Haar Wavelet Pyramid-Based Melanoma Skin Cancer Identification With Ensemble of Machine Learning Algorithms

Background

Despite only accounting for 1% of skin tumors globally, malignant melanoma is the cause of 60% of deaths from skin cancer (Khazaei et al., 2019). This is primarily because melanoma has been found to be highly likely to expand to other areas of the human body (American Cancer Society, 2020). While early detection does prove incredibly successful in ensuring high survival rates, this can only be facilitated by the presence of medical professionals, a scarcity of whom is seen across the globe, including countries like the US (Pham, 2019), the UK (Taylor, 2020), India (Hazarika 2013). To overcome these challenges, computer-assisted health informatics can prove to be a promising source of support (Kareh and Thoumy, 2018) (Jain and Singh, 2020).

In health informatics based melanoma detection, the approaches proposed can be broadly classified into three categories as methods that use machine learning-based algorithms with crafted dermoscopy skin image features, deep learning-based methods with skin images, and DNA profiling based methods. (Leachman et al., 2016)