An Approach to Detect Alopecia Areata Hair Disease Using Deep Learning

Hair is one of the defining characteristics of human beings. Hair is a protein filament that grows from follicles found in the dermis. Keratin is the source of hairs and nails. There are different kinds of hair diseases in humans like Alopecia Areata, Congenital Hypotrichosis, Telogen Effluvium, Scarring Alopecia, Anagen Effluvium, etc., in which Alopecia Areata is the most common. It is the subject of research for early identification. People typically lose so many hairs per day. Hair loss is a very common problem nowadays seen in more or less in every age group, Alopecia is similar to hair loss. The reason behind hair loss is some vitamin deficiency that helps to build the sources like keratin. At the age of 32–36 near 40% people are facing the problem of hair loss. In the dataset there are 609 images in which 310 images of normal hair and 299 images which are infected from Alopecia Areata. In this study, we used pre-trained models (VGG-16, VGG-19, SqueezeNet and Inception-V3) for feature extraction. The dataset is splitted into the ratio of 7:3 in which 70% and 30% used as training data and testing data, respectively. Then, apply the proposed algorithms like ANN, SVM, logistic regression, Naive Bayes and get the maximum 98.3% accuracy.