Researchers involving the DQBM Menze Lab have presented the ISLES’24 dataset, which combines hyperacute CT, postinterventional MRI and clinical outcomes from 245 patients treated for ischemic stroke. The resource includes detailed vessel occlusion and infarct annotations for developing and evaluating medical AI methods.
The public training set contains 149 patients, while a hidden test set of 96 supports unbiased benchmarking. By linking imaging before treatment with MRI acquired two to nine days after successful reperfusion and three-month follow-up data, the dataset enables research into infarct evolution, outcome prediction and clinical decision support.