Temporal and Spectral feature fusion of Dermoscopic Images for Skin Cancer Classification using Support Vector Machine

  • G.Prathima, Dr.R.Shoba Rani
Keywords: skin cancer, wave atom transform, feature fusion, support vector machine, spectral features.

Abstract

Skin cancer is a dermatological disease appearing like normal skin that might cause health problems. Cancer cells are abnormal and multiply in an uncontrollable way, particularly in their malignant stage. Among the skin cancers, malignant is the most dangerous one. Thus, an automated classification system for malignant melanoma is more important than studying other type of skin cancers. In this study, Temporal and Spectral Feature Fusion (TSFF) approach for skin cancer classification is developed using Wave Atom Transform (WAT) and Support Vector Machine (SVM). Temporal features are extracted directly from the raw images after preprocessing by median filtering whereas spectral features are extracted from WAT transformed image. After extracting features, SVM classifier is employed to conduct final diagnosis. Fused features are mapped into higher dimensional feature space by kernels such as SVM-Linear, SVM-Polynomial, SVM-Radial Basis Function (RBF) and SVM-Quadratic. Experimental results prove that the 1st stage (normal/abnormal) classification fives 98.5% accuracy and 2nd stage (benign/malignant) classification gives 100% accuracy for skin cancer classification on PH2 database images. 

Published
2021-09-10
How to Cite
Dr.R.Shoba Rani, G. (2021). Temporal and Spectral feature fusion of Dermoscopic Images for Skin Cancer Classification using Support Vector Machine. Design Engineering, 6982-6997. Retrieved from http://www.thedesignengineering.com/index.php/DE/article/view/4189
Section
Articles