Psoriasis Skin Disease Classification based on Clinical Images

Abstract

Psoriasis is an autoimmune skin disorder that causes skin plaques to develop into red and scaly patches. It affects millions of people globally. Dermatologists currently employ visual and haptic methods to determine a medical issue's severity. Intelligent medical imaging-based diagnosis systems are now a possibility because of the relatively recent development of deep learning technologies for medical image processing. These systems can help a human expert make better decisions about a patient's health. Convolutional neural networks, or CNNs, on the other hand, have achieved imaging performance levels comparable to, if not better than, those of humans. In the paper, a Dermnet dataset is used. Image preprocessing, fuzzy c-mean-based segmentation, MobileNet-based feature extraction, and a support vector machine (SVM) classification are used for skin disease classification. Dermnet's dataset was investigated for images of skin conditions using three classes Psoriasis, Dermatofibroma, and Melanoma are studied. The performance metrics such as accuracy, precision-recall, and f1-score are evaluated and compared for three classes of skin diseases. Despite working with a smaller dataset, MobileNet with Support Vector Machine outperforms ResNet in terms of accuracy (99.12%), precision (98.65%), and recall (99.66%).