Abstract
To reduce the incidence of cerebrovascular disease and mortality, identifying the risks of cerebrovascular disease in advance and taking certain preventive measures are significant. This article was aimed to investigate the risk factors of cerebrovascular disease (CVD) in the primary prevention, and to build an early warning model based on the existing technology. The authors use the information entropy algorithm of rough set theory to establish the index system suitable for early warning model. Then, using the limited Boltzmann machine and direction propagation algorithm, the depth trust network is established by building and stacking RBM, and the back propagation is used to fine-tune the parameters of the network at the top layer. Compared with the LM-BP early-warning model, the deep confidence network model is more effective than traditional artificial neural network, which can help to identify the risk of cerebrovascular disease in advance and promote the primary prevention.Article Preview
TopIntroduction
Cerebrovascular disease (CVD) has the characteristics of high mortality and high morbidity. In advance to identify the risk of CVD, taking certain preventive and control measures can reduce the incidence and mortality of CVD. Diagnosing CVD is a complex task, but the process is commonly based on the experience and assumptions of doctors. Prakash and Karthikeyan (2021) mentioned that it is possible to make misjudgments because of the doctors’ subjective diagnosis and machine learning technology could become the universal language in emergencies to effectively predict the risk. The diagnosis of CVD has the following problems: (1) The discovery of the disease is too late to do the early detection and treatment; (2) The traditional screening process of CVD risk is not meticulous enough with eight risk factors and the boundary of classification is not obvious, as shown in Figure 1. To optimize the traditional screening process, the risk early warning model is necessary which could select relatively complete indicators and detect the possibility of CVD timely.
Figure 1.
The Traditional Screening Process of a high-risk group of CVD
In this article, the authors use the information entropy algorithm of rough set theory to reduce the initial index system, extract high generalization and high sensitivity indexes, and finally establish the index system suitable for an early warning model. Then, using the limited Boltzmann machine and direction propagation algorithm, the depth trust network is established by stacking Restricted Boltzmann Machine (RBM), and the Back Propagation (BP) is used to fine-tune the parameters of the network at the top layer.
To analyze its applicability, the deep confidence network model was evaluated by the simulation data, and the results were compared with the results of the LM-BP early-warning model. It turned out that the early warning model of CVD risk based on a deep confidence network is more effective than a traditional artificial neural network, which can help identify the risk of CVD in advance.
The aim of this study is to investigate the risk factors of CVD in the primary prevention of CVD, and build an early warning model based on the existing technology. According to the traditional risk screening process, it aims to reduce the misdiagnosis rate caused by patients’ self-report and doctors’ subjective diagnosis. In the background, the authors summarize relevant researches about early warning for CVD risk. In the methods, they complete the design of the risk index system using rough set theory and establish an early warning model for CVD risk based on Deep Belief Networks (DBNs). In the results and discussion, medical data is used to test the model from the dimension of accuracy, training time, etc. The final section concludes this work. The overall frame diagram of the article is shown in figure 2.
Figure 2.
The Overall Frame Diagram of the Article