Research On Image Diagnosis Method For Face Skin Health

With the development of deep learning technology,medical image analysis technology based on deep learning has become a hot research topic in the field of Computer Aided Diagnosis.Due to the variety of skin diseases and the disunity of skin disease identification criteria and diagnostic criteria,how to achieve automatic diagnosis of skin diseases has become a bottleneck problem in the field of dermatological image diagnosis.Therefore,this paper combines deep learning to carry out related research on automatic dermatological image diagnosis methods.The work of this paper mainly includes the following two aspects:For the problem that the disunity of skin disease identification criteria and diagnostic criteria,and the current Computer Aided Diagnosis System of skin disease uses identification criteria,this paper propose an image automatic diagnosis for facial skin disease based on region proposal.The method can be divided into skin disease detection step and skin disease classification step.Firstly,based on improved Faster R-CNN algorithm,the lesion area of skin was detected and the number of lesions was counted.Secondly,the detection results are input into the residual network for classification.Finally,the classification results are combined with the number of lesions and personal information to automatically diagnose skin diseases.This method combines the object detection network with the classification network,avoiding the imbalance of the number of category samples in the facial skin disease data set.For the problem of imbalance sample number among different facial skin disease in data set.This paper proposes an image automatic diagnosis for facial skin disease based on Cycle-Consistent Adversarial Networks.Firstly,the method expands the skin disease image with less data in the data set of facial skin disease by using the Cycle-Consistent Adversarial Networks,and then the expanded results are evaluated and labeled by professionals,so that different types of samples in the data set can reach a balanced state.Secondly,on the basis of proposed region-based skin disease detection network,the balanced data set is used to train the skin disease detection network,so that it can independently detect the skin disease area,classify and count the number of lesion areas,and automatically diagnose the skin disease image combined with personal information.This method solves the problem of imbalance of classes in facial skin disease data set from the point of view of expanding data set,and improves the accuracy of image automatic diagnosis algorithm.In this paper,combined with deep learning technology,aiming at the problem of the inconsistency of the criteria for differentiation and diagnosing skin diseases.The automatic diagnosis technology of skin disease image based on unbalanced data set and balanced data set.The proposed method is to explore the automatic diagnosis technology of facial skin image.