR) 1900 ms. The field-of-view was 256x256x208 in the APxISxRL direction with a resolution of 1x1x1 mm. Axial T2-weighted turbo spin echo with radial sampling (BLADE) was acquired with FA 90 degrees, TE/TR 117/9480 ms, 41 slices with a thickness of 3 mm and slice-gap of 0.9 mm, matrix size 320 × 320 with a 230 × 230 mm FoV resulting in 0.72 × 0.72 mm in-plane resolution. Fluid attenuated inversion recovery (FLAIR) images were acquired using axial T2-weighted turbo inversion recovery magnitude (TIRM) with FA 130 degrees, TE/TI/TR 58/2500/9000 ms, 50 slices with a thickness of 3 mm and no gap. The acquisition matrix was 256 × 256 and in-plane resolution of 0.45 × 0.45 mm. Diffusion weighted images were acquired using an echo planar imaging (EPI) sequence with FA 90 degrees, TR/TE 4000/95 ms, 28 slices with a thickness of 3 mm and slice-gap of 0.9 mm, 128 × 128 matrix with 1.7×.1.7 mm in-plane resolution. Two diffusion weightings with b = 0 and b = 1000 s/mm2 in 30 directions and four averages were used. In the last two averages the phase-encoding direction was reversed to enable restoration of image-distortions due to susceptibility differences. The DCE perfusion was acquired using a fast 3D spoiled gradient echo sequence with FA 14 degrees, TR/TE 3.63/1.02 ms, 30 axial slices, 180 time-frames and a temporal resolution of 2.6 s. Field-of-view was 187x230x150 in the APxRLxIS direction with a resolution of 2.4 × 2.4 × 5 mm. Images for T1-mapping were acquired before contrast injection using variable flip angles (8, 14 and 20, or 4, 8, 14 and 20 degrees) and otherwise identical parameters. Two half-dose (0.05 mL/kg) boluses of contrast agent (Gadovist 0.1 mmol/mL) were injected using a power injector approximately 18 and 85 s after the dynamic DCE acquisition was started.

Data processing

Diffusion tensor MRI was processed using FSL [36]. Data were first corrected for susceptibility induced distortions and eddy currents. Secondly, a diffusion tensor was fitted to each voxel resulting in maps of mean diffusivity (MD) and fractional anisotropy (FA). The DCE perfusion data was processed using in-house software written in MATLAB (The MathWorks, Inc., Natick, MA, USA) as described previously [12]. Shortly, the MRI signal was converted to relative signal time curves [37]. Blood flow (F) and mean transit time (MTT) were estimated using a model-free deconvolution regularized by Tikhonov’s method [38]. Maps of vascular permeability (Ki), intravascular blood volume (Vb) and volume of the extra-vascular, extra-cellular space (Ve) were computed by fitting a two-compartment model to the concentration-time curves, with F fixed to the estimate from model-free deconvolution. The CT was used for attenuation correction of both FET and FDG [35]. Within each patient, 18F-FDG-uptake was normalized to a manually defined healthy appearing region in centrum semi-ovale and 18F-FET-uptake was normalized to healthy appearing contralateral cortex including grey and white matter using a standard method [14].

Binomial logistic regression model

Binomial logistic regression models were fitted using the lassoglm function in MATLAB (Statistics and Machine Learning toolbox, Matlab R2017a, The MathWorks, Inc., Natick, MA, USA) that implements a general linear model (GLM) with elastic net regularization. The regularization procedure has two hyperparameters, α and λ. The former reduces the effect of correlated variables, whereas the latter is adjusted to prevent overfitting. Five values for α and 100 different values for λ were investigated for each model and the optimal parameters were determined by 10-fold cross-validation with the objective to minimize binomial deviance. The α levels were selected from the set [0.05, 0.2, 0.5, 0.95, 1] and the experimental values for λ were automatically determined by lassoglm – see Friedman et al. [39] for details.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lundemann, M., Munck af Rosenschöld, P., Muhic, A. et al. Feasibility of multi-parametric PET and MRI for prediction of tumour recurrence in patients with glioblastoma. Eur J Nucl Med Mol Imaging 46, 603–613 (2019). https://doi.org/10.1007/s00259-018-4180-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00259-018-4180-3

Keywords

Navigation