Our purpose was to compare the detectability and detection rate of acute ischaemic cerebral hemisphere infarcts on CT and diffusion-weighted MRI (DWI). We investigated 32 consecutive patients with acute hemisphere stroke with unenhanced CT and DWI within 6 h of stroke onset. The interval between CT and DWI ranged from 15 to 180 min (mean 60 min). Infarct detectability on CT and DWI was determined by comparing the initial CT, DWI and later reference images in a consensus reading of five independent examiners. The "true" detection rate was assessed by analysing all single readings. Two patients had intracerebral haematomas on DWI and CT and were excluded. There were 27 patients with ischaemic infarcts; all were visible on DWI and proven by follow-up. DWI was negative in three patients without a final diagnosis of infarct (100% sensitivity, 100% specificity, chi2 = 30, P < 0.0001). Ischaemic infarcts were visible on 15 and not seen on 12 CT studies (55 % sensitivity, 100% specificity, chi2 = 1.48, P = 0.224). With regard to the single readings (30 examinations x 5 examiners = 150 readings), 63 CT readings were true positive and 72 false negative (sensitivity 47 %, specificity 86%, chi2 = 2.88, P = 0.089). Of the DWI readings 128 were true positive and 7 false negative (sensitivity 95%, specificity 87 %, chi2 = 70.67, P < 0.0001). Interobserver agreement was substantial for CT (chi = 0.72, 95 % confidence interval, 0.6-0.84) and DWI (chi = 0.82, 95 % confidence interval, 0.46-1). Taken together, detectability and detection rate of acute (< 6 h) hemisphere infarcts are significantly higher with DWI than with CT.