Classification of Dermatological Shap Asymmetry Measures of Skin Lesion

Authors

  • Zofia Stawska Faculty of Physics and Applied Informatics, University of Lodz
  • Piotr Milczarski Faculty of Physics and Applied Informatics, University of Lodz

Keywords:

dermatological asymmetry of skin lesion, dermatological features segmentation, skin lesions classification, support vector machine, kNN, neural networks

Abstract

In the paper, a classification of the skin lesion asymmetry using different methods is discussed. We present shortly dermatological asymmetry measure in shape (DASMShape) and show its two implementations. Comparing the results of DASMShape measures and a lesion asymmetry given by the experts in PH2 dataset we achieved the best accuracy (83.2%) using SVM with RBF kernel function for the DASMShape. Although the NN results are lower by 9.5% it is always overestimating the asymmetry.

Downloads

Published

2018-05-19

Issue

2018: INFORMATION TECHNOLOGIES AND COMPUTER MODELLING proceedings of the International Scientific Conference

Section

Section 1 Information technologies in technical and special purpose systems, information technologies in society, education, medicine, economics, management, ecology and law