Detection and Classification of Skin Cancer Using Deep Convolutional Neural Networks (CNN) via KNIME Analytics Platform Software

Abstract

The use of technologies from many fields, such as mass spectrometry, next-generation sequencing, or image processing, is common in experiments in the life sciences. Complex scripts are frequently used to govern data flow, data transformation, and statistical analysis when passing data between such tools. Such scripts not only tend to be platform dependant, but also tend to expand as the experiment goes on and are rarely clearly documented, which makes the experiment harder to reproduce. Workflow systems like KNIME Analytics Platform, which offers a platform for graphically linking tools and ensures the same results across various operating systems, aim to address these issues. systems that are frequently employed in the biological sciences and describe how they compare and contrast with KNIME. KNIME is an open source program that enables programmers and scientists to share their own extensions with the scientific community. The unified data model of KNIME allows for interoperability, and we describe a few additions from the life sciences that make it easier to explore, analyze, and visualize data. In addition, we mention additional workflow.  According to the American Cancer Society, skin cancer is the most prevalent form of malignancy in humans. It is typically identified visually, with first clinical screenings, dermoscopic (skin-related) analysis, a biopsy, and histological examinations as potential follow-up steps. Errors (mutations) in the DNA of skin cells are the cause of skin cancer. The cells proliferate out of control and aggregate into a mass of cancer cells as a result of the mutations. In this paper, convolutional neural networks are used to attempt to categorize photos of skin lesions. The deep neural networks demonstrate enormous potential for classifying images while taking into account the extreme environmental heterogeneity. Due to the current state of technology, it is imperative to use machines rather than people to address the widespread problem of skin cancer. One of the best ways to address skin cancer issues is deep learning. Huge data, virtual reality, augmented reality, and mini-services are all used in the new research area of deep learning in contemporary technology. The advent of powerful arithmetic capabilities enabled deep learning applications using Mobile net (CNN) to revolutionize image classification. The various forms of skin cancer can be categorized using deep learning‪. Transfer Learning was used during the training on many models. The model's best level of accuracy was over 77.333 %. To guarantee the validity and reproducibility of the aforementioned result, the dataset employed is openly accessible.