Comparison of diffusion-weighted MRI and CT in acute stroke

Objective: To compare diffusion-weighted MRI (DWI) and CT with respect to accuracy of localizing acute cerebral infarction; sensitivity, specificity, and interrater reliability for identifying more than one-third middle cerebral artery (MCA) territory involvement; and correlation of acute lesion volume with final infarct volume. Method: Nineteen consecutive stroke patients underwent CT and DWI within 7 hours of stroke onset and a follow-up DWI examination 36 hours after symptom onset, which served as the "gold standard" for lesion location and extent of MCA involvement. Each scan was evaluated for acute ischemic lesions by two experienced observers. After 30 days, T2-weighted MRI was obtained for assessment of the final infarct volume. Results: The acute CT and DWI scans were obtained on average 2.6 and 5.1 hours after symptom onset. On DWI the acute lesion was identified correctly in all instances and on CT it was identified correctly in 42 to 63% of patients. Sensitivity for detection of more than 33% MCA involvement was better for DWI (57 to 86%) than for CT (14 to 43%), whereas specificity was excellent for both. Interrater reliability was moderately good for both (kappa, 0.6 for DWI; 0.5 for CT). A positive correlation (r = 0.79; p = 0.001) existed between lesion volume on acute DWI and final infarct volume, whereas no correlation was found between CT volume and final infarct volume. Conclusion: When compared with CT, DWI was more accurate for identifying acute infarction and more sensitive for detection of more than 33% MCA involvement. In addition, lesion volume on acute DWI, but not on acute CT, correlated strongly with final infarct volume. Additional studies are required to demonstrate whether these advantages of DWI are clinically relevant in the management of patients with acute stroke.