Comperative assessment of taylor water cycle optimization (TWCO) -based deep residual network (DRN) for skin cancer detection using deep learning

Abstract

This paper devises a hybrid optimization driven deep technique for automated detection of skin cancer. Here, the pre-trained deep learning model is utilized for the skin cancer detection. The pre-processing of skin images is performed that helped to reduce the ill impacts and several artifacts such as hair that may be contained in the dermoscopy images. In addition, the DeepJoint model is used to perform segmentation in order to attain improved outcomes. The data augmentation helped to make the image suitable for improved processing helps to quantify the images for effective classification. The skin cancer detection is done with Deep Residual Network (DRN), which is trained using newly devised technique, namely Taylor Water Cycle Optimization (TWCO) algorithm. The proposed TWCO-based DRN outperformed with highest testing accuracy of 92.3%, True positive rate (TPR) of 93.5% and True negative rate (TNR) of 90.5%.