Program Information
Evaluation of Fully Automatic Volumetric GBM Segmentation in the TCGA-GBM Dataset: Prognosis and Correlation with VASARI Features
E Rios Velazquez1*, R Meier2 , W Dunn3 , B Alexander4 , R Wiest5 , S Bauer6 , D Gutman7 , M Reyes8 , H Aerts9 , (1) Dana-Farber Cancer Institute | Harvard Medical School, Boston, MA, (2) Institute for Surgical Technology and Biomechanics, Bern, NA, (3) Emory University School of Medicine, Atlanta, GA, (4) Dana-Farber Cancer Institute, Brigham and Womens Hospital, Harvard Medic, Boston, MA, (5) Institute for Surgical Technology and Biomechanics, University of Bern, Bern, NA, (6) Institute for Surgical Technology and Biomechanics, Support Center for Adva, Bern, NA, (7) Emory University School of Medicine, Atlanta, GA, (8) Institute for Surgical Technology and Biomechanics, University of Bern, Bern, NA, (9) Dana-Farber/Brigham Womens Cancer Center, Boston, MA
Presentations
TU-AB-BRA-11 (Tuesday, July 14, 2015) 7:30 AM - 9:30 AM Room: Ballroom A
Purpose: Reproducible definition and quantification of imaging biomarkers is essential. We evaluated a fully automatic MR-based segmentation method by comparing it to manually defined sub-volumes by experienced radiologists in the TCGA-GBM dataset, in terms of sub-volume prognosis and association with VASARI features.
Methods: MRI sets of 67 GBM patients were downloaded from the Cancer Imaging archive. GBM sub-compartments were defined manually and automatically using the Brain Tumor Image Analysis (BraTumIA), including necrosis, edema, contrast enhancing and non-enhancing tumor. Spearman’s correlation was used to evaluate the agreement with VASARI features. Prognostic significance was assessed using the C-index.
Results: Auto-segmented sub-volumes showed high agreement with manually delineated volumes (range (r): 0.65 – 0.91). Also showed higher correlation with VASARI features (auto r = 0.35, 0.60 and 0.59; manual r = 0.29, 0.50, 0.43, for contrast-enhancing, necrosis and edema, respectively). The contrast-enhancing volume and post-contrast abnormal volume showed the highest C-index (0.73 and 0.72), comparable to manually defined volumes (p = 0.22 and p = 0.07, respectively). The non-enhancing region defined by BraTumIA showed a significantly higher prognostic value (CI = 0.71) than the edema (CI = 0.60), both of which could not be distinguished by manual delineation.
Conclusion: BraTumIA tumor sub-compartments showed higher correlation with VASARI data, and equivalent performance in terms of prognosis compared to manual sub-volumes. This method can enable more reproducible definition and quantification of imaging based biomarkers and has a large potential in high-throughput medical imaging research.
Contact Email: