Prediction model of cervical spine disease established by genetic programming

In order to improve the efficiency of medical diagnosis, this research proposed a predict model that evaluated the cervical spine condition of patients. According to three main index including the severity, cervical curvature and alignment, the predict model we proposed consisted of two stages, including data input and predicting mechanism. In the first stage, based on the magnetic resonance images (MRI) provided by diagnostic radiology department, we used 42 measurements out of 8 attributes as the parameters that may affect the cervical spine condition. Then, in the second stage, the genetic programming (GP) were adopted as the core refereeing engine to construct the prediction tree by training data set. The operation of GP was choosing the different nodes to undergo selection, crossover, and mutation, and, producing complete offspring section in our model. After being classified by the correct ratio, the offspring section went through the process again and again until finding out the most suitable solution. Finally, after ten times average test results from forecasting rules of GP and adjustment the predicting accuracy reached up to 90%.