Skin Disease Classification with Help of Artificial Intelligence Techniques

Medical domain has seen rapid growth of disease detection using different artificial intelligence techniques. The disease can be detected by analyzing images of diseased organs or using pathological test numeric values. This study explores the usage of traditional machine learning algorithms and deep learning methods to diagnose nine different kinds of skin diseases using infected skin images. As convolution neural networks (CNN) have shown huge promise in image classification hence three pre-trained models have been used to detect skin disease. A self-proposed CNN has also been used for the diagnosis of skin disease from images. Five traditional machine Learning models have also been used for skin disease diagnosis using a few hand-crafted image features. The best performance using machine learning methods was obtained using Random Forest giving an accuracy of 64.83% and F1-score of 0.7538. The best accuracy of 77.77% was obtained using the proposed CNN. The reason for not being able to attain high accuracy can be due to the similarity in a few types of skin diseases.