Comparison of Models for Predicting the Outcome of Craniocerebral Injury by Using Machine Learning
Introduction
Craniocerebral injury leads to a high probability of death and disability, the accurate and timely prediction of the outcome of this clinical condition is the key point in diagnosis and treatment. However, the traditional evaluation systems of craniocerebral injury mainly depend on the experience of experts and is often not objective enough. Therefore, we built models to predict the outcome by using machine learning to improve the prediction accuracy. However, the generalization error of existing tools may result in wandering in accuracy in different machine learning model.. Therefore, based on the clinical data of patients with craniocerebral injury, we established multiple models using different algorithms to find the appropriate model to improve the accuracy and objective of prediction.
Methods
Date was collected from patients with craniocerebral injury admitted to the Department of trauma surgery in Sichuan Provincial People's Hospital, Sichuan Academy of Medical Sciences. Inclusion criteria: 1) age 4-81, 2) the head injury history.. We used these data to build a variety of machine learning models including Decision tree, Linear Discriminant, SVM , KNN and Boostedtrap. and compared their performances by means of Receiver Operating Characteristic (ROC) and Area under Curve (AUC), Accuracy, F-Score, Precision Ratio and Recall Ratio, Training Time. All of these results are compared to the classical Logistic regression results. Model building and evaluation using MATLAB2016a(MathWorks, America)on the windows10.
Results
127 patients with craniocerebral injury were enrolled. The accuracy of all machine learning models was between 86.6% and 94.5%, and the Logistic regression's accuracy is …., which indicated that the establishment of machine learning models can be regarded as an effective way to predict the outcome of patients with craniocerebral injury.. Different machine learning models result data have differ performance in our clinical dataset. but except for ****, all of the performance of our algorithm is better than that of the classical Logistic Regression.
Conclusion
In this study, we found that using machine learning models can predict the outcome of patients with craniocerebral injury in a better way. In addition to more accurate predictions, some machine-learning algorithms can provide better interpretative analysis of clinical data