Brain connectivity of glioblastoma patients using MR-PET and DTI data
Ana Carina Mendes1, Ana-Maria Oros-Peusquens2, André Santos Ribeiro1,3, Karl-Josef Langen2, Carolin Weiß Lucas4, Nadim Jon Shah2, and Hugo Alexandre Ferreira1

1Institute of Biophysics and Biomedical Engineering, Faculty of Sciences of the University of Lisbon, Lisbon, Portugal, 2Forschungszentrum Juelich GmbH, Institute of Neurosciences and Medicine-INM4, Juelich, Germany, 3Centre for Neuropsychopharmacology, Division of Brain Sciences, Department of Medicine, Imperial College London, London, United Kingdom, 4Center of Neurosurgery, University of Cologne, Cologne, Germany

Synopsis

Methods capable of mapping brain connectivity pathways may prove useful by providing valuable information in order to prevent sequelae following a surgical intervention. This study presents an approach for the whole-brain connectivity evaluation of nine patients with lateralized gliobastoma, using the Multimodal Imaging Brain Connectivity Analysis (MIBCA) toolbox to process MR and PET data. Results show changes in connectivity metrics across both hemispheres for all patients accompanied by an increased number of fibres which may result from reorganization of connectivity pathways caused by the disruption of the original ones by the tumour.

Purpose

Planning of surgery involves the difficult balancing act between removing all areas likely to be affected by tumour infiltration and preserving function by removing as little functional tissue as possible.1 The purpose of this work was to study the whole-brain connectivity of glioblastoma patients, paving the way for the development of an improved connectivity-based pre-surgical planning protocol. Therefore, the Multimodal Imaging Brain Connectivity Analysis (MIBCA) toolbox was used.2

Material and methods

Nine patients with glioblastoma (5 left / 4 right hemisphere, all adjacent to the primary motor area, Figures 1 and 2) underwent simultaneous magnetic resonance imaging (MRI) and dynamic 18F-fluoro-ethyl-tyrosine (18F-FET) positron emission tomography (PET) scans. The control group comprised twenty-two healthy volunteers (only MRI was performed). Imaging data were acquired on a hybrid MR-PET scanner, consisting of a 3T Siemens scanner with a BrainPET insert.3 A birdcage transmit coil and an 8-element receive array were used for radiofrequency transmit and signal receive, respectively. The MRI protocol included volumetric T1-weighted (T1-w) MPRAGE (1x1x1 mm3), diffusion tensor imaging (DTI) (dir=30, b-value=800 s/mm2, 2 averages, 2x2x2 mm3) and contrast enhanced volumetric T1-w MPRAGE (1x1x1 mm³) sequences. The amino acid 18F-FET was produced via nucleophilic 18F-fluorination with a specific radioactivity of more than 200 GBq/μmol.4 Time activity curves of FET uptake were generated and mean and maximum tumour-to-brain ratios were determined by region-of-interest (ROI) analysis. Tumour volumes in FET-PET images were calculated from images containing the integrated intensity (20–60 min) using threshold-based volume-of-interest analyses that included voxels with a tumour-to-brain ratio of at least 1.6. This cutoff was based on a biopsy-controlled study in cerebral gliomas in that a lesion-to-brain ratio of 1.6 best separated tumour from peritumour tissue.5 The MIBCA toolbox was used to automatically pre-process MR-PET data (including brain parcellation) and to derive imaging and connectivity metrics from the multimodal data, that is, cortical thickness from contrast enhanced T1-w data; mean diffusivity (MD), fractional anisotropy (FA), node degree, clustering coefficient and pairwise ROI fibre tracking from DTI data; and standardized uptake value (SUV) from PET data. Differences in whole-brain connectivity between patients and controls were analyzed. Mean and standard deviation (SD) values were obtained for the control data, and latter used to threshold patient data in order to constrain the results to significant differences. To do so, intervals with the mean and triple SD were obtained in order to threshold patient data (by verifying which ROIs of each metric fall off the range). Results were visualized in a connectogram and both structural connectivity and metrics were studied in regions surrounding lesions (to the extent of the present oedema), identified by increased uptake in FET-PET.

Results and Discussion

In Table 1 an overview of the information obtained from the connectograms concerning the affected metrics is presented, divided by peritumour and more distant regions. All patients had regions in which one or more parameters included in the connectogram deviated by more than 3 SD from the mean value based on data from healthy volunteers. For most patients, the number of affected regions is higher in the contralateral, as compared to the ipsilateral hemisphere. In particular, patient E showed regions with more than two affected metrics only for non-peritumour regions. Connectograms with thresholded data can be found in Figures 3 and 4. Data presented in the connectograms concerning the distribution of fibres suggest an increase of connections, particularly in the lesion side, more evident in patients with the tumour in the right hemisphere (Figure 4, patients G and H). Findings concerning changes in metric values, particularly increased FA values, may be related to fibre packing caused by the tumour mass effect. Changes found in more distant areas may result from structural re-organization in response to the presence of the tumour.

