Related Experiment Videos
[Development of a model for the diagnosis and risk classification on anthrax through artificial neural network]
Jia-xin Han1, Hong-yan Xiong, Ting-hui Zhang
1Department of Epidemiology, School of Military Preventive Medicine, Third Military Medical University, Chongqing 400038, China.
Objective:
Based on data through clinical and epidemiological studies, a model regarding the diagnosis and risk classification on anthrax was developed by artificial neural network (ANN). The model could integrally diagnose anthrax cases, judge the risk tendency in time, and increase the ability of recognizing the anthrax accidents.
Methods:
Clinical, laboratory and epidemiological data from anthrax cases was collected and analyzed. The important factors which could greatly influence the results on diagnosis and judgment was chosen and used as the neural units. Through the use of artificial neural network analytic method (back propagation, BP), an intelligent model on the diagnosis and risk classification was developed.
Results:
Results from the multivariate analysis revealed that: 11 factors including incubation period, chest radiographic and microscopic findings, characteristics on professions etc. were associated with the judgment on the diagnosis and intensity of the epidemics. Through 500 times training on the neural network, the performance error decreased from 6.669 59 to 5.051 19 x 10(-11). The model was then validated. With 100% average correct rate, the predictive value was good.
Conclusion:
It was feasible to use the disease information to develop a diagnosis and risk classification model on anthrax by artificial neural network. With 100% average correct rate, the established model was valuable in practice.
