Study On Prognostic Prediction Of Spinal Cord Injury And Intertrochanteric Fracture Based On ANN

Because the interdisciplinary collaboration of medical science and engineering has attracted more and more attention as a new hot spot, this thesis is mainly about to find the joint point of medical science and engineerin and unfold a deep research on prognostic prediction of related orthopedic injuries based on ANN so as to comply with the needs of doctors and patients, which is supported by a certain grade-three general hospital in Western Liaoning. Considering the complexity and diversity of prognostic prediction of related orthopedic injuries, cervical spinal cord injury and intertrochanteric fracture are chosen for deep study on the optimization problem of prediction on key points of medical process, and the real value of interdisciplinary collaboration is embedded in a orthopedic injury prognosis prediction system.As is known to all, there are distinct nonlinear characteristics in prognostic prediction of cervical spinal cord injury and intertrochanteric fracture, which is difficult to establish the accurate mathematical model. Whereas Artificial Neural Network is well suited for predicting the injury recovery in consideration of its excellent abilities of nonlinear fitting and approximation. So the intensive research work of prognostic prediction of related orthopedic injuries is carried out for the first time based on ANN.Firstly, the Back Propagation Neural Network (BP) and Probabilistic Neural Network (PNN) are selected for prediction tools, meanwhile, a comprehensive evaluation index system of prognostic prediction with weights order is established based on Analytic Hierarchy Process so as to provide standards and methods of choosing appropriate network.Secondly, combining with the literature, the reasons of low precision in preliminary tests and defect of traditional neural network algorithm are found based on the sufficient analysis. And then, a particle swarm optimization based on variety inertia weight is used to optimized the smoothing parameters of PNN because of its lack of diversity which cannot meet the adaptive requirement. Meanwhile, a.improved genetic algorithm based on PSO is used to optimized the hidden layer nodes and link weights of BP network because of the low convergence rate and local minimal, on which the IPSO-PNN prediction model and PGA-BP prediction model are build for prognostic prediction of related orthopedic injuries.Thirdly, the models are trained by a large-number of training samples with statistically significant and tested by testing samples to test the generalization ability. Next, simulation results show that the performance indicators such as prediction accuracy, stability and generalizetion ability are satisfied. And then, on the basis of the simulation results, the advantages and disadvantages of BP model and PNN model are summarized by comparing the output characteristic of ANN prediction model and echoing the weights order of comprehensive evaluation index system in this paper, at the same time, highlights the advantages of ANN by comparing ANN model and logistic model in order to provide the diversified choice for medical staffs.Finally, a orthopedic injury prognosis prediction system is designed based on theoretical study and simulation experiment with GUI, which is promoted on a small scale in cooperative hospital, and the feedback shows that the prediction system is running well and reaches the anticipated object.