Summary Background Many patients with stroke are precluded from thrombolysis treatment because the time from onset of their symptoms is unknown. We aimed to test whether a mismatch in visibility of an acute ischaemic lesion between diffusion-weighted MRI (DWI) and fluid-attenuated inversion recovery (FLAIR) MRI (DWI-FLAIR mismatch) can be used to detect patients within the recommended time window for thrombolysis. Methods In this multicentre observational study, we analysed clinical and MRI data from patients presenting between Jan 1, 2001, and May 31, 2009, with acute stroke for whom DWI and FLAIR were done within 12 h of observed symptom onset. Two neurologists masked to clinical data judged the visibility of acute ischaemic lesions on DWI and FLAIR imaging, and DWI-FLAIR mismatch was diagnosed by consensus. We calculated predictive values of DWI-FLAIR mismatch for the identification of patients with symptom onset within 4·5 h and within 6 h and did multivariate regression analysis to identify potential confounding covariates. This study is registered with ClinicalTrials.gov , number NCT01021319. Findings The final analysis included 543 patients. Mean age was 66·0 years (95% CI 64·7–67·3) and median National Institutes of Health Stroke Scale score was 8 (IQR 4–15). Acute ischaemic lesions were identified on DWI in 516 patients (95%) and on FLAIR in 271 patients (50%). Interobserver agreement for acute ischaemic lesion visibility on FLAIR imaging was moderate (κ=0·569, 95% CI 0·504–0·634). DWI-FLAIR mismatch identified patients within 4·5 h of symptom onset with 62% (95% CI 57–67) sensitivity, 78% (72–84) specificity, 83% (79–88) positive predictive value, and 54% (48–60) negative predictive value. Multivariate regression analysis identified a longer time to MRI (p<0·0001), a lower age (p=0·0009), and a larger DWI lesion volume (p=0·0226) as independent predictors of lesion visibility on FLAIR imaging. Interpretation Patients with an acute ischaemic lesion detected with DWI but not with FLAIR imaging are likely to be within a time window for which thrombolysis is safe and effective. These findings lend support to the use of DWI-FLAIR mismatch for selection of patients in a future randomised trial of thrombolysis in patients with unknown time of symptom onset. Funding Else Kröner-Fresenius-Stiftung, National Institutes of Health.