| Lung cancer is one of the malignant tumors which is the greatest threat to human health and human life with the highest incidence of morbidity and mortality.Some experts pointed out that the lung cancer is caused by the haze,and also mentioned the incidence of lung cancer wouldn't show a downward trend in the next period of time,all of these have showed the importance of prevention of lung disease.At present,the way to prevent lung cancer is to diagnose early and treat early,that is to say we have to diagnose the patient in the state of lung nodule.In recent years,the deep learning has excellent performance in many areas,many large companies which have large data have also turned their attention to the deep learning direction.The deep learning,in fact,is a sub-direction of machine learning,the ultimate goal is to get the feature which can representative samples best from the sample associated with each other in the high-dimensional dimensionality information,then input the feature to the computer and explain it,by the way,the computer would has a simple thinking ability similar to human.At present,There are many products contain this technology in market,it can be said that the deep learning is the most successful research area in machine learning in recent years.In this paper,we first construct the database based on the data samples hand-intercepted from the data,and apply it to the traditional convolutional neural network,which proves the feasibility of this method.Then applied the whole model to the Keras framework,and we analyzed various techniques in traditional model,and we got a better accuracy.Finally,an improved network model is proposed.Compared with the traditional convolutional neural network model,we introduce a more excellent algorithm,which reduces the time consumption.Experiments show that the method can obtain representative features without complex image preprocessing and overcome many inherent problems. |
PDF Full Text Request