Publication

Abstract : Skin is the largest organ of human body which acts as an effective barrier against external threats. Normally, skin diseases arise from bacteria, viruses, fungi, autoimmune issues, UV exposure, pollutants, and genetic factors and the symptoms can range from mild irritations like rashes to serious conditions like melanoma. But, unfortunately, many people ignore early signs of skin variations that led to delayed diagnosis and reduced treatment process. Many traditional methods in were slow and prone to human error. Hence, to overcome the drawbacks of traditional methodologies and to improve the diagnostic accuracy, this study combines advanced image preprocessing techniques with effective segmentation methods to enhance lesion visibility. After validating with multiple techniques, the selected best-performing techniques namely Gaussian Filters (GFs) for contrast enhancement and Morphological Operations (MOs) for segmentation are integrated with a variety of Machine Learning (ML) and Deep Learning (DL) classifiers to optimize SD detection. Followed by this, the hybrid models, especially the Convolutional Neural Network-based Support Vector Machine (CNN-SVM) combination, showed outstanding performance by achieving high classification accuracy of 98.77%. Hence, this proposed model effectively outperformed the conventional methods and also helped the dermatologists for faster and more reliable SD diagnosis.