An artificial neural network model for the evaluation of carotid artery stenting prognosis using a national-wide
Abstract:
Stroke is a serious health problem in many countries. About 20% of ischemia stroke involves carotid stenosis. Neck carotid ultrasound is fast, secure and convenient way to detect carotid artery stenosis. Carotid artery stenting (CAS) has become a popular treatment for cerebrovascular stenosis in recent years. However, CAS may also induce the occurrence of major adverse cardiovascular events (MACE) in older patients. Hence the evaluation the CAS prognosis is important. In this study, we attempted to construct a model for the evaluation of CAS prognosis by artificial neural network (ANN). The data of 317 patients from Taiwan Nation Health Insurance Research Database (NHIRD) was used to train and test the constructed ANN model. The input features contain 13 clinical risk factors and the output is the occurrence of MACE. In results, an ANN model of multilayer perceptron with 18 neurons in hidden layer was developed. The performance of this model is with sensitivity 89.4%, specificity 57.4%, and accuracy 82.5% in testing group as well as with sensitivity 85.8%, specificity 60.8% and accuracy 80.76% in overall patients. The results revealed that the created ANN model achieved a good performance in prediction of MACE in patients needing CAS treatment. Such a model will be helpful for prevention of high-risked patients with CAS and could serve as a reference of communication when neurologists refer patients and before patients are treated by cardiologists.
