Research On The Cerebrovascular Diseases Diagnosis Expert System Based On The Back Propagation Neural Network

Cerebrovascular disease in the elder has become endanger the health and lives of major diseases. According to the pathological change cerebrovascular disease divided into hemorrhagic and ischemic two categories. The main symptoms are headache, dizziness, nausea, vomiting, hemiplegia, migraine are sensory dysfunction, aphasia, drowsiness or coma.Artificial neural network is the fastest development in the near term outcome of the research in the field of artificial intelligence. BP neural network model is the most widely used category in neural networks. It gets the minimum mean square difference between actual output and the desired output through gradient algorithm, and reverse flow and amend errors . With that error back-propagation ongoing, the network on the correct input mode response rate has been increasing. With appropriate parameters, the network will converge to have a smaller variance.This paper introduces the theory of BP neural network. Because of the slow convergence of the defects, the paper replaces the gradient descent algorithm with L-M algorithm . The convergence of L-M algorithm is faster than gradient descent algorithm. The model consists of 16 nodes input layer, hidden layer 11 nodes, 4 output layer nodes. Error learning algorithm used numerical L-M Optimization algorithm, which greatly accelerates the speed of a network of training than gradient descent algorithm.In this paper,the BP neural network model is designed by Visual C + + 6.0. In order to enhance modularity and facilitate, the training algorithm and matrix used dynamic link library. The system is divided into two parts: Training and diagnostic. First, the system is trained according to sample data set, and then save trained weight matrix. At the time of diagnosis the system inputs the trained weights matrix and caculate , and then compare the output with the standard output, last get the diagnosis result. The system that used for diagnostic tests achieved good results. That artificial neural network method in the diagnosis of the disease is feasible and effective.