Skin Lesion Segmentation and Classification Model using Optimal Shannon Entropy with Artificial Neural Network

  • Dr. Artheeswari S., Dr. Kalaivany S.

Abstract

Skin lesion segmentation plays a crucial role in the earlier and precise diagnosis of skin cancer from dermoscopic images methods. But the automated segmentation and classification of skin lesion is a difficult process due to the presence of artefacts (hair, gel bubble, ruler marker), unclear boundaries, low contrast and varying sizes and shapes of the lesion images. This paper presents a new grasshopper optimization algorithm (GOA) based Shannon entropy for skin lesion segmentation and artificial neural network (ANN) based classification model called GOASE-ANN to detect the presence of lesion in the dermoscopic images. The GOASE-ANN initially performs preprocessing using Gabor Kuwahara Filter to remove the noise exist in the image. Then, the preprocessed image undergoes segmentationusingGOASE model, where the optimal threshold values of SE is determined by GOA. Next, the feature extraction process is carried out using GoogleNet model. Finally, the feature vectors are effectively classified into appropriate class labels using ANN. The performance validation of the GOASE-ANN model takes place using ISIC dataset and the experimental results pointed out the superior nature of the GOASE-ANN model over the compared methods.

Published
2020-03-30
How to Cite
Dr. Artheeswari S., Dr. Kalaivany S. (2020). Skin Lesion Segmentation and Classification Model using Optimal Shannon Entropy with Artificial Neural Network. International Journal of Advanced Science and Technology, 29(3), 13188 -. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/31518
Section
Articles