INTRODUCTION

Derived from glial cells, glioblastoma multiforme (GBM) is an extremely aggressive malignant tumor in adults. Generating MR-derived growth pattern models for GBM have been of an attractive approach in neuro-oncology, suggesting a distinct pattern of lesion spread with tendency in growing along the white fiber direction for the invasive component (1,2). Here we report preliminary results from a retrospective analysis, designed to provide a brain atlas predicting the dominant directions of tumor lesion expansion/shrinkage prior to surgery.

METHODS

Forty three adult glioblastoma patients were examined at two time-points; diagnostic and pre-operation, with minimum 14 days between acquisitions with contrast-enhanced anatomical MRI (ce-T1). The study was approved by the regional ethics committee and adhered to the Helsinki Declaration. The tumor volumes were delineated by two radiology experts. The image analysis software BrainVoyager QX (Version 2.3, Brain Innovation B.V., Maastricht, The Netherlands) was used for semi-automatic tumor segmentation. The widely accepted Advanced Normalization Tools (ANTs) script packages (3) and customized MATLAB scripts were used for registration of MR images. The 3D data analysis pipeline (Fig. 1) was started with a few pre-processing steps including up-sampling for identical voxel number/size, N4 bias field correction (from ANTs package) and skull stripping using Brain Extraction Tool (BET) (4). Using segmented tumor masks, fine non-linear registration was performed to generate deformation fields of tumor lesions delineated from T1-diagnostic and T1-pre-operation images. Using non-linear transformation, all solutions were transformed into the MNI template space and compared with a DTI WM atlas (5). Angles between the vector fields of DTI WM and tumor growth directions were calculated (within sub-regions with prominent lesion overlaps) to estimate an angle map of lesion growth to that of white matter tracts.

RESULTS

To compute the degree of alignment between tumor growth vector fields and white matter tracts, an angle map was calculated. The results of the generated maps were color-coded to visualize the alignment agreements between two dataset, as yellow color indicates the maximum alignment (θ<20º or θ>160º, Fig. 1). In all sub-regions, tumor displacement showed a reproducible tendency in moving along the white matter tracts (Fig. 1), as evidenced by voxel-wise parallel alignments of the mean vector fields towards the tensor field of the WM atlas.

CONCLUSION

The resulting GBM atlas of deformation vector field provides information on tumor growth pattern prior to surgery. We mapped the delineated growth pattern on DTI WM atlas, investigating the hypothesis that tumor cells tend to track along the white matter. Our findings represent a first step in investigating the hypothesis that tumor cells tend to track along the white matter.

References

1. Stensjoen AL, Solheim O, Kvistad KA, Haberg AK, Salvesen O, Berntsen EM. Growth dynamics of untreated glioblastomas in vivo. Neuro Oncol 2015;17(10):1402-1411. 2. Bondiau PY, Clatz O, Sermesant M, Marcy PY, Delingette H, Frenay M, Ayache N. Biocomputing: numerical simulation of glioblastoma growth using diffusion tensor imaging. Phys Med Biol 2008;53(4):879-893. 3. Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, Gee JC. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging 2010;29(6):1310-1320. 4. Muschelli J, Ullman NL, Mould WA, Vespa P, Hanley DF, Crainiceanu CM. Validated automatic brain extraction of head CT images. Neuroimage 2015;114:379-385. 5. Zhang H, Awate SP, Das SR, Woo JH, Melhem ER, Gee JC, Yushkevich PA. A tract-specific framework for white matter morphometry combining macroscopic and microscopic tract features. Med Image Anal 2010;14(5):666-673.