A Deep Hybrid Learning Approach For Nail Diseases Classification

In medical domains, the appearance of fingernails can provide clues to underlying systemic diseases or nutritional imbalance; the neglection of such clues could lead to unwanted health complications and less chance for recovery. In this paper, a Deep Hybrid Learning (DHL) approach was proposed to detect nail-based diseases, where a Deep Learning (DL) model is used for feature extraction, and a traditional Machine Learning (ML) classifier is used for classification. The aim is to classify three nail diseases: melanoma, beau's nails, and eczema, in addition to healthy nails. Further, the proposed approach is compared to the transfer learning approach, where a pre-trained model is used for feature extraction and classification. The experiment results indicate that the DHL approach is superior to the transfer learning approach. Specifically, the architecture where the DenseNet201 pre-trained model is used for feature extraction and the SGDClassifier is used for classification, as it achieved an accuracy of 94%.