Prediction of the prognosis of ischemic stroke patients after intravenous thrombolysis using artificial neural networks

In general, around 80% of all strokes are ischemic. Take caring of the patients who have suffered an ischemic stroke is both expensive and time consuming. It is known that thrombolysis in patients with ischemic stroke can reduce the disability and increase the survival rate, however some patients still have poor outcomes. Therefore, to be able to predict the outcome of ischemic stroke patients after intravenous thrombolysis would be useful while making clinical decisions. In this study, we collected retrospective data of 82 ischemic stroke patients who received intravenous thrombolysis from July 2005 to June 2012 in Tri-service General Hospital. Of these patients, 10 died within 3 months, and only 36 patients made a good recovery. We used STATISTICA 10 software to select the best artificial neural network. The parameters of model 1 were age, blood sugar, onset to treatment time, National Institute of Health Stroke Scale (NIHSS) score, dense cerebral artery sign, and old stroke to predict 3-month outcomes. The parameters of model 2 were age, onset to treatment time, NIHSS score, hypertension, heart disease, diabetes and old stroke to predict the 3-month prognosis. The sensitivity, specificity and accuracy for model 1 were 77.78%, 80.43% and 79.27%, respectively, and 94.44%, 95.65% and 95.12%, respectively, for model 2. Artificial neural networks are used to establish prediction models with good performance to predict thrombolysis outcomes. These models may be able to help physicians to discuss and explain the likely outcomes to patients and their families before thrombolysis treatment.