Comparing the Performance of Mathematical Models for Surgical Decisions on Head Injury Patients

This paper compares three mathematical models for surgical decisions on head injury patients. A logistic regression and two neural network models were developed using a large clinical database. Using randomly selected 9480 cases as the training group and another 3160 cases as the validation group. We evaluated the performance of a logistic regression model, a multi-layer perceptron (MLP) neural net and a radial-basis-function (RBF) neural net in terms of their accuracy in predicting physician’s decision on openskull surgery. The resultant area under ROC curve for logistic regression, MLP and RBF neural nets are 0.761, 0.897 and 0.880 respectively. The results suggest that neural networks may be a better solution for complex, non-linear medical decision support systems than conventional statistical techniques such as logistic regression.