Acute Ischemic Stroke: Accuracy of Diffusion-weighted MR Imaging—Effects of b Value and Cerebrospinal Fluid Suppression

Abstract

Purpose: To prospectively determine which diffusion-weighted magnetic resonance (MR) imaging technique (ie, conventional diffusion-weighted MR imaging [b=1000 or 1500 sec/mm2] or fluid-inversion prepared diffusion [FLIPD] MR imaging [b=1500 sec/mm2]) is most accurate in depicting acute ischemic stroke at 3 T. Materials and Methods: The Health Research Ethics Board approved this study; written informed consent was provided by all participants or their surrogate. Diffusion-weighted MR imaging was performed in 75 consecutive patients (43 men, 32 women; mean age, 64.0 years) with acute ischemic stroke. Two experienced neuroradiologists determined the presence of hyperacute stroke lesions at diffusion-weighted MR imaging by locating areas of hyperintensity that corresponded to regions with a decreased diffusion coefficient. These findings were used as the reference standard. Four raters who were blinded to patient history assessed all images and apparent diffusion coefficient maps for the presence of changes that were consistent with acute ischemic stroke. Accuracy, sensitivity, specificity, negative predictive value, positive predictive value, and inter- and intrarater reliability scores were calculated for each technique. Results: Specificity, positive predictive value, and accuracy were not significantly different among the techniques. FLIPD MR images obtained with a b value of 1500 sec/mm2 had decreased sensitivity for acute ischemic stroke (mean, 61.8%; 95% confidence interval [CI]: 55.4%, 67.9%) compared with conventional diffusion-weighted MR images obtained with a b value of either 1000 sec/mm2 (mean, 82.5%; 95% CI: 77.1%, 87.0%) or 1500 sec/mm2 (mean, 84.5%; 95% CI: 79.3%, 88.9%). FLIPD MR images also had decreased negative predictive value (mean, 96.5%; 95% CI: 95.7%, 97.2%) compared with conventional diffusion-weighted MR images obtained with a b value of either 1000 sec/mm2 (mean, 98.4%; 95% CI: 97.8%, 98.8%) or 1500 sec/mm2 (mean, 98.6%; 95% CI: 98.1%, 99.0%). Intra- and interrater reliability scores were generally excellent for all three techniques. Conclusion: FLIPD MR images obtained with a b value of 1500 sec/mm2 are less suitable for the detection of acute ischemic stroke owing to a decreased sensitivity and negative predictive value. The performance of the two conventional diffusionweighted MR imaging techniques (b 1000 and 1500 sec/mm2) was equivalent.