Quantitative prediction of ischemic stroke tissue fate

Accurate prediction of ischemic tissue fate could aid clinical decision-making in the treatment of acute stroke. We investigated predictions of tissue fate for three (30-min, 60-min and permanent) stroke models in rats. Quantitative cerebral blood flow (CBF), apparent diffusion coefficient (ADC) and spin-spin relaxation time constant (T₂) were acquired during the acute phase and at the end point followed by histological examination. Probability-of-infarct profiles based on ADC and CBF data were constructed using a training dataset. Probability-of-infarct maps were predicted using only acute stroke data from a separate experimental dataset, revealing the likelihood of future infarction. Performance measures of sensitivity and specificity showed accurate predictions. Sensitivities (mean ± SD) for the 30-min, 60-min and permanent stroke were, respectively, 82 ± 6%, 82 ± 7%, and 86 ± 4%, specificities were 83 ± 5%, 86 ± 5%, and 89 ± 6%, and the areas under the receiver operating curve were 87 ± 3%, 90 ± 4%, and 93 ± 3%. Importantly, to improve prediction accuracy, we took into account regional susceptibility to infarction. Spatial frequency-of-infarct maps were constructed and predictions were made by taking the weighted average of the probability-of-infarct map and spatial frequency-of-infarct map. The optimal weighting coefficient of spatial frequency-of-infarct was small (10%) for the permanent occlusion group but surprisingly large (40%) for the reperfusion groups, indicating that regional susceptibility of infarction was important for accurate prediction in reperfusion stroke. We concluded that the likelihood of cerebral infarction in rats can be accurately predicted and that accounting for regional susceptibility of infarct further improves prediction accuracy. Predictive models have the potential to provide a valuable quantitative framework for clinicians to consider different stroke treatment options.