Unknown onset time is a common contradiction for anti-thrombolytic treatment of ischemic stroke. T2 relaxation times within the lesion estimate onset time, but accuracy may be affected by dependence on the non-ischemic hemisphere as reference for pre-ischemic values. In hyperacute ischemic stroke patients, we tested a reference-independent approach shown to be an accurate timer in a preclinical stroke model. This involved modelling the T2 distribution within ADC defined lesions to design linear regression models for onset time estimation. The reference-independent approach was accurate in grey matter and for lesions containing grey and white matter, the reference-dependent approach was more accurate.