Combined 18F-FET PET and diffusion kurtosis MRI in posttreatment glioblastoma: differentiation of true progression from treatment-related changes

Francesco Annibale D'Amore, Farida Grinberg, Jörg Mauler, Norbert Galldiks, Ganna Blazhenets, Ezequiel Farrher, Christian Filß, Gabriele Stoffels, Felix M. Mottaghy, Philipp Lohmann, N. Jon Shah, Karl‐Josef LangenNeuro-Oncology 2021 · DOI

TL;DR

Distinguishing between true tumor progression and treatment-related changes in post-treatment glioblastoma is a critical clinical challenge. This study demonstrates that combining 18F-FET PET and diffusion kurtosis MRI (DKI) significantly improves diagnostic accuracy, achieving 94% accuracy in differentiating these two conditions. This hybrid approach offers a promising new tool for more precise patient management.

What Problem Does This Paper Address?

Glioblastoma, a highly aggressive brain tumor, often recurs after initial treatment. A major hurdle in managing these patients is accurately differentiating between true tumor progression (TPR) and treatment-related changes (TRC), such as pseudoprogression or radiation necrosis. TRCs can mimic the appearance of tumor growth on standard MRI, leading to potentially incorrect treatment decisions. While 18F-FET PET has shown promise in identifying metabolically active tumor tissue, and advanced MRI techniques like DKI offer insights into tissue microstructure, their combined diagnostic power for this specific challenge in the post-treatment setting has not been fully explored. This paper aims to bridge this gap by investigating the synergistic value of these modalities.

Methods

The researchers employed a retrospective study design involving 32 patients with pretreated glioblastoma who exhibited signs suggestive of TPR. They utilized a hybrid PET/MRI scanner to acquire both 18F-FET PET and diffusion-weighted MRI data, including DKI sequences.

Key methodological aspects include:

  • 18F-FET PET Imaging: Dynamic PET scans were acquired after injecting 18F-FET. Regions of interest (ROIs) were defined based on a tumor-to-brain ratio (TBR) of at least 1.6, focusing on metabolically active areas. Parameters like mean and maximum TBR, time-to-peak (TTP), and slope of the time-activity curve were extracted.
  • Diffusion Kurtosis MRI (DKI) Acquisition: DKI sequences were acquired with multiple b-values (0, 1000, 2500 s/mm²) and gradient directions.
  • DKI Image Processing: Raw diffusion-weighted images were corrected for motion and eddy currents. DKI analysis was performed to generate diffusion tensor (DT) and kurtosis tensor (KT) parameter maps, including mean diffusivity (MD), radial diffusivity (RD), axial diffusivity (AD), mean kurtosis (MK), radial kurtosis (RK), and axial kurtosis (AK).
  • Histogram Analysis: 3D ROIs from the PET-guided segmentation were used to generate voxel-wise histograms for DT and KT parameters. Key histogram metrics (means, centiles) were extracted.
  • Statistical Analysis: Mann-Whitney U tests were used for intergroup comparisons. Receiver operating characteristic (ROC) curve analysis was employed to determine diagnostic accuracy (AUC, sensitivity, specificity, cutoff values). Multivariate logistic regression was used to assess the combined predictive value of parameters.
  • FET-DKI Index: A novel index was developed by combining weighted values of TBRmax (from PET) and MK C90 (from DKI) to create a single, clinically applicable score.

Datasets

The study utilized data from 32 patients with histopathologically defined and pretreated glioblastoma. Twenty-one patients were assigned to the true progression (TPR) group, and 11 to the treatment-related changes (TRC) group. The diagnosis was confirmed by histopathology or clinicoradiological follow-up. No new benchmark datasets were introduced.

