Computer Aided Deep Transfer Learning Based Skin Lesion Segmentation and Classification using Dermoscopic Images

  • Arulmurugan A., Hema Rajini N., Pradeep G.

Abstract

At present days, skin cancer becomesa widespread type of cancer, not only in the United States but all over the globe. Earlier identification of skin cancer is essential to reduce the mortality rate. Owing to the fast development in the number of skin cancer, there is an urgent requirement for computer aided diagnosis (CAD) model for skin lesion detection. Though several models have existed in the literature, deep learning (DL) based techniques have exhibited better performance on thedermoscopic skin lesion images, which does not require any hand crafted features. Therefore, this paper develops a DL based CAD model for dermoscopic skin lesion detection and classification. The presented model initially performs the preprocessing task to detect the region of interest (RoI) and remove noise that exist in the image. For segmentation, density based fuzzy c means (DFCM) technique is applied to identify the affected regions in the image. In addition, MobileNet model is employed as a feature extractor to derive a useful set of feature vectors. Finally, two classification models namely Gaussian naïve bayes (GNB) and support vector machine (SVM) models are appended at the final layer of MobileNet, to derive MN-GNB and MN-SVM techniques to identify and classify the different types of skin cancer. To verify the diagnostic outcome of the MN-GNB and MN-SVM models, an extensive set of simulations will take place on the benchmark ISIC dataset. The detailed experimental outcome ensured the betterment of the proposed models over the compared ones.

Published
2020-11-05
How to Cite
Arulmurugan A., Hema Rajini N., Pradeep G. (2020). Computer Aided Deep Transfer Learning Based Skin Lesion Segmentation and Classification using Dermoscopic Images. International Journal of Advanced Science and Technology, 29(04), 10644–10662. Retrieved from http://sersc.org/journals/index.php/IJAST/article/view/33577