Research On Medical Diagnostic Methods Based On Deep Learning

On the traditional medical field,patients must talk with their doctors face to face about their own symptoms,and then doctors use their own experience to determine the patient’s condition.On the hand,this method makes some patients unable to get expert guidance in a timely manner.On the other hand,difficulties in registration,scarcity of field experts,high cost and other issues.In response to these problems and in recent years deep learning has been widely used and applied in all walks of life,this paper introduces the deep learning model into medical disease diagnosis so that the patient can describe his/her illness conditions and the system gives the most possible risk disease results.This paper first uses the named entity recognition work to to extract the data content related with disease diagnosis in the patient’s disease description,followed by disease diagnosis.In the work of named entity recognition,based on the original Bi-directional Long Short-Term Memory neural network and Conditional Random Field model.By introducing a variety of different embedding layers,the accuracy of named entity recognition work is improved.In the disease diagnosis,the introduction of convolutional neural networks to solve the traditional methods of disease diagnosis which is cumbersome and low accuracy.At the same time,aiming at the over-fitting problem of the original convolutional neural network text classification model,a convolutional neural network based on the mean feature layer and a convolutional neural network text classification model based on the all features connected layer are proposed.And the new model is more excellent than the original one under the experiments.Because named entity recognition work and disease diagnosis work as two different network models,it is easy to cause deviation explosion.At the same time,the complexity is higher in data preprocessing,training model and updating the model.Therefore,this paper combines the first two parts of the work and proposes an end-to-end medical disease diagnosis model.Through experimental verification,the end-to-end network model is not only more convenient,but also has a better performance in medical disease diagnosis.