An Integrated Deep Learning Framework Approach for Nail Disease Identification

Abstract

Nail Diseases refer to some kind of deformity in the nail unit. Although the nail unit is a skin accessory, it has its own distinct class of diseases as these diseases have their own set of signs, symptoms, causes and effects that may or may not relate to other medical conditions. Recognizing nail diseases still remains an unexplored and a challenging endeavor in itself. This paper proposes a novel deep learning framework to detect and classify nail diseases from images. A distinct class of eleven diseases i.e. onychomycosis, subungulal hematoma, beau's lines, yellow nail syndrome, psoriasis, hyperpigmentation, koilonychias, paroncychia, pincer nails, leukonychia, and onychorrhexis. The framework uses a hybrid of Convolutional Neural Network (CNNs) for feature extraction. Due to the non-existence of a meticulous dataset, a new dataset was built for testing the enactment of our proposed framework. This work has been tested on our dataset and has also been compared with other state-of-the-art algorithms (SVM, ANN, KNN, and RF) that have been shown to have an excelled performance in the area of feature extraction. The results have shown a comparable performance, in terms of differentiating amongst the wide spectrum of nail diseases and are able to recognize them with an accuracy of 84.58%.