Disease Prediction and Early Intervention System Based on Symptom Similarity Analysis
Peiying Zhang, Xingzhe Huang, Maozhen Li · IEEE Access · 2019
With the development of computer technology, the electronic of medical data has become a reality. Now, how to analyze the data sufficiently to predict patient's disease and conduct early intervention has become a focused research direction. The patient's intuitive expression of feelings is also an aspect that cannot be ignored. Doctors record the pathological characteristics of patients in system. In the paper, we proposed a sentence similarity model to carry out symptom similarity analysis to achieve elementary disease prediction and early intervention, which makes use of word embedding and convolutional neural network (CNN) to extract a sentence vector that contains keyword information about the patient's feelings and symptoms. In order to increase the accuracy of sentence similarity computation, this model integrated syntactic tree and neural network into the computation process. Our main innovation is to use symptom similarity analysis model for disease prediction and early intervention. In addition, the SPO kernel is also one of the innovations. Finally, the results of experiment on Microsoft research paraphrase identification (MSRP) indicated that our model can achieve an excellent performance reached 83.9% in the terms of F1 and accuracy. Furthermore, we also conducted experiments on the data of the Semantic Textual Similarity task. Pearson correlation coefficient indicates that our result is closer to the gold standard scores, which illustrates that it can extract the key information of sentence well to realize the prediction of disease and carry out early intervention.