Classification of Benign and Malignant Skin Lesions using Convolutional Neural Networks (CNN)

Skin cancer is one of the common forms of cancers, and if it is diagnosed early, then the possibility of survival increases greatly. Currently, the diagnosis is carried out based on expertise and visual inspection, which are cumbersome and prone to human errors. In the present paper use CNN for skin cancer detection using a deep learning approach. A large dataset of dermoscopic images of several categories of skin lesions is used as training data, which may contain both malignant and benign samples, for example, melanoma, basal cell carcinoma, squamous cell carcinoma, and seborrheic keratosis. In fact, the proposed CNN architecture extracts the relevant information of the images, such as texture, shapes, and patterns, to classify such skin lesions into any one of these pre-specified categories. The model will highly determine both benign and malignant lesions with the highest accuracy, thus providing an accurate machine for automated screening of skin cancer. Experimental results showed that it outperformed the techniques used for diagnostics in speed, accuracy, and scalable early detection of skin cancers. This would help in more informed decisions by the dermatologists and thus lead to better patient outcomes while burdening the least members of the healthcare providers. This paper aims at proving the possibility of using deep learning to improve and automate skin-cancer diagnosis. The objective of this work is to leverage deep learning to classify skin lesions into malignant (melanoma, basal cell carcinoma, squamous cell carcinoma) and benign (seborrheic keratosis) categories with high accuracy.