Prioritizing interpretability in an algorithm for thrombolysis selection following stroke

  • Selecting patients who would benefit from thrombolysis following acute ischemic stroke can be achieved by a predictive algorithm. Algorithm interpretability is critical for widespread clinical uptake.
  • The advanced version of classical-k-nearest neighbors classification algorithm (KNN) outperformed the classical KNN algorithm in terms of predictive power (P=0.019). The advanced algorithm identified clinical features (onset time, diabetes, and baseline National Institutes of Health Stroke Scale scores) with significant effects on the output.
  • The authors concluded that advanced classical KNN offers an accurate and easy to interpret algorithm for the selection of patients for thrombolysis.