: Skin cancer is regarded as the cardinal cause of morbidity and mortality globally, with death count increasing at an alarming rate. It has a higher chance of being cured if diagnosis is done in its initial stages; proper diagnosis of skin cancer is crucial to enable proper treatments. Highly skilled dermatologists and skin specialist doctors are capable of accurately detecting skin cancer at an early stage. Expert dermatologists are limited in number, so systems that automatically detect the cancerous growth at early stage with high performance are a useful tool. So, this study presents a deep learning (DL) technique to classify images and detect skin cancer at an early stage. We have trained our model using images of harmless, that is, benign images and tumor-based images; we have used Convolutional neural network (CNN) on those images to classify whether the image is a suspect of skin cancer or not. This proposed approach achieves an accuracy of 86% and is compared to the DCNN model, which was introduced earlier, before our work. Also, an additional approach using the ResNet-50, a 50-layer deep CNN has been implemented which has proved useful in further improving the accuracy to over 90%.
Authors: Sidharth Purohit, Shubhra Suman, Avinash Kumar, Sobhangi Sarkar, Chittaranjan Pradhan, Jyotir Moy Chatterjee
DOI: https://doi.org/10.1515/9783110708127-009
Publish Year: 2021
