Hybrid Deep Learning and SVM Approach for Accurate Nail Disease Classification

Nail diseases, such as fungal infections, melanomas, and psoriasis, affect millions of people globally, and their accurate diagnosis is critical for timely treatment. Though these might be subjective and time-consuming for dermatologists, we propose a hybrid approach, in this paper, based on a deep learning model and a machine learning algorithm, for the automatic classification of nail diseases. A deep convolutional neural network, VGG16 CNN, is used as the feature extractor and combined with a Support Vector Machine SVM classifier. This research dataset is a collection of nail images categorized into six classes, namely Acral Lentiginous Melanoma, Healthy Nail, Onychogryphosis, Blue Finger, Clubbing, and Pitting. The VGG16 model, pretrained on ImageNet, extracts the high-level features from the nail images. These features are passed to the SVM classifier for the final classification. Extensive experiments are performed on the nail disease dataset, and the proposed model finally achieves a test set accuracy of 94.36%. The results prove the effectiveness of a combination with VGG16 for feature extraction and SVM for classification, in contrast to the performance of stand-alone deep learning models. This approach uses transfer learning and classical machine learning to efficiently and accurately construct a system for classifying nail diseases. A model developed with this system may be used as a tool for dermatologists that would help in reducing the time and effort spent on diagnosis.