Identifying lesion growth with MR imaging in acute ischemic stroke

Michael S. Bristow, Brett W. Poulin, Jessica E. Simon, Michael D. Hill, Jayme C. Kosior, Shelagh B. Coutts, Richard Frayne, J. Ross Mitchell, Andrew M. Demchuk

Research output: Contribution to journalArticlepeer-review

9 Scopus citations


Purpose: To determine whether different MR diffusion- and perfusion-weighted imaging (DWT and PWI) parameters are important in distinguishing lesion growth from the acute lesion and from oligemia. Materials and Methods: MR DWI and PWI were acquired from thirteen patients. We defined three regions: (i) LESION - intersection of acute and final lesions. (ii) GROWTH - portion of final lesion not part of acute lesion. and (iii) OLIGEMIA - region of perfusion abnormality not part of either the acute or final lesions. We used logistic regression modeling to distinguish GROWTH from LESION and from OLIGEMIA on a voxel-wise basis using DWI- and PWI-based parameters. Final models were selected based on the Wald statistic and validated by cross-validation using the mean (± standard deviation) area under the curve (AUC) from receiver operating characteristic analysis. Results: The final model for differentiating GROWTH from LESION included DWI, the apparent diffusion coefficient (ADC), cerebral blood flow (CBF) and tissue type (AUC = 0.939 ± 0.028). The final model for differentiating GROWTH from OLIGEMIA included DWI, ADC, CBF, and time-to-peak (AUC = 0.793 ± 0.106). Conclusion: Different MR parameters are important in differentiating lesion growth from acute lesion and from oligemia in acute ischemic stroke.

Original languageEnglish (US)
Pages (from-to)837-846
Number of pages10
JournalJournal of Magnetic Resonance Imaging
Issue number4
StatePublished - Oct 2008


  • Acute stroke
  • Diffusion weighted MRI
  • Logistic regression
  • Perfusion weighted MRI

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'Identifying lesion growth with MR imaging in acute ischemic stroke'. Together they form a unique fingerprint.

Cite this