Daniel Paul Barboriak MD, Presenter: Advisory Board, General Electric Company

Zheng Zhang PhD, Abstract Co-Author: Nothing to Disclose

Pratik Desai, Abstract Co-Author: Nothing to Disclose

Yair Safriel MBBCh, Abstract Co-Author: Principal, PharmaScan Clinical Trials

Robert C. McKinstry MD, PhD, Abstract Co-Author: Travel support, Siemens AG Speaker, Siemens AG

Felix Bokstein, Abstract Co-Author: Nothing to Disclose

A. Gregory Sorensen MD, Abstract Co-Author: CEO, Siemens USA Consultant, sanofi-aventis Group Research support, sanofi-aventis Group Consultant, Bayer AG Research support, Exelixis, Inc Research support, Schering-Plough Corporation Consultant, Mitsubishi Corporation Consultant, Biogen Idec Inc Research support, Takeda Pharmaceutical Company Limited

Mark Gilbert, Abstract Co-Author: Nothing to Disclose

Jerrold L. Boxerman MD, PhD, Abstract Co-Author: Medical Advisor, Imaging Biometrics, LLC

To study factors associated with inter-reader variability (IRV) in initial area under the gadolinium concentration curve (IAUGC), transfer constant (Ktrans) and fractional volume of the extracellular extravascular space (ve) parameters derived from DCE-MRI.

18 patients (mean age 57.1±13.3 years, 10 men) with recurrent glioblastoma treated with bevacizumab and either temozolomide or irinotecan had analyzable DCE-MRI scans as part of the ACRIN 6677/RTOG 0625 trial. Two teams consisting of a trained neuroradiologist and technologist independently measured volumes of tumor-related enhancement (vTRE) on 3D SPGR images. Volumes of interest were registered to DCE-MRI images and parameter maps. IAUGC was defined as the area under the gadolinium contrast agent-time curve (ACTC) in TRE divided by ACTC in a vessel over the first 90 seconds. One open source software was used to derive Ktrans and ve from a standard Tofts model (www.dcemri.org). The effect of time of scan (baseline vs. post-treatment), parameter map quality (graded on a 1–3 scale), scanner magnet / sequence (3T GE vs. 1.5T Siemens vs. 1.5T GE), average tumor volume, and reader variability in tumor volume (percent difference in vTRE) on IRV (percent difference in parameter) was studied using intraclass correlation coefficient(ICC) and linear mixed models.

78 (14 baseline and 64 post-treatment) scans from 4 centers were evaluated. Overall mean IRV for mean and median IAUGC, mean and median Ktrans and median ve were 17.6±24.1, 16.8±23.3, 27.2±34.4, 16.1±27.0 and 26.6±34.3% respectively. ICCs for these metrics ranged from 0.905 (mean Ktrans) to 0.996 (median Ktrans) on baseline scans and from 0.477 (mean Ktrans) to 0.760 (median IAUGC) for post-treatment scans. Lack of reader agreement on tumor volume was significantly related to IRV for all parameters (p<0.002). Scanner magnet/sequence and average tumor volume were related to IRV only for mean Ktrans (p<0.05). Image quality score was related to IRV only for median Ktrans (p=0.02).

Differences in reader performance on tumor segmentation is a significant source of IRV for all parameters.

Improving reader agreement in deriving segmentations or, alternatively, use of automated segmentation techniques may reduce the variability in parameters derived from DCE-MRI.  Funded by NCI U01-CA080098 and U01-CA079778.

Barboriak, D, Zhang, Z, Desai, P, Safriel, Y, McKinstry, R, Bokstein, F, Sorensen, A, Gilbert, M, Boxerman, J, Inter-reader Variability in Dynamic Contrast-enhanced MR Imaging (DCE-MRI) of Patients with Recurrent Glioblastoma: Results from the Multi-center ACRIN 6677 / RTOG 0625 Study.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045441.html