Conclusions

The use of multimodal imaging in this study allows the integration of different type of data that result in a richer characterization of brain connectivity. Reported results suggest that tumour infiltration may alter both local and more distant structural connections. Future work will focus on the inclusion of a larger patient group so as to increase robustness of these findings as well as diffusion kurtosis information to study its influence on the identification of connectivity changes produced by tumours. Ultimately, the goal of the approach taken in this study is to investigate if and how connectivity studies can help pre-surgical planning. For instance, by studying specific brain networks in order to determine the best anatomical approach to tumour resection and influence the process of decision-making during surgery, regarding the prevention of sequelae.

Acknowledgements

Research supported by the Edmond J. Safra Philanthropic Foundation, and by the Fundação para a Ciência e Tecnologia (FCT) and Ministério da Ciência e Educação (MCE) Portugal (PIDDAC) under grants UID/BIO/00645/2013 and PTDC/SAU-ENB/120718/2010.

References

1. S. M. Krieg, J. Sabih, L. Bulubasova, T. Obermueller, C. Negwer, I. Janssen, E. Shiban, B. Meyer, and F. Ringel, “Preoperative motor mapping by navigated transcranial magnetic brain stimulation improves outcome for motor eloquent lesions,” Neuro-Oncology, vol. 16, pp. 1274–1282, Sept. 2014

2. A. Santos Ribeiro, L. M. Lacerda, and H. A. Ferreira, “Multimodal Imaging Brain Connectivity Analysis toolbox (MIBCA),” PeerJ 3:e1078, Jul. 2015

3. H. Herzog, K.-J. Langen, C. Weirich, E. Rota Kops, J. Kaffanke, L. Tellmann, J. Scheins, I. Neuner, G. Stoffels, K. Fischer, L. Caldeira, H. H. Coenen, and N. J. Shah, “High resolution BrainPET combined with simultaneous MRI:,” Nuklearmedizin, vol. 50, no. 2, pp. 74–82, Feb. 2011

4. K. Hamacher and H. H. Coenen, “Efficient routine production of the 18f-labelled amino acid O-(2-[18f]fluoroethyl)-l-tyrosine,” Applied Radiation and Isotopes, vol. 57, pp. 853–856, Dec. 2002

5. D. Pauleit, F. Floeth, K. Hamacher, M. J. Riemen-schneider, G. Reifenberger, H.-W. Müller, K. Zilles, H. H. Coenen, and K.-J. Langen, “O-(2-[18f]fluoroethyl)-l-tyrosine PET combined with MRI improves the diagnostic assessment of cerebral gliomas,” Brain, vol. 128, pp. 678–687, Mar. 2005

Figures

Figure 1. T1-weighted axial, coronal and sagittal cuts for patients with tumours in the left hemisphere, showing the surrounding regions affected by the tumours. In the first row plain T1-weighted images are shown and in the second row the correspondent parcellations obtained with Freesurfer are displayed.

Figure 2. T1-weighted axial, coronal and sagittal cuts for patients with tumours in the right hemisphere, showing the surrounding regions affected by the tumours. In the first row plain T1-weighted images are shown and in the second row the correspondent parcellations obtained with Freesurfer are displayed.

Figure 3. Left-sided tumour patients thresholded connectograms (uncorrected p-val=0.0015). From outer to inner rings: left (dark grey) and right (light grey) cortical and subcortical (black) regions/structures; standardized uptake value (SUV); cortical thickness (CtxT); fractional anisotropy (FA); mean diffusivity (MD); clustering coefficient (ClusC) and node degree (Deg). Connections correspond to DTI tractography (middle).

Figure 4. Right-sided tumour patients thresholded connectograms (uncorrected p-val=0.0015). From outer to inner rings: left (dark grey) and right (light grey) cortical and subcortical (black) regions/structures; standardized uptake value (SUV); cortical thickness (CtxT); fractional anisotropy (FA); mean diffusivity (MD); clustering coefficient (ClusC) and node degree (Deg). Connections correspond to DTI tractography (middle).

Table 1. Peritumour regions for each patient and also more distant affected regions. Represented changes correspond to the application of the triple value of the threshold.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0956