Abstract: ObjectiveTo evaluate a standardised ’stoke window’ in non-enhanced computed tomography (NECT) of the brain in the diagnosis of early ischemic changes (EIC). MethodsRetrospective analysis was performed using clinical data from 168 patients who had been diagnosed with EIC by MRI or DSA within the first 360 minutes of an onset. Two resident physicians of the Radiology Department used the default window (100 HU, 35 HU) and a ’stoke window’ (30 HU, 35 HU) to assess EIC and its various signs every other two weeks. The final confirmation by two attending physicians was set as the gold standard. The diagnostic accuracy of the two windows was compared. ResultsA total of 113 patients (67.3%) were confirmed with EIC signs, including 59 patients with hyperdense middle cerebral artery sign (HMCAS), 5 with hyperdense posterior cerebral artery sign (HPCAS), 42 with disappearing basal ganglia sign, 35 with loss of the insular ribbon sign, 29 with cortical sulcal effacement sign and 31 with focal hypodensity sign. The resident physicians detected 72 EIC cases (42.9%) using the default window, which was remarkably reduced compared with "the gold standards" (P<0.01). In contrast, 107 cases (63.7%) were detected when the "stroke window" was applied, which was not statistical different from the results by the gold standards (P>0.05). Comparing with the default window, marked more EIC cases were found using the "stroke window" (P<0.01). Except for the HPCAS, the detection rate by the default window based on other EIC signs were substantially decreased compared with the "gold standard" (P<0.05). No statistical difference was found through fewer signs were detected by the residents using the ’stroke window’ than the "gold standard" (P>0.05). Remarkable more signs were detected by the residents using the ’stroke window’ compared with the default window (P<0.05). Conclusions"Stroke window" observation can significantly facilitate the residents to improve their detection rate of EIC.