Leveraging Deep Learning for Nail Disease Diagnostic
Mahendra Mehra, Steve D'Costa, Ryan D’Mello, Joseph George, Dhananjay Kalbande
Abstract
There are many types of nail diseases, and although the nail is just a small part of our body, the nail unit can be a significant sign of some underlying disease based upon its features. Subungual Melanoma remains a life-threatening disease. Although it can be cured in its early stages, it is difficult to diagnose it during that time. It often leads to a late disease diagnosis, which makes it difficult to cure the disease. The present medical tests for disease diagnosis are costly and not available in rural parts. This project proposes an AI approach to detect and classify nail diseases from images. A distinct class of two diseases i.e., yellow nail syndrome and Subungual Melanoma, is classified in this project. The project uses an Artificial Neural Network based model for training and testing. We have used the concept of transfer learning for the training model because making a model from scratch is not feasible with fewer data and less GPU. The model is an implementation of VGG16 by Keras framework with two added layers of ANN. Since we could not find any dataset, we made a new dataset for our proposed framework. This work has been tested on our dataset and has shown to have an excellent performance in identifying diseases.