Key Insights and Ideas

  • Synergistic Power of Hybrid Imaging: The core insight is that combining functional information from 18F-FET PET (amino acid transport) with microstructural information from DKI provides a more comprehensive picture of tissue changes than either modality alone.
  • Kurtosis Metrics Outperform Diffusivity: Diffusion kurtosis (KT) parameters, particularly MK, showed significantly higher diagnostic accuracy in differentiating TPR from TRC compared to diffusion (DT) parameters. This suggests that kurtosis is more sensitive to the subtle microstructural alterations occurring in progressing tumors versus treatment effects.
  • PET-Guided ROIs are Crucial: Segmenting ROIs based on 18F-FET uptake, rather than solely on contrast enhancement on MRI, was key. This approach focuses the analysis on metabolically active areas, potentially excluding non-specific edema or reactive changes that can confound standard MRI interpretation.
  • A Simple, Robust Index: The development of the FET-DKI index, a single score derived from a combination of PET and DKI metrics, offers a practical tool for clinical application, simplifying the interpretation of complex imaging data.

Key Results

  • Individual Modality Performance: Diffusion kurtosis MRI parameters alone achieved diagnostic accuracy up to 88% for differentiating TPR from TRC. 18F-FET PET parameters showed moderate accuracy, with TBRmax being the most significant individual parameter.
  • Combined Approach Superiority: The combination of 18F-FET PET (TBRmax) and DKI (MK C90) achieved the highest diagnostic accuracy of 94% (AUC = 0.97, sensitivity = 94%, specificity = 91%).
  • FET-DKI Index Performance: The developed FET-DKI index demonstrated excellent diagnostic performance, with an AUC of 0.97, a cutoff of 41, a sensitivity of 95%, and a specificity of 91%.
  • Statistical Significance: Significant differences were observed in KT parameters between the TPR and TRC groups, with TPR generally showing higher kurtosis. DT parameters showed a tendency but did not reach statistical significance after Bonferroni correction.

Limitations and Future Work

The authors acknowledge several limitations:

  • Small Sample Size: The study included a relatively small number of patients (n=32), and the TRC group was particularly small (n=11).
  • Absence of a Validation Set: The findings were not validated on an independent dataset.
  • Monocentric Approach: Data was collected from a single center.
  • Limited Biopsies in TRC Group: The diagnosis in the TRC group relied on clinical follow-up in many cases, with fewer histopathological confirmations.
  • Variability in Treatment Timing: The timing of various treatments prior to imaging varied, which could potentially influence results.
  • Potential Influence of Anti-angiogenic Drugs/Steroids: While 18F-FET uptake is less affected by BBB disruption, these treatments could influence contrast enhancement on MRI.

Future work suggested includes:

  • Larger, Multicenter Studies: Validation in larger, multi-institutional cohorts is crucial.
  • Neuropathological Correlation: Further studies with neuropathological validation of neuroimaging findings are warranted.
  • Tumor Subregion Analysis: Investigating diffusion abnormalities within specific subregions of highest 18F-FET uptake and correlating them with histomolecular findings could provide deeper insights.
  • Longitudinal Studies: Exploring the dynamic changes in diffusion signals over time and their relationship with treatment response.

Summary updated

Abstract

The combined use of <sup>18</sup>F-FET PET and MRI diffusion kurtosis maps appears to be a promising approach to improve the differentiation of TPR from TRC in pretreated glioblastoma and warrants further investigation.
BibTeX
@inproceedings{damore2021combined,
  author = {Francesco Annibale D'Amore and Farida Grinberg and Jörg Mauler and Norbert Galldiks and Ganna Blazhenets and Ezequiel Farrher and Christian Filß and Gabriele Stoffels and Felix M. Mottaghy and Philipp Lohmann and N. Jon Shah and Karl‐Josef Langen},
  title = {{Combined 18F-FET PET and diffusion kurtosis MRI in posttreatment glioblastoma: differentiation of true progression from treatment-related changes}},
  booktitle = {Neuro-Oncology},
  year = {2021},
  doi = {10.1093/noajnl/vdab044},
}
Figure