Research And Implementation Of Early Skin Cancer Recognition System Based On Convolutional Neural Network

With the advent of the era of artificial intelligence,more and more fields began to rely on artificial intelligence technologies,especially in the medical field.Cancer is one of the biggest problems in the medical field,if early detection and early treatment can be achieved,the possibility of cure will greatly increase.Malignant skin cancer,as one of the cancer types with the highest lethal rate in recent years,rely highly on doctors' experience and is unable to be detected in time.Therefore,if artificial intelligence technology can be used to help doctors detect skin cancer earlier,or let everyone be able to detect skin lesion anytime and anywhere,it will have some practical significances.However,due to the problems of insufficient sample size and small differences between skin cancer classes in dermoscopy images,in this study,methods such as generative adversarial networks and data enhancement are mainly used to process the data,and we use convolutional neural networks to train the data,and finally implement an online skin cancer recognition system based on Java Language,which facilitates early skin cancer detection.The main research content and results of the paper are as follows:1.Because of the insufficiency of dermoscopy pictures,in the research,the dermoscopy pictures that have been collected are used to train the generative adversarial network for getting more images,so that a sufficient number of picture samples are obtained,and the next network training can be carried out.On the other hand,image augment methods such as random image rotation and random flip are also used in this study to further expand the sample picture data set size.2.Because the data set has reached a certain scale,the next step is to complete the selection of the neural network and the construction of the network model.After comparing many network parameters horizontally,the current network structure is selected,which can reach a good balance between training time and accuracy.3.The third part of the research is the design and implementation of the online skin cancer recognition system.From the three aspects of requirements analysis,summary design,overall design and implementation,the implementation of the system is analyzed and studied in detail.As result,the basic functions of the system are both realized.Finally,after completing the three parts of data generation,model design and system implementation,I summarized my research work,and do some analysis over the current improvement directions.