Scott Nyshin Hwang PhD,MD, Presenter: Nothing to Disclose

Chad Ashley Holder MD, Abstract Co-Author: Nothing to Disclose

Erich Huang PhD, Abstract Co-Author: Stockholder, Johnson & Johnson Stockholder, General Electric Company

Robert A. Clifford, Abstract Co-Author: Nothing to Disclose

Dima Hammoud MD, Abstract Co-Author: Nothing to Disclose

Prashant Raghavan MD, Abstract Co-Author: Nothing to Disclose

Manal Nicolasjilwan MD, Abstract Co-Author: Nothing to Disclose

Max Wintermark MD, Abstract Co-Author: Nothing to Disclose

David Gutman, Abstract Co-Author: Nothing to Disclose

Carlos Moreno, Abstract Co-Author: Nothing to Disclose

Lee Cooper, Abstract Co-Author: Nothing to Disclose

John B. Freymann BS, Abstract Co-Author: Nothing to Disclose

Justin Kirby, Abstract Co-Author: Nothing to Disclose

Arun Radha Krishnan MD, Abstract Co-Author: Nothing to Disclose

Seena Dehkharghani MD, Abstract Co-Author: Nothing to Disclose

Carl Jaffe, Abstract Co-Author: Nothing to Disclose

Joel Haskin Saltz MD, PhD, Abstract Co-Author: Nothing to Disclose

Adam E. Flanders MD, Abstract Co-Author: Research Consultant, Geron Corporation

Daniel J. Brat MD, PhD, Abstract Co-Author: Nothing to Disclose

Kenneth Buetow, Abstract Co-Author: Nothing to Disclose

Investigate quantitative computer-aided and qualitative visual imaging characteristics of primary glioblastomas that may predict genomic features.

As part of The Cancer Genome Atlas (TCGA) MRI characterization project of the National Cancer Institute (NCI), methods for image analysis were developed as macros for NIH ImageJ. Preprocessing with 3D Slicer (software partially sponsored by NIH/NCI) included anisotropic diffusion noise reduction, N4 intensity inhomogeneity correction, and image co-registration. Measurements included volumes and surface areas of enhancing, nonenhancing, and necrotic tumor. At least 3 neuroradiologists independently reviewed MRI scans of 75 glioblastoma patients in the TCGA database on the basis of 26 imaging features (VASARI, https://wiki.nci.nih.gov/display/CIP/VASARI). Tumors were classified as Proneural, Neural, Classic, or Mesenchymal subtypes based on gene expression. Proneural subtype is associated with improved outcome. Genomic data were obtained from TCGA’s publicly available information, including mutation status (e.g. TP53 gene). Subgroups were identified based on copy number data (e.g. CDKN2A homozygous deletions).

Proneural tumors had smaller volume of necrosis by computer-aided measurements (p=0.04). A suggestion that Proneural tumors have smaller enhancing tumor volume (p=0.06) supports an association between minimal enhancing tumor (≤5% proportion of the overall tumor) and Proneural classification (p=0.0006), as qualitatively assessed by neuroradiologists. Mesenchymal tumors were also associated with higher necrosis volume (p=0.05). A suggested association between larger necrosis volume and CDKN2A homozygous deletion (p=0.07) was noted. Morphology may also be an important feature, since Classic glioblastomas had greater surface area to volume ratio of enhancing tumor (p=0.002).

Quantitative computer-aided measurements and qualitative visual assessment of glioblastomas are associated with certain genomic features. In particular, Proneural classification is associated with decreased necrosis, and possibly with smaller enhancing tumor volume.

A combination of quantitative computer-aided measurements and radiologist qualitative assessment of glioblastomas may be potentially helpful in predicting genomic features and clinical outcome.A combi

Hwang, S, Holder, C, Huang, E, Clifford, R, Hammoud, D, Raghavan, P, Nicolasjilwan, M, Wintermark, M, Gutman, D, Moreno, C, Cooper, L, Freymann, J, Kirby, J, Krishnan, A, Dehkharghani, S, Jaffe, C, Saltz, J, Flanders, A, Brat, D, Buetow, K, Computer-aided Visual Image Analysis of Glioblastomas and Genomic Features.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11004309.html