Classification of skin cancer using deep learning techniques

• Main Authors: Srinivasa Rao, Perumalla, Satish Babu, Kaaparapu, Raju, M. C.
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial neural networks; Cancer; Datasets; Deep learning; Diagnosis; Evaluation; High speed cameras; Image classification; Lesions; Machine learning; Medical imaging; Skin cancer
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0212401

Early diagnosis and treatment outcomes for skin cancer depend on these factors. Recent studies have shown that deep learning models with dermoscopic images can classify skin cancer with good accuracy rates. In this article, we develop and evaluate deep-learning models for skin cancer classification using a large dataset of dermoscopic pictures. There are more than 200 different types of cancer. The first stage in the diagnosis of cancer is often clinical screening, which is then followed by dermoscopic examination and histological analysis (biopsy). Melanoma is one type of skin cancer that can be treated if found in its early stages. Using high-speed cameras, dermatologists take dermoscopic images of the skin lesions, which are subsequently evaluated. For the early identification of skin cancer, the use of suitable automated examination models can be highly advantageous. In this study, labels indicating the presence of one of six distinct types of skin cancer—were applied to over 30,000 dermoscopic images of skin lesions (vasc). The photos are preprocessed in order to standardize the intensities, resize them to a uniform size, and maybe enhance them in order to expand the dataset. For the categorization of skin cancer, we examine convolutional neural networks (CNNs) and transfer learning techniques. The models are tested on a different test set after being trained on a subset of the dataset to determine how accurate they are at classifying the various forms of skin cancer.