Detection of Benign/Malignant Skin Cancer with Deep Transfer Learning Using Fused Features
K. Vijayakumar, Mohammad Kanan, R. Appavu Raj, Bhukya Balakrishna, S. Prabha, N. Surya
Abstract
Skin is one of the large sensory organs in human physiology and infection in skin will lead to health issues. Skincancer is one of the severe skin issues, which needs early monitoring and treatment. Skin melanoma examination is a crucial procedure and detection of benign/malignant class is essential for treatment planning and management. This work proposes a deep-learning (DL) model based technique to analyse the skin-cancer images with better accuracy. The developed scheme consist the following stages; (i) labelled melanoma data collection and resizing, (ii) features extraction using DL-model and implementing SoftMax based classification, (iii) best model detection and developing fused-features (FF) using serial integration after 50% features reduction, and (iv) classification and 3-fold cross validation. This research considered 1500 images per category for the study. The obtained results with individual-features and FF along with SoftMax-classifier are verified. The outcome of this research confirms that the implemented study provided 97.67% accuracy with SoftMax classifier.