Title :
SVMs classification based on Insitu melanoma
Author :
Satheesha, T.Y. ; Satyanarayana, D. ; Giriprasad, M.N.
Author_Institution :
Dept. of ECE, NCET, Bangalore, India
Abstract :
This paper presents a comparative study of Support Vector Machines (SVMs) which is classified based on melanoma imaging technique. After the preprocessing and segmentation of a set of distinct 35 images, the extracted features were Asymmetry, Border, Color, Diameter,(ABCD) Entropy and Correlations respectively. Further the resultant data was fed into five different SVM classifiers namely linear, poly, quadratic, radial basic function and Multilayer Preceptor, which classifies different aspects of Insitu melanoma. Among these 35 images the most consistent result which was obtained stood unmatched at an accuracy of 77.77%. Finally each stage of pigmented skin lesion is represented by a histogram plot which further enriches the data quite significantly.
Keywords :
feature extraction; image classification; image representation; image segmentation; medical image processing; support vector machines; SVM classification; feature extraction; insitu melanoma; linear classifiers; melanoma imaging technique; multilayer preceptor; pigmented skin lesion; poly classifiers; quadratic classifiers; radial basic function; support vector machines; Accuracy; Cancer; Feature extraction; Image segmentation; Kernel; Malignant tumors; Support vector machines; ABCD; Histogram; Melanoma; SVM kernel;
Conference_Titel :
Circuits, Communication, Control and Computing (I4C), 2014 International Conference on
Print_ISBN :
978-1-4799-6545-8
DOI :
10.1109/CIMCA.2014.7057829
