Purpose: To assess the use of nonstandard, variable window width and level review settings in computed tomography (CT) without contrast material administration in the detection of acute stroke.
Materials and methods: Nonenhanced CT was performed in 21 patients with acute (< 6 hours) middle cerebral arterial stroke and nine control patients. Two blinded neuroradiologists rated all scans for presence of parenchymal hypoattenuation. Images were reviewed at a picture archiving and communication system (PACS) workstation, with standard, locally determined center level and window width settings of 20 and 80 HU and with variable soft-copy settings initially centered at a level of 32 HU with a width of 8 HU. Reviewers altered settings to accentuate gray and white matter contrast.
Results: With standard viewing parameters, sensitivity and specificity for stroke detection were 57% and 100%. Sensitivity increased to 71% with variable window width and center level settings, without loss of specificity. Receiver operating characteristic analysis revealed a significant improvement in accuracy with nonstandard, soft-copy review settings (P = .03, one-tailed z test).
Conclusion: In nonehanced CT of the head, detection of ischemic brain parenchyma is facilitated by soft-copy review with variable window width and center level settings to accentuate the contrast between normal and edematous tissue.