Repeatability of quantitative parameters derived from diffusion tensor imaging in patients with glioblastoma multiforme

Abstract

To quantify the repeatability of apparent diffusion coefficient (ADC) and fractional anisotropy (FA) in patients with glioblastoma multiforme. IRB approval and informed consent were obtained for this Health Insurance Portability and Accountability Act-compliant study. Sixteen patients with glioblastoma multiforme underwent MR imaging at two time points without interval intervention. ADC and FA maps were registered to the contrast-enhanced and fluid-attenuated inversion recovery (FLAIR) image volumes. Volumes of tumor-related enhancement (TRE) and FLAIR signal abnormality (FSA) were defined using a semiautomated segmentation technique. Repeated observations of mean ADC and mean FA were highly consistent within both TRE (ADC: r = 0.947,P < 0.0001; FA: r = 0.947, P < 0.0001) and FSA (ADC: r = 0.979, P < 0.0001; FA: r = 0.972, P < 0.0001). Within TRE, repeatability coefficients and 95% confidence intervals (CIs) for change measured 0.104 x 10(-3) mm(2)S(-1) and 7.4% (ADC) and 0.0196 and 13.9% (FA), respectively. Within FSA, repeatability coefficients and 95% CI for change measured 0.071 x 10(-3) mm(2)S(-1) and 5.2% (ADC) and 0.0159 and 8.7% (FA), respectively. To detect 10% changes in mean ADC, sample sizes of nine (TRE) and six (FSA) patients would be required. The same change in mean FA would require sample sizes of 21 (TRE) and 10 (FSA) patients, respectively. Changes after therapy greater than the repeatability coefficient or 95% CI for change are unlikely to be related to variability in the measurement of ADC and FA.