Automated Detection of skin and nail disorders using Convolutional Neural Networks

Abstract

Dermatological diseases are highly complex to diagnose due to their shared characteristics and subjectivity of human interpretation. Early diagnosis of these diseases helps in getting cured at the right time. Hence, a system has been developed to predict skin and nail disorders in the early stages, when the rate of recovery is high. This, in turn, aids in saving lives and protects a lot of patients from various other health issues. This paper proposes an automated system for the detection of skin and nail disorders by using Convolutional Neural Networks (CNN). The developed model is deployed on a web application named “DermaDoc” to predict these disorders and to aid the users by providing details about the lesion and remedies for temporary relief if available. The diseases considered are Psoriasis, Eczema, Guttate psoriasis, Sebaceous cysts, Paronychia and Yellow nail syndrome. Using various data augmentation techniques and Transfer learning approach on CNN, a higher accuracy of about 92.5% has been successfully achieved.