TY - JOUR
T1 - Differentiation of edema and glioma infiltration
T2 - Proposal of a DTI-based probability map
AU - Hoefnagels, Friso W.A.
AU - De Witt Hamer, Philip
AU - Sanz-Arigita, Ernesto
AU - Idema, Sander
AU - Kuijer, Joost P.A.
AU - Pouwels, Petra J.W.
AU - Barkhof, Frederik
AU - Vandertop, W. Peter
PY - 2014/9/27
Y1 - 2014/9/27
N2 - Conflicting results on differentiating edema and glioma by diffusion tensor imaging (DTI) are possibly attributable to dissimilar spatial distribution of the lesions. Combining DTI-parameters and enhanced registration might improve prediction. Regions of edema surrounding 22 metastases were compared to tumor-infiltrated regions from WHO grade 2 (12), 3 (10) and 4 (18) gliomas. DTI data was co-registered using Tract Based Spatial Statistics (TBSS), to measure Fractional Anisotropy (FA) and Mean Diffusivity (MD) for white matter only, and relative changes compared to matching reference regions (dFA and dMD). A two-factor principal component analysis (PCA) on metastasis and grade 2 glioma was performed to explore a possible differentiating combined factor. Edema demonstrated equal MD and higher FA compared to grade 2 and 3 glioma (P < 0.001), but did not differ from glioblastoma. Differences were non-significant when corrected for spatial distribution, since reference regions differed strongly (P < 0.001). The second component of the PCA (PCA-C2) did differentiate edema and low-grade tumor (sensitivity 91.7 %, specificity 86.4 %). PCA-C2 scores were plotted voxel-wise as a probability-map, discerning distinct areas of presumed edema or tumor infiltration. Correction of spatial dependency appears essential when differentiating glioma from edema. A tumor-infiltration probability-map is presented, based on supplementary information of multiple DTI parameters and spatial normalization.
AB - Conflicting results on differentiating edema and glioma by diffusion tensor imaging (DTI) are possibly attributable to dissimilar spatial distribution of the lesions. Combining DTI-parameters and enhanced registration might improve prediction. Regions of edema surrounding 22 metastases were compared to tumor-infiltrated regions from WHO grade 2 (12), 3 (10) and 4 (18) gliomas. DTI data was co-registered using Tract Based Spatial Statistics (TBSS), to measure Fractional Anisotropy (FA) and Mean Diffusivity (MD) for white matter only, and relative changes compared to matching reference regions (dFA and dMD). A two-factor principal component analysis (PCA) on metastasis and grade 2 glioma was performed to explore a possible differentiating combined factor. Edema demonstrated equal MD and higher FA compared to grade 2 and 3 glioma (P < 0.001), but did not differ from glioblastoma. Differences were non-significant when corrected for spatial distribution, since reference regions differed strongly (P < 0.001). The second component of the PCA (PCA-C2) did differentiate edema and low-grade tumor (sensitivity 91.7 %, specificity 86.4 %). PCA-C2 scores were plotted voxel-wise as a probability-map, discerning distinct areas of presumed edema or tumor infiltration. Correction of spatial dependency appears essential when differentiating glioma from edema. A tumor-infiltration probability-map is presented, based on supplementary information of multiple DTI parameters and spatial normalization.
KW - Edema
KW - Glioma infiltration
KW - MR diffusion tensor imaging
KW - Probability map
KW - Spatial normalization
UR - http://www.scopus.com/inward/record.url?scp=84910109757&partnerID=8YFLogxK
U2 - https://doi.org/10.1007/s11060-014-1544-9
DO - https://doi.org/10.1007/s11060-014-1544-9
M3 - Article
C2 - 25079117
SN - 0167-594X
VL - 120
SP - 187
EP - 198
JO - Journal of Neuro-Oncology
JF - Journal of Neuro-Oncology
IS - 1
ER -