Prognosing acute cerebral infarction using a synthesis of multiple indicators / 中华物理医学与康复杂志

ABSTRACT

Objective To predict the prognosis after acute cerebral infarction using a combination of indicators. Methods Two hundred and seventeen patients with acute cerebral infarction admitted from October 2005 to December 2008 were studied. Logistic regression analysis of the data from 112 of the patients admitted from October 2005 to March 2007 was used to select 20 indicators for study. The indicators were combined into prognostic indexes using a multi-layer perception (MLP) neural network (NN) model. Data on the subsequent 105 patients were usedto appraise the model. Results The agreement of the prediction results of the NN model with the real recovery observations was rated as "excellent" in 39 cases, "OK" in 27 and "bad" in 32. The sensitivities were 95.1%,87.1% and 96.9% respectively. On average, the differences between the predicted results with the NN model andthe real recovery were not significant. Conclusion The NN model delivered good precision in predicting the outcome of acute cerebral infarction and it is worthy of further investigation.